tighten

Reverts openai/codex#6485

.github/workflows/cla.yml

@@ -13,6 +13,9 @@ permissions:
 
 jobs:
   cla:
+    # Only run the CLA assistant for the canonical openai repo so forks are not blocked
+    # and contributors who signed previously do not receive duplicate CLA notifications.
+    if: ${{ github.repository_owner == 'openai' }}
     runs-on: ubuntu-latest
     steps:
       - uses: contributor-assistant/github-action@v2.6.1

.github/workflows/close-stale-contributor-prs.yml

@@ -0,0 +1,105 @@
+name: Close stale contributor PRs
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 6 * * *"
+
+permissions:
+  contents: read
+  issues: write
+  pull-requests: write
+
+jobs:
+  close-stale-contributor-prs:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Close inactive PRs from contributors
+        uses: actions/github-script@v7
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          script: |
+            const DAYS_INACTIVE = 14;
+            const cutoff = new Date(Date.now() - DAYS_INACTIVE * 24 * 60 * 60 * 1000);
+            const { owner, repo } = context.repo;
+            const dryRun = false;
+            const stalePrs = [];
+
+            core.info(`Dry run mode: ${dryRun}`);
+
+            const prs = await github.paginate(github.rest.pulls.list, {
+              owner,
+              repo,
+              state: "open",
+              per_page: 100,
+              sort: "updated",
+              direction: "asc",
+            });
+
+            for (const pr of prs) {
+              const lastUpdated = new Date(pr.updated_at);
+              if (lastUpdated > cutoff) {
+                core.info(`PR ${pr.number} is fresh`);
+                continue;
+              }
+
+              if (!pr.user || pr.user.type !== "User") {
+                core.info(`PR ${pr.number} wasn't created by a user`);
+                continue;
+              }
+
+              let permission;
+              try {
+                const permissionResponse = await github.rest.repos.getCollaboratorPermissionLevel({
+                  owner,
+                  repo,
+                  username: pr.user.login,
+                });
+                permission = permissionResponse.data.permission;
+              } catch (error) {
+                if (error.status === 404) {
+                  core.info(`Author ${pr.user.login} is not a collaborator; skipping #${pr.number}`);
+                  continue;
+                }
+                throw error;
+              }
+
+              const hasContributorAccess = ["admin", "maintain", "write"].includes(permission);
+              if (!hasContributorAccess) {
+                core.info(`Author ${pr.user.login} has ${permission} access; skipping #${pr.number}`);
+                continue;
+              }
+
+              stalePrs.push(pr);
+            }
+
+            if (!stalePrs.length) {
+              core.info("No stale contributor pull requests found.");
+              return;
+            }
+
+            for (const pr of stalePrs) {
+              const issue_number = pr.number;
+              const closeComment = `Closing this pull request because it has had no updates for more than ${DAYS_INACTIVE} days. If you plan to continue working on it, feel free to reopen or open a new PR.`;
+
+              if (dryRun) {
+                core.info(`[dry-run] Would close contributor PR #${issue_number} from ${pr.user.login}`);
+                continue;
+              }
+
+              await github.rest.issues.createComment({
+                owner,
+                repo,
+                issue_number,
+                body: closeComment,
+              });
+
+              await github.rest.pulls.update({
+                owner,
+                repo,
+                pull_number: issue_number,
+                state: "closed",
+              });
+
+              core.info(`Closed contributor PR #${issue_number} from ${pr.user.login}`);
+            }

.github/workflows/rust-ci.yml

@@ -95,8 +95,9 @@ jobs:
       run:
         working-directory: codex-rs
     env:
-      # Speed up repeated builds across CI runs by caching compiled objects.
-      RUSTC_WRAPPER: sccache
+      # Speed up repeated builds across CI runs by caching compiled objects (non-Windows).
+      USE_SCCACHE: ${{ startsWith(matrix.runner, 'windows') && 'false' || 'true' }}
+      RUSTC_WRAPPER: ${{ startsWith(matrix.runner, 'windows') && '' || 'sccache' }}
       CARGO_INCREMENTAL: "0"
       SCCACHE_CACHE_SIZE: 10G
 
@@ -170,12 +171,14 @@ jobs:
 
       # Install and restore sccache cache
       - name: Install sccache
+        if: ${{ env.USE_SCCACHE == 'true' }}
         uses: taiki-e/install-action@44c6d64aa62cd779e873306675c7a58e86d6d532 # v2
         with:
           tool: sccache
           version: 0.7.5
 
       - name: Configure sccache backend
+        if: ${{ env.USE_SCCACHE == 'true' }}
         shell: bash
         run: |
           set -euo pipefail
@@ -189,7 +192,7 @@ jobs:
           fi
 
       - name: Restore sccache cache (fallback)
-        if: ${{ env.SCCACHE_GHA_ENABLED != 'true' }}
+        if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
         id: cache_sccache_restore
         uses: actions/cache/restore@v4
         with:
@@ -274,7 +277,7 @@ jobs:
           key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ hashFiles('codex-rs/rust-toolchain.toml') }}
 
       - name: Save sccache cache (fallback)
-        if: always() && !cancelled() && env.SCCACHE_GHA_ENABLED != 'true'
+        if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
         continue-on-error: true
         uses: actions/cache/save@v4
         with:
@@ -282,12 +285,12 @@ jobs:
           key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ github.run_id }}
 
       - name: sccache stats
-        if: always()
+        if: always() && env.USE_SCCACHE == 'true'
         continue-on-error: true
         run: sccache --show-stats || true
 
       - name: sccache summary
-        if: always()
+        if: always() && env.USE_SCCACHE == 'true'
         shell: bash
         run: |
           {
@@ -326,7 +329,9 @@ jobs:
       run:
         working-directory: codex-rs
     env:
-      RUSTC_WRAPPER: sccache
+      # Speed up repeated builds across CI runs by caching compiled objects (non-Windows).
+      USE_SCCACHE: ${{ startsWith(matrix.runner, 'windows') && 'false' || 'true' }}
+      RUSTC_WRAPPER: ${{ startsWith(matrix.runner, 'windows') && '' || 'sccache' }}
       CARGO_INCREMENTAL: "0"
       SCCACHE_CACHE_SIZE: 10G
 
@@ -370,12 +375,14 @@ jobs:
             cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-
 
       - name: Install sccache
+        if: ${{ env.USE_SCCACHE == 'true' }}
         uses: taiki-e/install-action@44c6d64aa62cd779e873306675c7a58e86d6d532 # v2
         with:
           tool: sccache
           version: 0.7.5
 
       - name: Configure sccache backend
+        if: ${{ env.USE_SCCACHE == 'true' }}
         shell: bash
         run: |
           set -euo pipefail
@@ -389,7 +396,7 @@ jobs:
           fi
 
       - name: Restore sccache cache (fallback)
-        if: ${{ env.SCCACHE_GHA_ENABLED != 'true' }}
+        if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
         id: cache_sccache_restore
         uses: actions/cache/restore@v4
         with:
@@ -406,10 +413,10 @@ jobs:
 
       - name: tests
         id: test
-        continue-on-error: true
         run: cargo nextest run --all-features --no-fail-fast --target ${{ matrix.target }} --cargo-profile ci-test
         env:
           RUST_BACKTRACE: 1
+          NEXTEST_STATUS_LEVEL: leak
 
       - name: Save cargo home cache
         if: always() && !cancelled() && steps.cache_cargo_home_restore.outputs.cache-hit != 'true'
@@ -424,7 +431,7 @@ jobs:
           key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ hashFiles('codex-rs/rust-toolchain.toml') }}
 
       - name: Save sccache cache (fallback)
-        if: always() && !cancelled() && env.SCCACHE_GHA_ENABLED != 'true'
+        if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
         continue-on-error: true
         uses: actions/cache/save@v4
         with:
@@ -432,12 +439,12 @@ jobs:
           key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ github.run_id }}
 
       - name: sccache stats
-        if: always()
+        if: always() && env.USE_SCCACHE == 'true'
         continue-on-error: true
         run: sccache --show-stats || true
 
       - name: sccache summary
-        if: always()
+        if: always() && env.USE_SCCACHE == 'true'
         shell: bash
         run: |
           {

AGENTS.md

@@ -85,6 +85,7 @@ If you don’t have the tool:
 - `ResponsesRequest` exposes helpers (`body_json`, `input`, `function_call_output`, `custom_tool_call_output`, `call_output`, `header`, `path`, `query_param`) so assertions can target structured payloads instead of manual JSON digging.
 - Build SSE payloads with the provided `ev_*` constructors and the `sse(...)`.
 - Prefer `wait_for_event` over `wait_for_event_with_timeout`.
+- Prefer `mount_sse_once` over `mount_sse_once_match` or `mount_sse_sequence`
 
 - Typical pattern:
 

codex-rs/.config/nextest.toml

@@ -0,0 +1,9 @@
+[profile.default]
+# Do not increase, fix your test instead
+slow-timeout = { period = "15s", terminate-after = 2 }
+
+
+[[profile.default.overrides]]
+# Do not add new tests here
+filter = 'test(rmcp_client) | test(humanlike_typing_1000_chars_appears_live_no_placeholder)'
+slow-timeout = { period = "1m", terminate-after = 4 }

codex-rs/Cargo.lock

@@ -850,6 +850,7 @@ dependencies = [
  "codex-protocol",
  "codex-utils-json-to-toml",
  "core_test_support",
+ "mcp-types",
  "opentelemetry-appender-tracing",
  "os_info",
  "pretty_assertions",
@@ -883,6 +884,19 @@ dependencies = [
  "uuid",
 ]
 
+[[package]]
+name = "codex-app-server-test-client"
+version = "0.0.0"
+dependencies = [
+ "anyhow",
+ "clap",
+ "codex-app-server-protocol",
+ "codex-protocol",
+ "serde",
+ "serde_json",
+ "uuid",
+]
+
 [[package]]
 name = "codex-apply-patch"
 version = "0.0.0"
@@ -1048,6 +1062,7 @@ dependencies = [
  "codex-app-server-protocol",
  "codex-core",
  "codex-protocol",
+ "once_cell",
  "serde",
  "toml",
 ]
@@ -1114,6 +1129,7 @@ dependencies = [
  "similar",
  "strum_macros 0.27.2",
  "tempfile",
+ "test-case",
  "test-log",
  "thiserror 2.0.17",
  "time",
@@ -1548,9 +1564,11 @@ name = "codex-utils-tokenizer"
 version = "0.0.0"
 dependencies = [
  "anyhow",
+ "codex-utils-cache",
  "pretty_assertions",
  "thiserror 2.0.17",
  "tiktoken-rs",
+ "tokio",
 ]
 
 [[package]]
@@ -6160,6 +6178,39 @@ version = "0.5.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8f50febec83f5ee1df3015341d8bd429f2d1cc62bcba7ea2076759d315084683"
 
+[[package]]
+name = "test-case"
+version = "3.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eb2550dd13afcd286853192af8601920d959b14c401fcece38071d53bf0768a8"
+dependencies = [
+ "test-case-macros",
+]
+
+[[package]]
+name = "test-case-core"
+version = "3.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "adcb7fd841cd518e279be3d5a3eb0636409487998a4aff22f3de87b81e88384f"
+dependencies = [
+ "cfg-if",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.104",
+]
+
+[[package]]
+name = "test-case-macros"
+version = "3.3.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5c89e72a01ed4c579669add59014b9a524d609c0c88c6a585ce37485879f6ffb"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.104",
+ "test-case-core",
+]
+
 [[package]]
 name = "test-log"
 version = "0.2.18"
@@ -6267,9 +6318,9 @@ dependencies = [
 
 [[package]]
 name = "tiktoken-rs"
-version = "0.7.0"
+version = "0.9.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "25563eeba904d770acf527e8b370fe9a5547bacd20ff84a0b6c3bc41288e5625"
+checksum = "3a19830747d9034cd9da43a60eaa8e552dfda7712424aebf187b7a60126bae0d"
 dependencies = [
  "anyhow",
  "base64",

codex-rs/Cargo.toml

@@ -5,6 +5,7 @@ members = [
     "async-utils",
     "app-server",
     "app-server-protocol",
+    "app-server-test-client",
     "apply-patch",
     "arg0",
     "feedback",
@@ -122,8 +123,8 @@ eventsource-stream = "0.2.3"
 futures = { version = "0.3", default-features = false }
 http = "1.3.1"
 icu_decimal = "2.1"
-icu_provider = { version = "2.1", features = ["sync"] }
 icu_locale_core = "2.1"
+icu_provider = { version = "2.1", features = ["sync"] }
 ignore = "0.4.23"
 image = { version = "^0.25.8", default-features = false }
 indexmap = "2.12.0"
@@ -140,6 +141,7 @@ mime_guess = "2.0.5"
 multimap = "0.10.0"
 notify = "8.2.0"
 nucleo-matcher = "0.3.1"
+once_cell = "1"
 openssl-sys = "*"
 opentelemetry = "0.30.0"
 opentelemetry-appender-tracing = "0.30.0"
@@ -181,6 +183,7 @@ tempfile = "3.23.0"
 test-log = "0.2.18"
 textwrap = "0.16.2"
 thiserror = "2.0.17"
+tiktoken-rs = "0.9"
 time = "0.3"
 tiny_http = "0.12"
 tokio = "1"
@@ -278,8 +281,8 @@ opt-level = 0
 [patch.crates-io]
 # Uncomment to debug local changes.
 # ratatui = { path = "../../ratatui" }
-ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.29.0-patch" }
 crossterm = { git = "https://github.com/nornagon/crossterm", branch = "nornagon/color-query" }
+ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.29.0-patch" }
 
 # Uncomment to debug local changes.
 # rmcp = { path = "../../rust-sdk/crates/rmcp" }

codex-rs/app-server-protocol/src/export.rs

@@ -13,10 +13,7 @@ use crate::export_server_responses;
 use anyhow::Context;
 use anyhow::Result;
 use anyhow::anyhow;
-use codex_protocol::parse_command::ParsedCommand;
 use codex_protocol::protocol::EventMsg;
-use codex_protocol::protocol::FileChange;
-use codex_protocol::protocol::SandboxPolicy;
 use schemars::JsonSchema;
 use schemars::schema_for;
 use serde::Serialize;
@@ -120,10 +117,6 @@ pub fn generate_json(out_dir: &Path) -> Result<()> {
         |d| write_json_schema_with_return::<crate::ClientNotification>(d, "ClientNotification"),
         |d| write_json_schema_with_return::<crate::ServerNotification>(d, "ServerNotification"),
         |d| write_json_schema_with_return::<EventMsg>(d, "EventMsg"),
-        |d| write_json_schema_with_return::<FileChange>(d, "FileChange"),
-        |d| write_json_schema_with_return::<crate::protocol::v1::InputItem>(d, "InputItem"),
-        |d| write_json_schema_with_return::<ParsedCommand>(d, "ParsedCommand"),
-        |d| write_json_schema_with_return::<SandboxPolicy>(d, "SandboxPolicy"),
     ];
 
     let mut schemas: Vec<GeneratedSchema> = Vec::new();
@@ -152,13 +145,10 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
         "ClientNotification",
         "ClientRequest",
         "EventMsg",
-        "FileChange",
-        "InputItem",
-        "ParsedCommand",
-        "SandboxPolicy",
         "ServerNotification",
         "ServerRequest",
     ];
+    const IGNORED_DEFINITIONS: &[&str] = &["Option<()>"];
 
     let namespaced_types = collect_namespaced_types(&schemas);
     let mut definitions = Map::new();
@@ -171,6 +161,10 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
             in_v1_dir,
         } = schema;
 
+        if IGNORED_DEFINITIONS.contains(&logical_name.as_str()) {
+            continue;
+        }
+
         if let Some(ref ns) = namespace {
             rewrite_refs_to_namespace(&mut value, ns);
         }
@@ -181,6 +175,9 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
             && let Value::Object(defs_obj) = defs
         {
             for (def_name, mut def_schema) in defs_obj {
+                if IGNORED_DEFINITIONS.contains(&def_name.as_str()) {
+                    continue;
+                }
                 if SPECIAL_DEFINITIONS.contains(&def_name.as_str()) {
                     continue;
                 }
@@ -386,14 +383,6 @@ fn variant_definition_name(base: &str, variant: &Value) -> Option<String> {
             });
         }
 
-        if let Some(mode_literal) = literal_from_property(props, "mode") {
-            let pascal = to_pascal_case(mode_literal);
-            return Some(match base {
-                "SandboxPolicy" => format!("{pascal}SandboxPolicy"),
-                _ => format!("{pascal}{base}"),
-            });
-        }
-
         if props.len() == 1
             && let Some(key) = props.keys().next()
         {

codex-rs/app-server-protocol/src/protocol/common.rs

@@ -508,6 +508,9 @@ server_notification_definitions! {
     McpToolCallProgress => "item/mcpToolCall/progress" (v2::McpToolCallProgressNotification),
     AccountUpdated => "account/updated" (v2::AccountUpdatedNotification),
     AccountRateLimitsUpdated => "account/rateLimits/updated" (v2::AccountRateLimitsUpdatedNotification),
+    ReasoningSummaryTextDelta => "item/reasoning/summaryTextDelta" (v2::ReasoningSummaryTextDeltaNotification),
+    ReasoningSummaryPartAdded => "item/reasoning/summaryPartAdded" (v2::ReasoningSummaryPartAddedNotification),
+    ReasoningTextDelta => "item/reasoning/textDelta" (v2::ReasoningTextDeltaNotification),
 
     #[serde(rename = "account/login/completed")]
     #[ts(rename = "account/login/completed")]

codex-rs/app-server-protocol/src/protocol/v2.rs

@@ -8,6 +8,7 @@ use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::items::AgentMessageContent as CoreAgentMessageContent;
 use codex_protocol::items::TurnItem as CoreTurnItem;
+use codex_protocol::models::ResponseItem;
 use codex_protocol::protocol::RateLimitSnapshot as CoreRateLimitSnapshot;
 use codex_protocol::protocol::RateLimitWindow as CoreRateLimitWindow;
 use codex_protocol::user_input::UserInput as CoreUserInput;
@@ -56,8 +57,8 @@ v2_enum_from_core!(
 );
 
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
-#[serde(tag = "mode", rename_all = "camelCase")]
-#[ts(tag = "mode")]
+#[serde(tag = "type", rename_all = "camelCase")]
+#[ts(tag = "type")]
 #[ts(export_to = "v2/")]
 pub enum SandboxPolicy {
     DangerFullAccess,
@@ -290,11 +291,39 @@ pub struct ThreadStartResponse {
     pub thread: Thread,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[derive(Serialize, Deserialize, Debug, Default, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
+/// There are three ways to resume a thread:
+/// 1. By thread_id: load the thread from disk by thread_id and resume it.
+/// 2. By history: instantiate the thread from memory and resume it.
+/// 3. By path: load the thread from disk by path and resume it.
+///
+/// The precedence is: history > path > thread_id.
+/// If using history or path, the thread_id param will be ignored.
+///
+/// Prefer using thread_id whenever possible.
 pub struct ThreadResumeParams {
     pub thread_id: String,
+
+    /// [UNSTABLE] FOR CODEX CLOUD - DO NOT USE.
+    /// If specified, the thread will be resumed with the provided history
+    /// instead of loaded from disk.
+    pub history: Option<Vec<ResponseItem>>,
+
+    /// [UNSTABLE] Specify the rollout path to resume from.
+    /// If specified, the thread_id param will be ignored.
+    pub path: Option<PathBuf>,
+
+    /// Configuration overrides for the resumed thread, if any.
+    pub model: Option<String>,
+    pub model_provider: Option<String>,
+    pub cwd: Option<String>,
+    pub approval_policy: Option<AskForApproval>,
+    pub sandbox: Option<SandboxMode>,
+    pub config: Option<HashMap<String, serde_json::Value>>,
+    pub base_instructions: Option<String>,
+    pub developer_instructions: Option<String>,
 }
 
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -361,6 +390,8 @@ pub struct Thread {
     pub model_provider: String,
     /// Unix timestamp (in seconds) when the thread was created.
     pub created_at: i64,
+    /// [UNSTABLE] Path to the thread on disk.
+    pub path: PathBuf,
 }
 
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -485,7 +516,10 @@ pub enum ThreadItem {
     },
     Reasoning {
         id: String,
-        text: String,
+        #[serde(default)]
+        summary: Vec<String>,
+        #[serde(default)]
+        content: Vec<String>,
     },
     CommandExecution {
         id: String,
@@ -544,17 +578,11 @@ impl From<CoreTurnItem> for ThreadItem {
                     .collect::<String>();
                 ThreadItem::AgentMessage { id: agent.id, text }
             }
-            CoreTurnItem::Reasoning(reasoning) => {
-                let text = if !reasoning.summary_text.is_empty() {
-                    reasoning.summary_text.join("\n")
-                } else {
-                    reasoning.raw_content.join("\n")
-                };
-                ThreadItem::Reasoning {
-                    id: reasoning.id,
-                    text,
-                }
-            }
+            CoreTurnItem::Reasoning(reasoning) => ThreadItem::Reasoning {
+                id: reasoning.id,
+                summary: reasoning.summary_text,
+                content: reasoning.raw_content,
+            },
             CoreTurnItem::WebSearch(search) => ThreadItem::WebSearch {
                 id: search.id,
                 query: search.query,
@@ -612,7 +640,7 @@ pub enum McpToolCallStatus {
 #[ts(export_to = "v2/")]
 pub struct McpToolCallResult {
     pub content: Vec<McpContentBlock>,
-    pub structured_content: JsonValue,
+    pub structured_content: Option<JsonValue>,
 }
 
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -688,6 +716,32 @@ pub struct AgentMessageDeltaNotification {
     pub delta: String,
 }
 
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ReasoningSummaryTextDeltaNotification {
+    pub item_id: String,
+    pub delta: String,
+    pub summary_index: i64,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ReasoningSummaryPartAddedNotification {
+    pub item_id: String,
+    pub summary_index: i64,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ReasoningTextDeltaNotification {
+    pub item_id: String,
+    pub delta: String,
+    pub content_index: i64,
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
@@ -836,7 +890,8 @@ mod tests {
             ThreadItem::from(reasoning_item),
             ThreadItem::Reasoning {
                 id: "reasoning-1".to_string(),
-                text: "line one\nline two".to_string(),
+                summary: vec!["line one".to_string(), "line two".to_string()],
+                content: vec![],
             }
         );
 

codex-rs/app-server-test-client/Cargo.toml

@@ -0,0 +1,16 @@
+[package]
+name = "codex-app-server-test-client"
+version = { workspace = true }
+edition = "2024"
+
+[lints]
+workspace = true
+
+[dependencies]
+anyhow = { workspace = true }
+clap = { workspace = true, features = ["derive", "env"] }
+codex-app-server-protocol = { workspace = true }
+codex-protocol = { workspace = true }
+serde = { workspace = true, features = ["derive"] }
+serde_json = { workspace = true }
+uuid = { workspace = true, features = ["v4"] }

codex-rs/app-server-test-client/README.md

@@ -0,0 +1,2 @@
+# App Server Test Client
+Exercises simple `codex app-server` flows end-to-end, logging JSON-RPC messages sent between client and server to stdout.

codex-rs/app-server-test-client/src/main.rs

@@ -0,0 +1,691 @@
+use std::collections::VecDeque;
+use std::io::BufRead;
+use std::io::BufReader;
+use std::io::Write;
+use std::process::Child;
+use std::process::ChildStdin;
+use std::process::ChildStdout;
+use std::process::Command;
+use std::process::Stdio;
+use std::thread;
+use std::time::Duration;
+
+use anyhow::Context;
+use anyhow::Result;
+use anyhow::bail;
+use clap::Parser;
+use clap::Subcommand;
+use codex_app_server_protocol::AddConversationListenerParams;
+use codex_app_server_protocol::AddConversationSubscriptionResponse;
+use codex_app_server_protocol::AskForApproval;
+use codex_app_server_protocol::ClientInfo;
+use codex_app_server_protocol::ClientRequest;
+use codex_app_server_protocol::GetAccountRateLimitsResponse;
+use codex_app_server_protocol::InitializeParams;
+use codex_app_server_protocol::InitializeResponse;
+use codex_app_server_protocol::InputItem;
+use codex_app_server_protocol::JSONRPCMessage;
+use codex_app_server_protocol::JSONRPCNotification;
+use codex_app_server_protocol::JSONRPCResponse;
+use codex_app_server_protocol::LoginChatGptCompleteNotification;
+use codex_app_server_protocol::LoginChatGptResponse;
+use codex_app_server_protocol::NewConversationParams;
+use codex_app_server_protocol::NewConversationResponse;
+use codex_app_server_protocol::RequestId;
+use codex_app_server_protocol::SandboxPolicy;
+use codex_app_server_protocol::SendUserMessageParams;
+use codex_app_server_protocol::SendUserMessageResponse;
+use codex_app_server_protocol::ServerNotification;
+use codex_app_server_protocol::ThreadStartParams;
+use codex_app_server_protocol::ThreadStartResponse;
+use codex_app_server_protocol::TurnStartParams;
+use codex_app_server_protocol::TurnStartResponse;
+use codex_app_server_protocol::UserInput as V2UserInput;
+use codex_protocol::ConversationId;
+use codex_protocol::protocol::Event;
+use codex_protocol::protocol::EventMsg;
+use serde::de::DeserializeOwned;
+use serde_json::Value;
+use uuid::Uuid;
+
+/// Minimal launcher that initializes the Codex app-server and logs the handshake.
+#[derive(Parser)]
+#[command(author = "Codex", version, about = "Bootstrap Codex app-server", long_about = None)]
+struct Cli {
+    /// Path to the `codex` CLI binary.
+    #[arg(long, env = "CODEX_BIN", default_value = "codex")]
+    codex_bin: String,
+
+    #[command(subcommand)]
+    command: CliCommand,
+}
+
+#[derive(Subcommand)]
+enum CliCommand {
+    /// Send a user message through the Codex app-server.
+    SendMessage {
+        /// User message to send to Codex.
+        #[arg()]
+        user_message: String,
+    },
+    /// Send a user message through the app-server V2 thread/turn APIs.
+    SendMessageV2 {
+        /// User message to send to Codex.
+        #[arg()]
+        user_message: String,
+    },
+    /// Start a V2 turn that elicits an ExecCommand approval.
+    #[command(name = "trigger-cmd-approval")]
+    TriggerCmdApproval {
+        /// Optional prompt; defaults to a simple python command.
+        #[arg()]
+        user_message: Option<String>,
+    },
+    /// Start a V2 turn that elicits an ApplyPatch approval.
+    #[command(name = "trigger-patch-approval")]
+    TriggerPatchApproval {
+        /// Optional prompt; defaults to creating a file via apply_patch.
+        #[arg()]
+        user_message: Option<String>,
+    },
+    /// Start a V2 turn that should not elicit an ExecCommand approval.
+    #[command(name = "no-trigger-cmd-approval")]
+    NoTriggerCmdApproval,
+    /// Trigger the ChatGPT login flow and wait for completion.
+    TestLogin,
+    /// Fetch the current account rate limits from the Codex app-server.
+    GetAccountRateLimits,
+}
+
+fn main() -> Result<()> {
+    let Cli { codex_bin, command } = Cli::parse();
+
+    match command {
+        CliCommand::SendMessage { user_message } => send_message(codex_bin, user_message),
+        CliCommand::SendMessageV2 { user_message } => send_message_v2(codex_bin, user_message),
+        CliCommand::TriggerCmdApproval { user_message } => {
+            trigger_cmd_approval(codex_bin, user_message)
+        }
+        CliCommand::TriggerPatchApproval { user_message } => {
+            trigger_patch_approval(codex_bin, user_message)
+        }
+        CliCommand::NoTriggerCmdApproval => no_trigger_cmd_approval(codex_bin),
+        CliCommand::TestLogin => test_login(codex_bin),
+        CliCommand::GetAccountRateLimits => get_account_rate_limits(codex_bin),
+    }
+}
+
+fn send_message(codex_bin: String, user_message: String) -> Result<()> {
+    let mut client = CodexClient::spawn(codex_bin)?;
+
+    let initialize = client.initialize()?;
+    println!("< initialize response: {initialize:?}");
+
+    let conversation = client.new_conversation()?;
+    println!("< newConversation response: {conversation:?}");
+
+    let subscription = client.add_conversation_listener(&conversation.conversation_id)?;
+    println!("< addConversationListener response: {subscription:?}");
+
+    let send_response = client.send_user_message(&conversation.conversation_id, &user_message)?;
+    println!("< sendUserMessage response: {send_response:?}");
+
+    client.stream_conversation(&conversation.conversation_id)?;
+
+    client.remove_conversation_listener(subscription.subscription_id)?;
+
+    Ok(())
+}
+
+fn send_message_v2(codex_bin: String, user_message: String) -> Result<()> {
+    send_message_v2_with_policies(codex_bin, user_message, None, None)
+}
+
+fn trigger_cmd_approval(codex_bin: String, user_message: Option<String>) -> Result<()> {
+    let default_prompt =
+        "Run `touch /tmp/should-trigger-approval` so I can confirm the file exists.";
+    let message = user_message.unwrap_or_else(|| default_prompt.to_string());
+    send_message_v2_with_policies(
+        codex_bin,
+        message,
+        Some(AskForApproval::OnRequest),
+        Some(SandboxPolicy::ReadOnly),
+    )
+}
+
+fn trigger_patch_approval(codex_bin: String, user_message: Option<String>) -> Result<()> {
+    let default_prompt =
+        "Create a file named APPROVAL_DEMO.txt containing a short hello message using apply_patch.";
+    let message = user_message.unwrap_or_else(|| default_prompt.to_string());
+    send_message_v2_with_policies(
+        codex_bin,
+        message,
+        Some(AskForApproval::OnRequest),
+        Some(SandboxPolicy::ReadOnly),
+    )
+}
+
+fn no_trigger_cmd_approval(codex_bin: String) -> Result<()> {
+    let prompt = "Run `touch should_not_trigger_approval.txt`";
+    send_message_v2_with_policies(codex_bin, prompt.to_string(), None, None)
+}
+
+fn send_message_v2_with_policies(
+    codex_bin: String,
+    user_message: String,
+    approval_policy: Option<AskForApproval>,
+    sandbox_policy: Option<SandboxPolicy>,
+) -> Result<()> {
+    let mut client = CodexClient::spawn(codex_bin)?;
+
+    let initialize = client.initialize()?;
+    println!("< initialize response: {initialize:?}");
+
+    let thread_response = client.thread_start(ThreadStartParams::default())?;
+    println!("< thread/start response: {thread_response:?}");
+    let mut turn_params = TurnStartParams {
+        thread_id: thread_response.thread.id.clone(),
+        input: vec![V2UserInput::Text { text: user_message }],
+        ..Default::default()
+    };
+    turn_params.approval_policy = approval_policy;
+    turn_params.sandbox_policy = sandbox_policy;
+
+    let turn_response = client.turn_start(turn_params)?;
+    println!("< turn/start response: {turn_response:?}");
+
+    client.stream_turn(&thread_response.thread.id, &turn_response.turn.id)?;
+
+    Ok(())
+}
+
+fn test_login(codex_bin: String) -> Result<()> {
+    let mut client = CodexClient::spawn(codex_bin)?;
+
+    let initialize = client.initialize()?;
+    println!("< initialize response: {initialize:?}");
+
+    let login_response = client.login_chat_gpt()?;
+    println!("< loginChatGpt response: {login_response:?}");
+    println!(
+        "Open the following URL in your browser to continue:\n{}",
+        login_response.auth_url
+    );
+
+    let completion = client.wait_for_login_completion(&login_response.login_id)?;
+    println!("< loginChatGptComplete notification: {completion:?}");
+
+    if completion.success {
+        println!("Login succeeded.");
+        Ok(())
+    } else {
+        bail!(
+            "login failed: {}",
+            completion
+                .error
+                .as_deref()
+                .unwrap_or("unknown error from loginChatGptComplete")
+        );
+    }
+}
+
+fn get_account_rate_limits(codex_bin: String) -> Result<()> {
+    let mut client = CodexClient::spawn(codex_bin)?;
+
+    let initialize = client.initialize()?;
+    println!("< initialize response: {initialize:?}");
+
+    let response = client.get_account_rate_limits()?;
+    println!("< account/rateLimits/read response: {response:?}");
+
+    Ok(())
+}
+
+struct CodexClient {
+    child: Child,
+    stdin: Option<ChildStdin>,
+    stdout: BufReader<ChildStdout>,
+    pending_notifications: VecDeque<JSONRPCNotification>,
+}
+
+impl CodexClient {
+    fn spawn(codex_bin: String) -> Result<Self> {
+        let mut codex_app_server = Command::new(&codex_bin)
+            .arg("app-server")
+            .stdin(Stdio::piped())
+            .stdout(Stdio::piped())
+            .stderr(Stdio::inherit())
+            .spawn()
+            .with_context(|| format!("failed to start `{codex_bin}` app-server"))?;
+
+        let stdin = codex_app_server
+            .stdin
+            .take()
+            .context("codex app-server stdin unavailable")?;
+        let stdout = codex_app_server
+            .stdout
+            .take()
+            .context("codex app-server stdout unavailable")?;
+
+        Ok(Self {
+            child: codex_app_server,
+            stdin: Some(stdin),
+            stdout: BufReader::new(stdout),
+            pending_notifications: VecDeque::new(),
+        })
+    }
+
+    fn initialize(&mut self) -> Result<InitializeResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::Initialize {
+            request_id: request_id.clone(),
+            params: InitializeParams {
+                client_info: ClientInfo {
+                    name: "codex-toy-app-server".to_string(),
+                    title: Some("Codex Toy App Server".to_string()),
+                    version: env!("CARGO_PKG_VERSION").to_string(),
+                },
+            },
+        };
+
+        self.send_request(request, request_id, "initialize")
+    }
+
+    fn new_conversation(&mut self) -> Result<NewConversationResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::NewConversation {
+            request_id: request_id.clone(),
+            params: NewConversationParams::default(),
+        };
+
+        self.send_request(request, request_id, "newConversation")
+    }
+
+    fn add_conversation_listener(
+        &mut self,
+        conversation_id: &ConversationId,
+    ) -> Result<AddConversationSubscriptionResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::AddConversationListener {
+            request_id: request_id.clone(),
+            params: AddConversationListenerParams {
+                conversation_id: *conversation_id,
+                experimental_raw_events: false,
+            },
+        };
+
+        self.send_request(request, request_id, "addConversationListener")
+    }
+
+    fn remove_conversation_listener(&mut self, subscription_id: Uuid) -> Result<()> {
+        let request_id = self.request_id();
+        let request = ClientRequest::RemoveConversationListener {
+            request_id: request_id.clone(),
+            params: codex_app_server_protocol::RemoveConversationListenerParams { subscription_id },
+        };
+
+        self.send_request::<codex_app_server_protocol::RemoveConversationSubscriptionResponse>(
+            request,
+            request_id,
+            "removeConversationListener",
+        )?;
+
+        Ok(())
+    }
+
+    fn send_user_message(
+        &mut self,
+        conversation_id: &ConversationId,
+        message: &str,
+    ) -> Result<SendUserMessageResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::SendUserMessage {
+            request_id: request_id.clone(),
+            params: SendUserMessageParams {
+                conversation_id: *conversation_id,
+                items: vec![InputItem::Text {
+                    text: message.to_string(),
+                }],
+            },
+        };
+
+        self.send_request(request, request_id, "sendUserMessage")
+    }
+
+    fn thread_start(&mut self, params: ThreadStartParams) -> Result<ThreadStartResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::ThreadStart {
+            request_id: request_id.clone(),
+            params,
+        };
+
+        self.send_request(request, request_id, "thread/start")
+    }
+
+    fn turn_start(&mut self, params: TurnStartParams) -> Result<TurnStartResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::TurnStart {
+            request_id: request_id.clone(),
+            params,
+        };
+
+        self.send_request(request, request_id, "turn/start")
+    }
+
+    fn login_chat_gpt(&mut self) -> Result<LoginChatGptResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::LoginChatGpt {
+            request_id: request_id.clone(),
+            params: None,
+        };
+
+        self.send_request(request, request_id, "loginChatGpt")
+    }
+
+    fn get_account_rate_limits(&mut self) -> Result<GetAccountRateLimitsResponse> {
+        let request_id = self.request_id();
+        let request = ClientRequest::GetAccountRateLimits {
+            request_id: request_id.clone(),
+            params: None,
+        };
+
+        self.send_request(request, request_id, "account/rateLimits/read")
+    }
+
+    fn stream_conversation(&mut self, conversation_id: &ConversationId) -> Result<()> {
+        loop {
+            let notification = self.next_notification()?;
+
+            if !notification.method.starts_with("codex/event/") {
+                continue;
+            }
+
+            if let Some(event) = self.extract_event(notification, conversation_id)? {
+                match &event.msg {
+                    EventMsg::AgentMessage(event) => {
+                        println!("{}", event.message);
+                    }
+                    EventMsg::AgentMessageDelta(event) => {
+                        print!("{}", event.delta);
+                        std::io::stdout().flush().ok();
+                    }
+                    EventMsg::TaskComplete(event) => {
+                        println!("\n[task complete: {event:?}]");
+                        break;
+                    }
+                    EventMsg::TurnAborted(event) => {
+                        println!("\n[turn aborted: {:?}]", event.reason);
+                        break;
+                    }
+                    EventMsg::Error(event) => {
+                        println!("[error] {event:?}");
+                    }
+                    _ => {
+                        println!("[UNKNOWN EVENT] {:?}", event.msg);
+                    }
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    fn wait_for_login_completion(
+        &mut self,
+        expected_login_id: &Uuid,
+    ) -> Result<LoginChatGptCompleteNotification> {
+        loop {
+            let notification = self.next_notification()?;
+
+            if let Ok(server_notification) = ServerNotification::try_from(notification) {
+                match server_notification {
+                    ServerNotification::LoginChatGptComplete(completion) => {
+                        if &completion.login_id == expected_login_id {
+                            return Ok(completion);
+                        }
+
+                        println!(
+                            "[ignoring loginChatGptComplete for unexpected login_id: {}]",
+                            completion.login_id
+                        );
+                    }
+                    ServerNotification::AuthStatusChange(status) => {
+                        println!("< authStatusChange notification: {status:?}");
+                    }
+                    ServerNotification::AccountRateLimitsUpdated(snapshot) => {
+                        println!("< accountRateLimitsUpdated notification: {snapshot:?}");
+                    }
+                    ServerNotification::SessionConfigured(_) => {
+                        // SessionConfigured notifications are unrelated to login; skip.
+                    }
+                    _ => {}
+                }
+            }
+
+            // Not a server notification (likely a conversation event); keep waiting.
+        }
+    }
+
+    fn stream_turn(&mut self, thread_id: &str, turn_id: &str) -> Result<()> {
+        loop {
+            let notification = self.next_notification()?;
+
+            let Ok(server_notification) = ServerNotification::try_from(notification) else {
+                continue;
+            };
+
+            match server_notification {
+                ServerNotification::ThreadStarted(payload) => {
+                    if payload.thread.id == thread_id {
+                        println!("< thread/started notification: {:?}", payload.thread);
+                    }
+                }
+                ServerNotification::TurnStarted(payload) => {
+                    if payload.turn.id == turn_id {
+                        println!("< turn/started notification: {:?}", payload.turn.status);
+                    }
+                }
+                ServerNotification::AgentMessageDelta(delta) => {
+                    print!("{}", delta.delta);
+                    std::io::stdout().flush().ok();
+                }
+                ServerNotification::CommandExecutionOutputDelta(delta) => {
+                    print!("{}", delta.delta);
+                    std::io::stdout().flush().ok();
+                }
+                ServerNotification::ItemStarted(payload) => {
+                    println!("\n< item started: {:?}", payload.item);
+                }
+                ServerNotification::ItemCompleted(payload) => {
+                    println!("< item completed: {:?}", payload.item);
+                }
+                ServerNotification::TurnCompleted(payload) => {
+                    if payload.turn.id == turn_id {
+                        println!("\n< turn/completed notification: {:?}", payload.turn.status);
+                        if let Some(error) = payload.turn.error {
+                            println!("[turn error] {}", error.message);
+                        }
+                        println!("< usage: {:?}", payload.usage);
+                        break;
+                    }
+                }
+                ServerNotification::McpToolCallProgress(payload) => {
+                    println!("< MCP tool progress: {}", payload.message);
+                }
+                _ => {
+                    println!("[UNKNOWN SERVER NOTIFICATION] {server_notification:?}");
+                }
+            }
+        }
+
+        Ok(())
+    }
+
+    fn extract_event(
+        &self,
+        notification: JSONRPCNotification,
+        conversation_id: &ConversationId,
+    ) -> Result<Option<Event>> {
+        let params = notification
+            .params
+            .context("event notification missing params")?;
+
+        let mut map = match params {
+            Value::Object(map) => map,
+            other => bail!("unexpected params shape: {other:?}"),
+        };
+
+        let conversation_value = map
+            .remove("conversationId")
+            .context("event missing conversationId")?;
+        let notification_conversation: ConversationId = serde_json::from_value(conversation_value)
+            .context("conversationId was not a valid UUID")?;
+
+        if &notification_conversation != conversation_id {
+            return Ok(None);
+        }
+
+        let event_value = Value::Object(map);
+        let event: Event =
+            serde_json::from_value(event_value).context("failed to decode event payload")?;
+        Ok(Some(event))
+    }
+
+    fn send_request<T>(
+        &mut self,
+        request: ClientRequest,
+        request_id: RequestId,
+        method: &str,
+    ) -> Result<T>
+    where
+        T: DeserializeOwned,
+    {
+        self.write_request(&request)?;
+        self.wait_for_response(request_id, method)
+    }
+
+    fn write_request(&mut self, request: &ClientRequest) -> Result<()> {
+        let request_json = serde_json::to_string(request)?;
+        let request_pretty = serde_json::to_string_pretty(request)?;
+        print_multiline_with_prefix("> ", &request_pretty);
+
+        if let Some(stdin) = self.stdin.as_mut() {
+            writeln!(stdin, "{request_json}")?;
+            stdin
+                .flush()
+                .context("failed to flush request to codex app-server")?;
+        } else {
+            bail!("codex app-server stdin closed");
+        }
+
+        Ok(())
+    }
+
+    fn wait_for_response<T>(&mut self, request_id: RequestId, method: &str) -> Result<T>
+    where
+        T: DeserializeOwned,
+    {
+        loop {
+            let message = self.read_jsonrpc_message()?;
+
+            match message {
+                JSONRPCMessage::Response(JSONRPCResponse { id, result }) => {
+                    if id == request_id {
+                        return serde_json::from_value(result)
+                            .with_context(|| format!("{method} response missing payload"));
+                    }
+                }
+                JSONRPCMessage::Error(err) => {
+                    if err.id == request_id {
+                        bail!("{method} failed: {err:?}");
+                    }
+                }
+                JSONRPCMessage::Notification(notification) => {
+                    self.pending_notifications.push_back(notification);
+                }
+                JSONRPCMessage::Request(_) => {
+                    bail!("unexpected request from codex app-server");
+                }
+            }
+        }
+    }
+
+    fn next_notification(&mut self) -> Result<JSONRPCNotification> {
+        if let Some(notification) = self.pending_notifications.pop_front() {
+            return Ok(notification);
+        }
+
+        loop {
+            let message = self.read_jsonrpc_message()?;
+
+            match message {
+                JSONRPCMessage::Notification(notification) => return Ok(notification),
+                JSONRPCMessage::Response(_) | JSONRPCMessage::Error(_) => {
+                    // No outstanding requests, so ignore stray responses/errors for now.
+                    continue;
+                }
+                JSONRPCMessage::Request(_) => {
+                    bail!("unexpected request from codex app-server");
+                }
+            }
+        }
+    }
+
+    fn read_jsonrpc_message(&mut self) -> Result<JSONRPCMessage> {
+        loop {
+            let mut response_line = String::new();
+            let bytes = self
+                .stdout
+                .read_line(&mut response_line)
+                .context("failed to read from codex app-server")?;
+
+            if bytes == 0 {
+                bail!("codex app-server closed stdout");
+            }
+
+            let trimmed = response_line.trim();
+            if trimmed.is_empty() {
+                continue;
+            }
+
+            let parsed: Value =
+                serde_json::from_str(trimmed).context("response was not valid JSON-RPC")?;
+            let pretty = serde_json::to_string_pretty(&parsed)?;
+            print_multiline_with_prefix("< ", &pretty);
+            let message: JSONRPCMessage = serde_json::from_value(parsed)
+                .context("response was not a valid JSON-RPC message")?;
+            return Ok(message);
+        }
+    }
+
+    fn request_id(&self) -> RequestId {
+        RequestId::String(Uuid::new_v4().to_string())
+    }
+}
+
+fn print_multiline_with_prefix(prefix: &str, payload: &str) {
+    for line in payload.lines() {
+        println!("{prefix}{line}");
+    }
+}
+
+impl Drop for CodexClient {
+    fn drop(&mut self) {
+        let _ = self.stdin.take();
+
+        if let Ok(Some(status)) = self.child.try_wait() {
+            println!("[codex app-server exited: {status}]");
+            return;
+        }
+
+        thread::sleep(Duration::from_millis(100));
+
+        if let Ok(Some(status)) = self.child.try_wait() {
+            println!("[codex app-server exited: {status}]");
+            return;
+        }
+
+        let _ = self.child.kill();
+        let _ = self.child.wait();
+    }
+}

codex-rs/app-server/Cargo.toml

@@ -46,6 +46,7 @@ app_test_support = { workspace = true }
 assert_cmd = { workspace = true }
 base64 = { workspace = true }
 core_test_support = { workspace = true }
+mcp-types = { workspace = true }
 os_info = { workspace = true }
 pretty_assertions = { workspace = true }
 serial_test = { workspace = true }

codex-rs/app-server/src/bespoke_event_handling.rs

@@ -0,0 +1,483 @@
+use crate::codex_message_processor::ApiVersion;
+use crate::codex_message_processor::PendingInterrupts;
+use crate::outgoing_message::OutgoingMessageSender;
+use codex_app_server_protocol::AccountRateLimitsUpdatedNotification;
+use codex_app_server_protocol::AgentMessageDeltaNotification;
+use codex_app_server_protocol::ApplyPatchApprovalParams;
+use codex_app_server_protocol::ApplyPatchApprovalResponse;
+use codex_app_server_protocol::ExecCommandApprovalParams;
+use codex_app_server_protocol::ExecCommandApprovalResponse;
+use codex_app_server_protocol::InterruptConversationResponse;
+use codex_app_server_protocol::ItemCompletedNotification;
+use codex_app_server_protocol::ItemStartedNotification;
+use codex_app_server_protocol::McpToolCallError;
+use codex_app_server_protocol::McpToolCallResult;
+use codex_app_server_protocol::McpToolCallStatus;
+use codex_app_server_protocol::ReasoningSummaryPartAddedNotification;
+use codex_app_server_protocol::ReasoningSummaryTextDeltaNotification;
+use codex_app_server_protocol::ReasoningTextDeltaNotification;
+use codex_app_server_protocol::ServerNotification;
+use codex_app_server_protocol::ServerRequestPayload;
+use codex_app_server_protocol::ThreadItem;
+use codex_app_server_protocol::TurnInterruptResponse;
+use codex_core::CodexConversation;
+use codex_core::protocol::ApplyPatchApprovalRequestEvent;
+use codex_core::protocol::Event;
+use codex_core::protocol::EventMsg;
+use codex_core::protocol::ExecApprovalRequestEvent;
+use codex_core::protocol::McpToolCallBeginEvent;
+use codex_core::protocol::McpToolCallEndEvent;
+use codex_core::protocol::Op;
+use codex_core::protocol::ReviewDecision;
+use codex_protocol::ConversationId;
+use std::sync::Arc;
+use tokio::sync::oneshot;
+use tracing::error;
+
+type JsonRpcResult = serde_json::Value;
+
+pub(crate) async fn apply_bespoke_event_handling(
+    event: Event,
+    conversation_id: ConversationId,
+    conversation: Arc<CodexConversation>,
+    outgoing: Arc<OutgoingMessageSender>,
+    pending_interrupts: PendingInterrupts,
+) {
+    let Event { id: event_id, msg } = event;
+    match msg {
+        EventMsg::ApplyPatchApprovalRequest(ApplyPatchApprovalRequestEvent {
+            call_id,
+            changes,
+            reason,
+            grant_root,
+        }) => {
+            let params = ApplyPatchApprovalParams {
+                conversation_id,
+                call_id,
+                file_changes: changes,
+                reason,
+                grant_root,
+            };
+            let rx = outgoing
+                .send_request(ServerRequestPayload::ApplyPatchApproval(params))
+                .await;
+            // TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
+            tokio::spawn(async move {
+                on_patch_approval_response(event_id, rx, conversation).await;
+            });
+        }
+        // TODO(celia): properly construct McpToolCall TurnItem in core.
+        EventMsg::McpToolCallBegin(begin_event) => {
+            let notification = construct_mcp_tool_call_notification(begin_event).await;
+            outgoing
+                .send_server_notification(ServerNotification::ItemStarted(notification))
+                .await;
+        }
+        EventMsg::McpToolCallEnd(end_event) => {
+            let notification = construct_mcp_tool_call_end_notification(end_event).await;
+            outgoing
+                .send_server_notification(ServerNotification::ItemCompleted(notification))
+                .await;
+        }
+        EventMsg::AgentMessageContentDelta(event) => {
+            let notification = AgentMessageDeltaNotification {
+                item_id: event.item_id,
+                delta: event.delta,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::AgentMessageDelta(notification))
+                .await;
+        }
+        EventMsg::ReasoningContentDelta(event) => {
+            let notification = ReasoningSummaryTextDeltaNotification {
+                item_id: event.item_id,
+                delta: event.delta,
+                summary_index: event.summary_index,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ReasoningSummaryTextDelta(
+                    notification,
+                ))
+                .await;
+        }
+        EventMsg::ReasoningRawContentDelta(event) => {
+            let notification = ReasoningTextDeltaNotification {
+                item_id: event.item_id,
+                delta: event.delta,
+                content_index: event.content_index,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ReasoningTextDelta(notification))
+                .await;
+        }
+        EventMsg::AgentReasoningSectionBreak(event) => {
+            let notification = ReasoningSummaryPartAddedNotification {
+                item_id: event.item_id,
+                summary_index: event.summary_index,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ReasoningSummaryPartAdded(
+                    notification,
+                ))
+                .await;
+        }
+        EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
+            call_id,
+            command,
+            cwd,
+            reason,
+            risk,
+            parsed_cmd,
+        }) => {
+            let params = ExecCommandApprovalParams {
+                conversation_id,
+                call_id,
+                command,
+                cwd,
+                reason,
+                risk,
+                parsed_cmd,
+            };
+            let rx = outgoing
+                .send_request(ServerRequestPayload::ExecCommandApproval(params))
+                .await;
+
+            // TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
+            tokio::spawn(async move {
+                on_exec_approval_response(event_id, rx, conversation).await;
+            });
+        }
+        EventMsg::TokenCount(token_count_event) => {
+            if let Some(rate_limits) = token_count_event.rate_limits {
+                outgoing
+                    .send_server_notification(ServerNotification::AccountRateLimitsUpdated(
+                        AccountRateLimitsUpdatedNotification {
+                            rate_limits: rate_limits.into(),
+                        },
+                    ))
+                    .await;
+            }
+        }
+        EventMsg::ItemStarted(item_started_event) => {
+            let item: ThreadItem = item_started_event.item.clone().into();
+            let notification = ItemStartedNotification { item };
+            outgoing
+                .send_server_notification(ServerNotification::ItemStarted(notification))
+                .await;
+        }
+        EventMsg::ItemCompleted(item_completed_event) => {
+            let item: ThreadItem = item_completed_event.item.clone().into();
+            let notification = ItemCompletedNotification { item };
+            outgoing
+                .send_server_notification(ServerNotification::ItemCompleted(notification))
+                .await;
+        }
+        // If this is a TurnAborted, reply to any pending interrupt requests.
+        EventMsg::TurnAborted(turn_aborted_event) => {
+            let pending = {
+                let mut map = pending_interrupts.lock().await;
+                map.remove(&conversation_id).unwrap_or_default()
+            };
+            if !pending.is_empty() {
+                for (rid, ver) in pending {
+                    match ver {
+                        ApiVersion::V1 => {
+                            let response = InterruptConversationResponse {
+                                abort_reason: turn_aborted_event.reason.clone(),
+                            };
+                            outgoing.send_response(rid, response).await;
+                        }
+                        ApiVersion::V2 => {
+                            let response = TurnInterruptResponse {};
+                            outgoing.send_response(rid, response).await;
+                        }
+                    }
+                }
+            }
+        }
+
+        _ => {}
+    }
+}
+
+async fn on_patch_approval_response(
+    event_id: String,
+    receiver: oneshot::Receiver<JsonRpcResult>,
+    codex: Arc<CodexConversation>,
+) {
+    let response = receiver.await;
+    let value = match response {
+        Ok(value) => value,
+        Err(err) => {
+            error!("request failed: {err:?}");
+            if let Err(submit_err) = codex
+                .submit(Op::PatchApproval {
+                    id: event_id.clone(),
+                    decision: ReviewDecision::Denied,
+                })
+                .await
+            {
+                error!("failed to submit denied PatchApproval after request failure: {submit_err}");
+            }
+            return;
+        }
+    };
+
+    let response =
+        serde_json::from_value::<ApplyPatchApprovalResponse>(value).unwrap_or_else(|err| {
+            error!("failed to deserialize ApplyPatchApprovalResponse: {err}");
+            ApplyPatchApprovalResponse {
+                decision: ReviewDecision::Denied,
+            }
+        });
+
+    if let Err(err) = codex
+        .submit(Op::PatchApproval {
+            id: event_id,
+            decision: response.decision,
+        })
+        .await
+    {
+        error!("failed to submit PatchApproval: {err}");
+    }
+}
+
+async fn on_exec_approval_response(
+    event_id: String,
+    receiver: oneshot::Receiver<JsonRpcResult>,
+    conversation: Arc<CodexConversation>,
+) {
+    let response = receiver.await;
+    let value = match response {
+        Ok(value) => value,
+        Err(err) => {
+            error!("request failed: {err:?}");
+            return;
+        }
+    };
+
+    // Try to deserialize `value` and then make the appropriate call to `codex`.
+    let response =
+        serde_json::from_value::<ExecCommandApprovalResponse>(value).unwrap_or_else(|err| {
+            error!("failed to deserialize ExecCommandApprovalResponse: {err}");
+            // If we cannot deserialize the response, we deny the request to be
+            // conservative.
+            ExecCommandApprovalResponse {
+                decision: ReviewDecision::Denied,
+            }
+        });
+
+    if let Err(err) = conversation
+        .submit(Op::ExecApproval {
+            id: event_id,
+            decision: response.decision,
+        })
+        .await
+    {
+        error!("failed to submit ExecApproval: {err}");
+    }
+}
+
+/// similar to handle_mcp_tool_call_begin in exec
+async fn construct_mcp_tool_call_notification(
+    begin_event: McpToolCallBeginEvent,
+) -> ItemStartedNotification {
+    let item = ThreadItem::McpToolCall {
+        id: begin_event.call_id,
+        server: begin_event.invocation.server,
+        tool: begin_event.invocation.tool,
+        status: McpToolCallStatus::InProgress,
+        arguments: begin_event
+            .invocation
+            .arguments
+            .unwrap_or(JsonRpcResult::Null),
+        result: None,
+        error: None,
+    };
+    ItemStartedNotification { item }
+}
+
+/// simiilar to handle_mcp_tool_call_end in exec
+async fn construct_mcp_tool_call_end_notification(
+    end_event: McpToolCallEndEvent,
+) -> ItemCompletedNotification {
+    let status = if end_event.is_success() {
+        McpToolCallStatus::Completed
+    } else {
+        McpToolCallStatus::Failed
+    };
+
+    let (result, error) = match &end_event.result {
+        Ok(value) => (
+            Some(McpToolCallResult {
+                content: value.content.clone(),
+                structured_content: value.structured_content.clone(),
+            }),
+            None,
+        ),
+        Err(message) => (
+            None,
+            Some(McpToolCallError {
+                message: message.clone(),
+            }),
+        ),
+    };
+
+    let item = ThreadItem::McpToolCall {
+        id: end_event.call_id,
+        server: end_event.invocation.server,
+        tool: end_event.invocation.tool,
+        status,
+        arguments: end_event
+            .invocation
+            .arguments
+            .unwrap_or(JsonRpcResult::Null),
+        result,
+        error,
+    };
+    ItemCompletedNotification { item }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use codex_core::protocol::McpInvocation;
+    use mcp_types::CallToolResult;
+    use mcp_types::ContentBlock;
+    use mcp_types::TextContent;
+    use pretty_assertions::assert_eq;
+    use serde_json::Value as JsonValue;
+    use std::time::Duration;
+
+    #[tokio::test]
+    async fn test_construct_mcp_tool_call_begin_notification_with_args() {
+        let begin_event = McpToolCallBeginEvent {
+            call_id: "call_123".to_string(),
+            invocation: McpInvocation {
+                server: "codex".to_string(),
+                tool: "list_mcp_resources".to_string(),
+                arguments: Some(serde_json::json!({"server": ""})),
+            },
+        };
+
+        let notification = construct_mcp_tool_call_notification(begin_event.clone()).await;
+
+        let expected = ItemStartedNotification {
+            item: ThreadItem::McpToolCall {
+                id: begin_event.call_id,
+                server: begin_event.invocation.server,
+                tool: begin_event.invocation.tool,
+                status: McpToolCallStatus::InProgress,
+                arguments: serde_json::json!({"server": ""}),
+                result: None,
+                error: None,
+            },
+        };
+
+        assert_eq!(notification, expected);
+    }
+
+    #[tokio::test]
+    async fn test_construct_mcp_tool_call_begin_notification_without_args() {
+        let begin_event = McpToolCallBeginEvent {
+            call_id: "call_456".to_string(),
+            invocation: McpInvocation {
+                server: "codex".to_string(),
+                tool: "list_mcp_resources".to_string(),
+                arguments: None,
+            },
+        };
+
+        let notification = construct_mcp_tool_call_notification(begin_event.clone()).await;
+
+        let expected = ItemStartedNotification {
+            item: ThreadItem::McpToolCall {
+                id: begin_event.call_id,
+                server: begin_event.invocation.server,
+                tool: begin_event.invocation.tool,
+                status: McpToolCallStatus::InProgress,
+                arguments: JsonValue::Null,
+                result: None,
+                error: None,
+            },
+        };
+
+        assert_eq!(notification, expected);
+    }
+
+    #[tokio::test]
+    async fn test_construct_mcp_tool_call_end_notification_success() {
+        let content = vec![ContentBlock::TextContent(TextContent {
+            annotations: None,
+            text: "{\"resources\":[]}".to_string(),
+            r#type: "text".to_string(),
+        })];
+        let result = CallToolResult {
+            content: content.clone(),
+            is_error: Some(false),
+            structured_content: None,
+        };
+
+        let end_event = McpToolCallEndEvent {
+            call_id: "call_789".to_string(),
+            invocation: McpInvocation {
+                server: "codex".to_string(),
+                tool: "list_mcp_resources".to_string(),
+                arguments: Some(serde_json::json!({"server": ""})),
+            },
+            duration: Duration::from_nanos(92708),
+            result: Ok(result),
+        };
+
+        let notification = construct_mcp_tool_call_end_notification(end_event.clone()).await;
+
+        let expected = ItemCompletedNotification {
+            item: ThreadItem::McpToolCall {
+                id: end_event.call_id,
+                server: end_event.invocation.server,
+                tool: end_event.invocation.tool,
+                status: McpToolCallStatus::Completed,
+                arguments: serde_json::json!({"server": ""}),
+                result: Some(McpToolCallResult {
+                    content,
+                    structured_content: None,
+                }),
+                error: None,
+            },
+        };
+
+        assert_eq!(notification, expected);
+    }
+
+    #[tokio::test]
+    async fn test_construct_mcp_tool_call_end_notification_error() {
+        let end_event = McpToolCallEndEvent {
+            call_id: "call_err".to_string(),
+            invocation: McpInvocation {
+                server: "codex".to_string(),
+                tool: "list_mcp_resources".to_string(),
+                arguments: None,
+            },
+            duration: Duration::from_millis(1),
+            result: Err("boom".to_string()),
+        };
+
+        let notification = construct_mcp_tool_call_end_notification(end_event.clone()).await;
+
+        let expected = ItemCompletedNotification {
+            item: ThreadItem::McpToolCall {
+                id: end_event.call_id,
+                server: end_event.invocation.server,
+                tool: end_event.invocation.tool,
+                status: McpToolCallStatus::Failed,
+                arguments: JsonValue::Null,
+                result: None,
+                error: Some(McpToolCallError {
+                    message: "boom".to_string(),
+                }),
+            },
+        };
+
+        assert_eq!(notification, expected);
+    }
+}

codex-rs/app-server/src/codex_message_processor.rs

@@ -1,3 +1,4 @@
+use crate::bespoke_event_handling::apply_bespoke_event_handling;
 use crate::error_code::INTERNAL_ERROR_CODE;
 use crate::error_code::INVALID_REQUEST_ERROR_CODE;
 use crate::fuzzy_file_search::run_fuzzy_file_search;
@@ -8,12 +9,9 @@ use chrono::DateTime;
 use chrono::Utc;
 use codex_app_server_protocol::Account;
 use codex_app_server_protocol::AccountLoginCompletedNotification;
-use codex_app_server_protocol::AccountRateLimitsUpdatedNotification;
 use codex_app_server_protocol::AccountUpdatedNotification;
 use codex_app_server_protocol::AddConversationListenerParams;
 use codex_app_server_protocol::AddConversationSubscriptionResponse;
-use codex_app_server_protocol::ApplyPatchApprovalParams;
-use codex_app_server_protocol::ApplyPatchApprovalResponse;
 use codex_app_server_protocol::ArchiveConversationParams;
 use codex_app_server_protocol::ArchiveConversationResponse;
 use codex_app_server_protocol::AskForApproval;
@@ -25,8 +23,6 @@ use codex_app_server_protocol::CancelLoginChatGptResponse;
 use codex_app_server_protocol::ClientRequest;
 use codex_app_server_protocol::ConversationGitInfo;
 use codex_app_server_protocol::ConversationSummary;
-use codex_app_server_protocol::ExecCommandApprovalParams;
-use codex_app_server_protocol::ExecCommandApprovalResponse;
 use codex_app_server_protocol::ExecOneOffCommandParams;
 use codex_app_server_protocol::ExecOneOffCommandResponse;
 use codex_app_server_protocol::FeedbackUploadParams;
@@ -45,9 +41,6 @@ use codex_app_server_protocol::GetUserSavedConfigResponse;
 use codex_app_server_protocol::GitDiffToRemoteResponse;
 use codex_app_server_protocol::InputItem as WireInputItem;
 use codex_app_server_protocol::InterruptConversationParams;
-use codex_app_server_protocol::InterruptConversationResponse;
-use codex_app_server_protocol::ItemCompletedNotification;
-use codex_app_server_protocol::ItemStartedNotification;
 use codex_app_server_protocol::JSONRPCErrorError;
 use codex_app_server_protocol::ListConversationsParams;
 use codex_app_server_protocol::ListConversationsResponse;
@@ -65,7 +58,6 @@ use codex_app_server_protocol::NewConversationResponse;
 use codex_app_server_protocol::RemoveConversationListenerParams;
 use codex_app_server_protocol::RemoveConversationSubscriptionResponse;
 use codex_app_server_protocol::RequestId;
-use codex_app_server_protocol::Result as JsonRpcResult;
 use codex_app_server_protocol::ResumeConversationParams;
 use codex_app_server_protocol::ResumeConversationResponse;
 use codex_app_server_protocol::SandboxMode;
@@ -74,7 +66,6 @@ use codex_app_server_protocol::SendUserMessageResponse;
 use codex_app_server_protocol::SendUserTurnParams;
 use codex_app_server_protocol::SendUserTurnResponse;
 use codex_app_server_protocol::ServerNotification;
-use codex_app_server_protocol::ServerRequestPayload;
 use codex_app_server_protocol::SessionConfiguredNotification;
 use codex_app_server_protocol::SetDefaultModelParams;
 use codex_app_server_protocol::SetDefaultModelResponse;
@@ -91,7 +82,6 @@ use codex_app_server_protocol::ThreadStartResponse;
 use codex_app_server_protocol::ThreadStartedNotification;
 use codex_app_server_protocol::Turn;
 use codex_app_server_protocol::TurnInterruptParams;
-use codex_app_server_protocol::TurnInterruptResponse;
 use codex_app_server_protocol::TurnStartParams;
 use codex_app_server_protocol::TurnStartResponse;
 use codex_app_server_protocol::TurnStartedNotification;
@@ -123,12 +113,8 @@ use codex_core::find_conversation_path_by_id_str;
 use codex_core::get_platform_sandbox;
 use codex_core::git_info::git_diff_to_remote;
 use codex_core::parse_cursor;
-use codex_core::protocol::ApplyPatchApprovalRequestEvent;
-use codex_core::protocol::Event;
 use codex_core::protocol::EventMsg;
-use codex_core::protocol::ExecApprovalRequestEvent;
 use codex_core::protocol::Op;
-use codex_core::protocol::ReviewDecision;
 use codex_core::read_head_for_summary;
 use codex_feedback::CodexFeedback;
 use codex_login::ServerOptions as LoginServerOptions;
@@ -163,7 +149,7 @@ use tracing::warn;
 use uuid::Uuid;
 
 type PendingInterruptQueue = Vec<(RequestId, ApiVersion)>;
-type PendingInterrupts = Arc<Mutex<HashMap<ConversationId, PendingInterruptQueue>>>;
+pub(crate) type PendingInterrupts = Arc<Mutex<HashMap<ConversationId, PendingInterruptQueue>>>;
 
 // Duration before a ChatGPT login attempt is abandoned.
 const LOGIN_CHATGPT_TIMEOUT: Duration = Duration::from_secs(10 * 60);
@@ -194,7 +180,7 @@ pub(crate) struct CodexMessageProcessor {
 }
 
 #[derive(Clone, Copy, Debug)]
-enum ApiVersion {
+pub(crate) enum ApiVersion {
     V1,
     V2,
 }
@@ -1201,21 +1187,17 @@ impl CodexMessageProcessor {
     }
 
     async fn thread_start(&mut self, request_id: RequestId, params: ThreadStartParams) {
-        // Build ConfigOverrides directly from ThreadStartParams for config derivation.
-        let cli_overrides = params.config;
-        let overrides = ConfigOverrides {
-            model: params.model,
-            cwd: params.cwd.map(PathBuf::from),
-            approval_policy: params.approval_policy.map(AskForApproval::to_core),
-            sandbox_mode: params.sandbox.map(SandboxMode::to_core),
-            model_provider: params.model_provider,
-            codex_linux_sandbox_exe: self.codex_linux_sandbox_exe.clone(),
-            base_instructions: params.base_instructions,
-            developer_instructions: params.developer_instructions,
-            ..Default::default()
-        };
+        let overrides = self.build_thread_config_overrides(
+            params.model,
+            params.model_provider,
+            params.cwd,
+            params.approval_policy,
+            params.sandbox,
+            params.base_instructions,
+            params.developer_instructions,
+        );
 
-        let config = match derive_config_from_params(overrides, cli_overrides).await {
+        let config = match derive_config_from_params(overrides, params.config).await {
             Ok(config) => config,
             Err(err) => {
                 let error = JSONRPCErrorError {
@@ -1244,16 +1226,15 @@ impl CodexMessageProcessor {
                 {
                     Ok(summary) => summary_to_thread(summary),
                     Err(err) => {
-                        warn!(
-                            "failed to load summary for new thread {}: {}",
-                            conversation_id, err
-                        );
-                        Thread {
-                            id: conversation_id.to_string(),
-                            preview: String::new(),
-                            model_provider: self.config.model_provider_id.clone(),
-                            created_at: chrono::Utc::now().timestamp(),
-                        }
+                        self.send_internal_error(
+                            request_id,
+                            format!(
+                                "failed to load rollout `{}` for conversation {conversation_id}: {err}",
+                                rollout_path.display()
+                            ),
+                        )
+                        .await;
+                        return;
                     }
                 };
 
@@ -1292,6 +1273,31 @@ impl CodexMessageProcessor {
         }
     }
 
+    #[allow(clippy::too_many_arguments)]
+    fn build_thread_config_overrides(
+        &self,
+        model: Option<String>,
+        model_provider: Option<String>,
+        cwd: Option<String>,
+        approval_policy: Option<codex_app_server_protocol::AskForApproval>,
+        sandbox: Option<SandboxMode>,
+        base_instructions: Option<String>,
+        developer_instructions: Option<String>,
+    ) -> ConfigOverrides {
+        ConfigOverrides {
+            model,
+            model_provider,
+            cwd: cwd.map(PathBuf::from),
+            approval_policy: approval_policy
+                .map(codex_app_server_protocol::AskForApproval::to_core),
+            sandbox_mode: sandbox.map(SandboxMode::to_core),
+            codex_linux_sandbox_exe: self.codex_linux_sandbox_exe.clone(),
+            base_instructions,
+            developer_instructions,
+            ..Default::default()
+        }
+    }
+
     async fn thread_archive(&mut self, request_id: RequestId, params: ThreadArchiveParams) {
         let conversation_id = match ConversationId::from_string(&params.thread_id) {
             Ok(id) => id,
@@ -1374,88 +1380,147 @@ impl CodexMessageProcessor {
     }
 
     async fn thread_resume(&mut self, request_id: RequestId, params: ThreadResumeParams) {
-        let conversation_id = match ConversationId::from_string(&params.thread_id) {
-            Ok(id) => id,
-            Err(err) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!("invalid thread id: {err}"),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
+        let ThreadResumeParams {
+            thread_id,
+            history,
+            path,
+            model,
+            model_provider,
+            cwd,
+            approval_policy,
+            sandbox,
+            config: cli_overrides,
+            base_instructions,
+            developer_instructions,
+        } = params;
+
+        let overrides_requested = model.is_some()
+            || model_provider.is_some()
+            || cwd.is_some()
+            || approval_policy.is_some()
+            || sandbox.is_some()
+            || cli_overrides.is_some()
+            || base_instructions.is_some()
+            || developer_instructions.is_some();
+
+        let config = if overrides_requested {
+            let overrides = self.build_thread_config_overrides(
+                model,
+                model_provider,
+                cwd,
+                approval_policy,
+                sandbox,
+                base_instructions,
+                developer_instructions,
+            );
+            match derive_config_from_params(overrides, cli_overrides).await {
+                Ok(config) => config,
+                Err(err) => {
+                    let error = JSONRPCErrorError {
+                        code: INVALID_REQUEST_ERROR_CODE,
+                        message: format!("error deriving config: {err}"),
+                        data: None,
+                    };
+                    self.outgoing.send_error(request_id, error).await;
+                    return;
+                }
+            }
+        } else {
+            self.config.as_ref().clone()
+        };
+
+        let conversation_history = if let Some(history) = history {
+            if history.is_empty() {
+                self.send_invalid_request_error(
+                    request_id,
+                    "history must not be empty".to_string(),
+                )
+                .await;
                 return;
             }
+            InitialHistory::Forked(history.into_iter().map(RolloutItem::ResponseItem).collect())
+        } else if let Some(path) = path {
+            match RolloutRecorder::get_rollout_history(&path).await {
+                Ok(initial_history) => initial_history,
+                Err(err) => {
+                    self.send_invalid_request_error(
+                        request_id,
+                        format!("failed to load rollout `{}`: {err}", path.display()),
+                    )
+                    .await;
+                    return;
+                }
+            }
+        } else {
+            let existing_conversation_id = match ConversationId::from_string(&thread_id) {
+                Ok(id) => id,
+                Err(err) => {
+                    let error = JSONRPCErrorError {
+                        code: INVALID_REQUEST_ERROR_CODE,
+                        message: format!("invalid thread id: {err}"),
+                        data: None,
+                    };
+                    self.outgoing.send_error(request_id, error).await;
+                    return;
+                }
+            };
+
+            let path = match find_conversation_path_by_id_str(
+                &self.config.codex_home,
+                &existing_conversation_id.to_string(),
+            )
+            .await
+            {
+                Ok(Some(p)) => p,
+                Ok(None) => {
+                    self.send_invalid_request_error(
+                        request_id,
+                        format!("no rollout found for conversation id {existing_conversation_id}"),
+                    )
+                    .await;
+                    return;
+                }
+                Err(err) => {
+                    self.send_invalid_request_error(
+                        request_id,
+                        format!(
+                            "failed to locate conversation id {existing_conversation_id}: {err}"
+                        ),
+                    )
+                    .await;
+                    return;
+                }
+            };
+
+            match RolloutRecorder::get_rollout_history(&path).await {
+                Ok(initial_history) => initial_history,
+                Err(err) => {
+                    self.send_invalid_request_error(
+                        request_id,
+                        format!("failed to load rollout `{}`: {err}", path.display()),
+                    )
+                    .await;
+                    return;
+                }
+            }
         };
 
-        let path = match find_conversation_path_by_id_str(
-            &self.config.codex_home,
-            &conversation_id.to_string(),
-        )
-        .await
-        {
-            Ok(Some(p)) => p,
-            Ok(None) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!("no rollout found for conversation id {conversation_id}"),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
-                return;
-            }
-            Err(err) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!("failed to locate conversation id {conversation_id}: {err}"),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
-                return;
-            }
-        };
-
-        let fallback_provider = self.config.model_provider_id.as_str();
-        let summary = match read_summary_from_rollout(&path, fallback_provider).await {
-            Ok(s) => s,
-            Err(err) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!("failed to load rollout `{}`: {err}", path.display()),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
-                return;
-            }
-        };
-
-        let initial_history = match RolloutRecorder::get_rollout_history(&summary.path).await {
-            Ok(initial_history) => initial_history,
-            Err(err) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!(
-                        "failed to load rollout `{}` for conversation {conversation_id}: {err}",
-                        summary.path.display()
-                    ),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
-                return;
-            }
-        };
+        let fallback_model_provider = config.model_provider_id.clone();
 
         match self
             .conversation_manager
             .resume_conversation_with_history(
-                self.config.as_ref().clone(),
-                initial_history,
+                config,
+                conversation_history,
                 self.auth_manager.clone(),
             )
             .await
         {
-            Ok(_) => {
-                let thread = summary_to_thread(summary);
-
+            Ok(NewConversation {
+                conversation_id,
+                session_configured,
+                ..
+            }) => {
                 // Auto-attach a conversation listener when resuming a thread.
                 if let Err(err) = self
                     .attach_conversation_listener(conversation_id, false)
@@ -1468,6 +1533,25 @@ impl CodexMessageProcessor {
                     );
                 }
 
+                let thread = match read_summary_from_rollout(
+                    session_configured.rollout_path.as_path(),
+                    fallback_model_provider.as_str(),
+                )
+                .await
+                {
+                    Ok(summary) => summary_to_thread(summary),
+                    Err(err) => {
+                        self.send_internal_error(
+                            request_id,
+                            format!(
+                                "failed to load rollout `{}` for conversation {conversation_id}: {err}",
+                                session_configured.rollout_path.display()
+                            ),
+                        )
+                        .await;
+                        return;
+                    }
+                };
                 let response = ThreadResumeResponse { thread };
                 self.outgoing.send_response(request_id, response).await;
             }
@@ -1878,6 +1962,15 @@ impl CodexMessageProcessor {
         self.outgoing.send_error(request_id, error).await;
     }
 
+    async fn send_internal_error(&self, request_id: RequestId, message: String) {
+        let error = JSONRPCErrorError {
+            code: INTERNAL_ERROR_CODE,
+            message,
+            data: None,
+        };
+        self.outgoing.send_error(request_id, error).await;
+    }
+
     async fn archive_conversation(
         &mut self,
         request_id: RequestId,
@@ -2544,115 +2637,6 @@ impl CodexMessageProcessor {
     }
 }
 
-async fn apply_bespoke_event_handling(
-    event: Event,
-    conversation_id: ConversationId,
-    conversation: Arc<CodexConversation>,
-    outgoing: Arc<OutgoingMessageSender>,
-    pending_interrupts: PendingInterrupts,
-) {
-    let Event { id: event_id, msg } = event;
-    match msg {
-        EventMsg::ApplyPatchApprovalRequest(ApplyPatchApprovalRequestEvent {
-            call_id,
-            changes,
-            reason,
-            grant_root,
-        }) => {
-            let params = ApplyPatchApprovalParams {
-                conversation_id,
-                call_id,
-                file_changes: changes,
-                reason,
-                grant_root,
-            };
-            let rx = outgoing
-                .send_request(ServerRequestPayload::ApplyPatchApproval(params))
-                .await;
-            // TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
-            tokio::spawn(async move {
-                on_patch_approval_response(event_id, rx, conversation).await;
-            });
-        }
-        EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
-            call_id,
-            command,
-            cwd,
-            reason,
-            risk,
-            parsed_cmd,
-        }) => {
-            let params = ExecCommandApprovalParams {
-                conversation_id,
-                call_id,
-                command,
-                cwd,
-                reason,
-                risk,
-                parsed_cmd,
-            };
-            let rx = outgoing
-                .send_request(ServerRequestPayload::ExecCommandApproval(params))
-                .await;
-
-            // TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
-            tokio::spawn(async move {
-                on_exec_approval_response(event_id, rx, conversation).await;
-            });
-        }
-        EventMsg::TokenCount(token_count_event) => {
-            if let Some(rate_limits) = token_count_event.rate_limits {
-                outgoing
-                    .send_server_notification(ServerNotification::AccountRateLimitsUpdated(
-                        AccountRateLimitsUpdatedNotification {
-                            rate_limits: rate_limits.into(),
-                        },
-                    ))
-                    .await;
-            }
-        }
-        EventMsg::ItemStarted(item_started_event) => {
-            let item: ThreadItem = item_started_event.item.clone().into();
-            let notification = ItemStartedNotification { item };
-            outgoing
-                .send_server_notification(ServerNotification::ItemStarted(notification))
-                .await;
-        }
-        EventMsg::ItemCompleted(item_completed_event) => {
-            let item: ThreadItem = item_completed_event.item.clone().into();
-            let notification = ItemCompletedNotification { item };
-            outgoing
-                .send_server_notification(ServerNotification::ItemCompleted(notification))
-                .await;
-        }
-        // If this is a TurnAborted, reply to any pending interrupt requests.
-        EventMsg::TurnAborted(turn_aborted_event) => {
-            let pending = {
-                let mut map = pending_interrupts.lock().await;
-                map.remove(&conversation_id).unwrap_or_default()
-            };
-            if !pending.is_empty() {
-                for (rid, ver) in pending {
-                    match ver {
-                        ApiVersion::V1 => {
-                            let response = InterruptConversationResponse {
-                                abort_reason: turn_aborted_event.reason.clone(),
-                            };
-                            outgoing.send_response(rid, response).await;
-                        }
-                        ApiVersion::V2 => {
-                            let response = TurnInterruptResponse {};
-                            outgoing.send_response(rid, response).await;
-                        }
-                    }
-                }
-            }
-        }
-
-        _ => {}
-    }
-}
-
 async fn derive_config_from_params(
     overrides: ConfigOverrides,
     cli_overrides: Option<std::collections::HashMap<String, serde_json::Value>>,
@@ -2666,84 +2650,6 @@ async fn derive_config_from_params(
     Config::load_with_cli_overrides(cli_overrides, overrides).await
 }
 
-async fn on_patch_approval_response(
-    event_id: String,
-    receiver: oneshot::Receiver<JsonRpcResult>,
-    codex: Arc<CodexConversation>,
-) {
-    let response = receiver.await;
-    let value = match response {
-        Ok(value) => value,
-        Err(err) => {
-            error!("request failed: {err:?}");
-            if let Err(submit_err) = codex
-                .submit(Op::PatchApproval {
-                    id: event_id.clone(),
-                    decision: ReviewDecision::Denied,
-                })
-                .await
-            {
-                error!("failed to submit denied PatchApproval after request failure: {submit_err}");
-            }
-            return;
-        }
-    };
-
-    let response =
-        serde_json::from_value::<ApplyPatchApprovalResponse>(value).unwrap_or_else(|err| {
-            error!("failed to deserialize ApplyPatchApprovalResponse: {err}");
-            ApplyPatchApprovalResponse {
-                decision: ReviewDecision::Denied,
-            }
-        });
-
-    if let Err(err) = codex
-        .submit(Op::PatchApproval {
-            id: event_id,
-            decision: response.decision,
-        })
-        .await
-    {
-        error!("failed to submit PatchApproval: {err}");
-    }
-}
-
-async fn on_exec_approval_response(
-    event_id: String,
-    receiver: oneshot::Receiver<JsonRpcResult>,
-    conversation: Arc<CodexConversation>,
-) {
-    let response = receiver.await;
-    let value = match response {
-        Ok(value) => value,
-        Err(err) => {
-            error!("request failed: {err:?}");
-            return;
-        }
-    };
-
-    // Try to deserialize `value` and then make the appropriate call to `codex`.
-    let response =
-        serde_json::from_value::<ExecCommandApprovalResponse>(value).unwrap_or_else(|err| {
-            error!("failed to deserialize ExecCommandApprovalResponse: {err}");
-            // If we cannot deserialize the response, we deny the request to be
-            // conservative.
-            ExecCommandApprovalResponse {
-                decision: ReviewDecision::Denied,
-            }
-        });
-
-    if let Err(err) = conversation
-        .submit(Op::ExecApproval {
-            id: event_id,
-            decision: response.decision,
-        })
-        .await
-    {
-        error!("failed to submit ExecApproval: {err}");
-    }
-}
-
 async fn read_summary_from_rollout(
     path: &Path,
     fallback_provider: &str,
@@ -2867,6 +2773,7 @@ fn parse_datetime(timestamp: Option<&str>) -> Option<DateTime<Utc>> {
 fn summary_to_thread(summary: ConversationSummary) -> Thread {
     let ConversationSummary {
         conversation_id,
+        path,
         preview,
         timestamp,
         model_provider,
@@ -2880,6 +2787,7 @@ fn summary_to_thread(summary: ConversationSummary) -> Thread {
         preview,
         model_provider,
         created_at: created_at.map(|dt| dt.timestamp()).unwrap_or(0),
+        path,
     }
 }
 

codex-rs/app-server/src/lib.rs

@@ -28,6 +28,7 @@ use tracing_subscriber::filter::Targets;
 use tracing_subscriber::layer::SubscriberExt;
 use tracing_subscriber::util::SubscriberInitExt;
 
+mod bespoke_event_handling;
 mod codex_message_processor;
 mod error_code;
 mod fuzzy_file_search;

codex-rs/app-server/tests/suite/v2/model_list.rs

@@ -46,9 +46,9 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
 
     let expected_models = vec![
         Model {
-            id: "gpt-5-codex".to_string(),
-            model: "gpt-5-codex".to_string(),
-            display_name: "gpt-5-codex".to_string(),
+            id: "gpt-5.1-codex".to_string(),
+            model: "gpt-5.1-codex".to_string(),
+            display_name: "gpt-5.1-codex".to_string(),
             description: "Optimized for codex.".to_string(),
             supported_reasoning_efforts: vec![
                 ReasoningEffortOption {
@@ -69,15 +69,30 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
             is_default: true,
         },
         Model {
-            id: "gpt-5".to_string(),
-            model: "gpt-5".to_string(),
-            display_name: "gpt-5".to_string(),
-            description: "Broad world knowledge with strong general reasoning.".to_string(),
+            id: "gpt-5.1-codex-mini".to_string(),
+            model: "gpt-5.1-codex-mini".to_string(),
+            display_name: "gpt-5.1-codex-mini".to_string(),
+            description: "Optimized for codex. Cheaper, faster, but less capable.".to_string(),
             supported_reasoning_efforts: vec![
                 ReasoningEffortOption {
-                    reasoning_effort: ReasoningEffort::Minimal,
-                    description: "Fastest responses with little reasoning".to_string(),
+                    reasoning_effort: ReasoningEffort::Medium,
+                    description: "Dynamically adjusts reasoning based on the task".to_string(),
                 },
+                ReasoningEffortOption {
+                    reasoning_effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems"
+                        .to_string(),
+                },
+            ],
+            default_reasoning_effort: ReasoningEffort::Medium,
+            is_default: false,
+        },
+        Model {
+            id: "gpt-5.1".to_string(),
+            model: "gpt-5.1".to_string(),
+            display_name: "gpt-5.1".to_string(),
+            description: "Broad world knowledge with strong general reasoning.".to_string(),
+            supported_reasoning_efforts: vec![
                 ReasoningEffortOption {
                     reasoning_effort: ReasoningEffort::Low,
                     description: "Balances speed with some reasoning; useful for straightforward \
@@ -132,7 +147,7 @@ async fn list_models_pagination_works() -> Result<()> {
     } = to_response::<ModelListResponse>(first_response)?;
 
     assert_eq!(first_items.len(), 1);
-    assert_eq!(first_items[0].id, "gpt-5-codex");
+    assert_eq!(first_items[0].id, "gpt-5.1-codex");
     let next_cursor = first_cursor.ok_or_else(|| anyhow!("cursor for second page"))?;
 
     let second_request = mcp
@@ -154,8 +169,30 @@ async fn list_models_pagination_works() -> Result<()> {
     } = to_response::<ModelListResponse>(second_response)?;
 
     assert_eq!(second_items.len(), 1);
-    assert_eq!(second_items[0].id, "gpt-5");
-    assert!(second_cursor.is_none());
+    assert_eq!(second_items[0].id, "gpt-5.1-codex-mini");
+    let third_cursor = second_cursor.ok_or_else(|| anyhow!("cursor for third page"))?;
+
+    let third_request = mcp
+        .send_list_models_request(ModelListParams {
+            limit: Some(1),
+            cursor: Some(third_cursor.clone()),
+        })
+        .await?;
+
+    let third_response: JSONRPCResponse = timeout(
+        DEFAULT_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(third_request)),
+    )
+    .await??;
+
+    let ModelListResponse {
+        data: third_items,
+        next_cursor: third_cursor,
+    } = to_response::<ModelListResponse>(third_response)?;
+
+    assert_eq!(third_items.len(), 1);
+    assert_eq!(third_items[0].id, "gpt-5.1");
+    assert!(third_cursor.is_none());
     Ok(())
 }
 

codex-rs/app-server/tests/suite/v2/thread_list.rs

@@ -6,10 +6,8 @@ use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::RequestId;
 use codex_app_server_protocol::ThreadListParams;
 use codex_app_server_protocol::ThreadListResponse;
-use serde_json::json;
 use tempfile::TempDir;
 use tokio::time::timeout;
-use uuid::Uuid;
 
 const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
 
@@ -150,46 +148,13 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
         "X",
         Some("mock_provider"),
     )?; // mock_provider
-    // one with a different provider
-    let uuid = Uuid::new_v4();
-    let dir = codex_home
-        .path()
-        .join("sessions")
-        .join("2025")
-        .join("01")
-        .join("02");
-    std::fs::create_dir_all(&dir)?;
-    let file_path = dir.join(format!("rollout-2025-01-02T11-00-00-{uuid}.jsonl"));
-    let lines = [
-        json!({
-            "timestamp": "2025-01-02T11:00:00Z",
-            "type": "session_meta",
-            "payload": {
-                "id": uuid,
-                "timestamp": "2025-01-02T11:00:00Z",
-                "cwd": "/",
-                "originator": "codex",
-                "cli_version": "0.0.0",
-                "instructions": null,
-                "source": "vscode",
-                "model_provider": "other_provider"
-            }
-        })
-        .to_string(),
-        json!({
-            "timestamp": "2025-01-02T11:00:00Z",
-            "type":"response_item",
-            "payload": {"type":"message","role":"user","content":[{"type":"input_text","text":"X"}]}
-        })
-        .to_string(),
-        json!({
-            "timestamp": "2025-01-02T11:00:00Z",
-            "type":"event_msg",
-            "payload": {"type":"user_message","message":"X","kind":"plain"}
-        })
-        .to_string(),
-    ];
-    std::fs::write(file_path, lines.join("\n") + "\n")?;
+    let _b = create_fake_rollout(
+        codex_home.path(),
+        "2025-01-02T11-00-00",
+        "2025-01-02T11:00:00Z",
+        "X",
+        Some("other_provider"),
+    )?;
 
     let mut mcp = McpProcess::new(codex_home.path()).await?;
     timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;

codex-rs/app-server/tests/suite/v2/thread_resume.rs

@@ -8,13 +8,15 @@ use codex_app_server_protocol::ThreadResumeParams;
 use codex_app_server_protocol::ThreadResumeResponse;
 use codex_app_server_protocol::ThreadStartParams;
 use codex_app_server_protocol::ThreadStartResponse;
+use codex_protocol::models::ContentItem;
+use codex_protocol::models::ResponseItem;
 use tempfile::TempDir;
 use tokio::time::timeout;
 
 const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
 
 #[tokio::test]
-async fn thread_resume_returns_existing_thread() -> Result<()> {
+async fn thread_resume_returns_original_thread() -> Result<()> {
     let server = create_mock_chat_completions_server(vec![]).await;
     let codex_home = TempDir::new()?;
     create_config_toml(codex_home.path(), &server.uri())?;
@@ -40,6 +42,7 @@ async fn thread_resume_returns_existing_thread() -> Result<()> {
     let resume_id = mcp
         .send_thread_resume_request(ThreadResumeParams {
             thread_id: thread.id.clone(),
+            ..Default::default()
         })
         .await?;
     let resume_resp: JSONRPCResponse = timeout(
@@ -54,6 +57,105 @@ async fn thread_resume_returns_existing_thread() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test]
+async fn thread_resume_prefers_path_over_thread_id() -> Result<()> {
+    let server = create_mock_chat_completions_server(vec![]).await;
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let start_id = mcp
+        .send_thread_start_request(ThreadStartParams {
+            model: Some("gpt-5-codex".to_string()),
+            ..Default::default()
+        })
+        .await?;
+    let start_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(start_id)),
+    )
+    .await??;
+    let ThreadStartResponse { thread } = to_response::<ThreadStartResponse>(start_resp)?;
+
+    let thread_path = thread.path.clone();
+    let resume_id = mcp
+        .send_thread_resume_request(ThreadResumeParams {
+            thread_id: "not-a-valid-thread-id".to_string(),
+            path: Some(thread_path),
+            ..Default::default()
+        })
+        .await?;
+
+    let resume_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(resume_id)),
+    )
+    .await??;
+    let ThreadResumeResponse { thread: resumed } =
+        to_response::<ThreadResumeResponse>(resume_resp)?;
+    assert_eq!(resumed, thread);
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn thread_resume_supports_history_and_overrides() -> Result<()> {
+    let server = create_mock_chat_completions_server(vec![]).await;
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    // Start a thread.
+    let start_id = mcp
+        .send_thread_start_request(ThreadStartParams {
+            model: Some("gpt-5-codex".to_string()),
+            ..Default::default()
+        })
+        .await?;
+    let start_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(start_id)),
+    )
+    .await??;
+    let ThreadStartResponse { thread } = to_response::<ThreadStartResponse>(start_resp)?;
+
+    let history_text = "Hello from history";
+    let history = vec![ResponseItem::Message {
+        id: None,
+        role: "user".to_string(),
+        content: vec![ContentItem::InputText {
+            text: history_text.to_string(),
+        }],
+    }];
+
+    // Resume with explicit history and override the model.
+    let resume_id = mcp
+        .send_thread_resume_request(ThreadResumeParams {
+            thread_id: thread.id,
+            history: Some(history),
+            model: Some("mock-model".to_string()),
+            model_provider: Some("mock_provider".to_string()),
+            ..Default::default()
+        })
+        .await?;
+    let resume_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(resume_id)),
+    )
+    .await??;
+    let ThreadResumeResponse { thread: resumed } =
+        to_response::<ThreadResumeResponse>(resume_resp)?;
+    assert!(!resumed.id.is_empty());
+    assert_eq!(resumed.model_provider, "mock_provider");
+    assert_eq!(resumed.preview, history_text);
+
+    Ok(())
+}
+
 // Helper to create a config.toml pointing at the mock model server.
 fn create_config_toml(codex_home: &std::path::Path, server_uri: &str) -> std::io::Result<()> {
     let config_toml = codex_home.join("config.toml");

codex-rs/cloud-tasks/src/lib.rs

@@ -1721,6 +1721,7 @@ mod tests {
     use ratatui::layout::Rect;
 
     #[test]
+    #[ignore = "very slow"]
     fn composer_input_renders_typed_characters() {
         let mut composer = ComposerInput::new();
         let key = KeyEvent::new(KeyCode::Char('a'), KeyModifiers::NONE);

codex-rs/common/Cargo.toml

@@ -8,9 +8,10 @@ workspace = true
 
 [dependencies]
 clap = { workspace = true, features = ["derive", "wrap_help"], optional = true }
+codex-app-server-protocol = { workspace = true }
 codex-core = { workspace = true }
 codex-protocol = { workspace = true }
-codex-app-server-protocol = { workspace = true }
+once_cell = { workspace = true }
 serde = { workspace = true, optional = true }
 toml = { workspace = true, optional = true }
 

codex-rs/common/src/model_presets.rs

@@ -1,5 +1,8 @@
+use std::collections::HashMap;
+
 use codex_app_server_protocol::AuthMode;
 use codex_core::protocol_config_types::ReasoningEffort;
+use once_cell::sync::Lazy;
 
 /// A reasoning effort option that can be surfaced for a model.
 #[derive(Debug, Clone, Copy)]
@@ -10,8 +13,14 @@ pub struct ReasoningEffortPreset {
     pub description: &'static str,
 }
 
+#[derive(Debug, Clone)]
+pub struct ModelUpgrade {
+    pub id: &'static str,
+    pub reasoning_effort_mapping: Option<HashMap<ReasoningEffort, ReasoningEffort>>,
+}
+
 /// Metadata describing a Codex-supported model.
-#[derive(Debug, Clone, Copy)]
+#[derive(Debug, Clone)]
 pub struct ModelPreset {
     /// Stable identifier for the preset.
     pub id: &'static str,
@@ -27,86 +36,175 @@ pub struct ModelPreset {
     pub supported_reasoning_efforts: &'static [ReasoningEffortPreset],
     /// Whether this is the default model for new users.
     pub is_default: bool,
+    /// recommended upgrade model
+    pub upgrade: Option<ModelUpgrade>,
 }
 
-const PRESETS: &[ModelPreset] = &[
-    ModelPreset {
-        id: "gpt-5-codex",
-        model: "gpt-5-codex",
-        display_name: "gpt-5-codex",
-        description: "Optimized for codex.",
-        default_reasoning_effort: ReasoningEffort::Medium,
-        supported_reasoning_efforts: &[
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Low,
-                description: "Fastest responses with limited reasoning",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Medium,
-                description: "Dynamically adjusts reasoning based on the task",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::High,
-                description: "Maximizes reasoning depth for complex or ambiguous problems",
-            },
-        ],
-        is_default: true,
-    },
-    ModelPreset {
-        id: "gpt-5-codex-mini",
-        model: "gpt-5-codex-mini",
-        display_name: "gpt-5-codex-mini",
-        description: "Optimized for codex. Cheaper, faster, but less capable.",
-        default_reasoning_effort: ReasoningEffort::Medium,
-        supported_reasoning_efforts: &[
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Medium,
-                description: "Dynamically adjusts reasoning based on the task",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::High,
-                description: "Maximizes reasoning depth for complex or ambiguous problems",
-            },
-        ],
-        is_default: false,
-    },
-    ModelPreset {
-        id: "gpt-5",
-        model: "gpt-5",
-        display_name: "gpt-5",
-        description: "Broad world knowledge with strong general reasoning.",
-        default_reasoning_effort: ReasoningEffort::Medium,
-        supported_reasoning_efforts: &[
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Minimal,
-                description: "Fastest responses with little reasoning",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Low,
-                description: "Balances speed with some reasoning; useful for straightforward queries and short explanations",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::Medium,
-                description: "Provides a solid balance of reasoning depth and latency for general-purpose tasks",
-            },
-            ReasoningEffortPreset {
-                effort: ReasoningEffort::High,
-                description: "Maximizes reasoning depth for complex or ambiguous problems",
-            },
-        ],
-        is_default: false,
-    },
-];
+static PRESETS: Lazy<Vec<ModelPreset>> = Lazy::new(|| {
+    vec![
+        ModelPreset {
+            id: "gpt-5.1-codex",
+            model: "gpt-5.1-codex",
+            display_name: "gpt-5.1-codex",
+            description: "Optimized for codex.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Low,
+                    description: "Fastest responses with limited reasoning",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Dynamically adjusts reasoning based on the task",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: true,
+            upgrade: None,
+        },
+        ModelPreset {
+            id: "gpt-5.1-codex-mini",
+            model: "gpt-5.1-codex-mini",
+            display_name: "gpt-5.1-codex-mini",
+            description: "Optimized for codex. Cheaper, faster, but less capable.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Dynamically adjusts reasoning based on the task",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: false,
+            upgrade: None,
+        },
+        ModelPreset {
+            id: "gpt-5.1",
+            model: "gpt-5.1",
+            display_name: "gpt-5.1",
+            description: "Broad world knowledge with strong general reasoning.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Low,
+                    description: "Balances speed with some reasoning; useful for straightforward queries and short explanations",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Provides a solid balance of reasoning depth and latency for general-purpose tasks",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: false,
+            upgrade: None,
+        },
+        // Deprecated models.
+        ModelPreset {
+            id: "gpt-5-codex",
+            model: "gpt-5-codex",
+            display_name: "gpt-5-codex",
+            description: "Optimized for codex.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Low,
+                    description: "Fastest responses with limited reasoning",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Dynamically adjusts reasoning based on the task",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: false,
+            upgrade: Some(ModelUpgrade {
+                id: "gpt-5.1-codex",
+                reasoning_effort_mapping: None,
+            }),
+        },
+        ModelPreset {
+            id: "gpt-5-codex-mini",
+            model: "gpt-5-codex-mini",
+            display_name: "gpt-5-codex-mini",
+            description: "Optimized for codex. Cheaper, faster, but less capable.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Dynamically adjusts reasoning based on the task",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: false,
+            upgrade: Some(ModelUpgrade {
+                id: "gpt-5.1-codex-mini",
+                reasoning_effort_mapping: None,
+            }),
+        },
+        ModelPreset {
+            id: "gpt-5",
+            model: "gpt-5",
+            display_name: "gpt-5",
+            description: "Broad world knowledge with strong general reasoning.",
+            default_reasoning_effort: ReasoningEffort::Medium,
+            supported_reasoning_efforts: &[
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Minimal,
+                    description: "Fastest responses with little reasoning",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Low,
+                    description: "Balances speed with some reasoning; useful for straightforward queries and short explanations",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::Medium,
+                    description: "Provides a solid balance of reasoning depth and latency for general-purpose tasks",
+                },
+                ReasoningEffortPreset {
+                    effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems",
+                },
+            ],
+            is_default: false,
+            upgrade: Some(ModelUpgrade {
+                id: "gpt-5.1",
+                reasoning_effort_mapping: Some(HashMap::from([(
+                    ReasoningEffort::Minimal,
+                    ReasoningEffort::Low,
+                )])),
+            }),
+        },
+    ]
+});
 
-pub fn builtin_model_presets(auth_mode: Option<AuthMode>) -> Vec<ModelPreset> {
-    let allow_codex_mini = matches!(auth_mode, Some(AuthMode::ChatGPT));
+pub fn builtin_model_presets(_auth_mode: Option<AuthMode>) -> Vec<ModelPreset> {
+    // leave auth mode for later use
     PRESETS
         .iter()
-        .filter(|preset| allow_codex_mini || preset.id != "gpt-5-codex-mini")
-        .copied()
+        .filter(|preset| preset.upgrade.is_none())
+        .cloned()
         .collect()
 }
 
+pub fn all_model_presets() -> &'static Vec<ModelPreset> {
+    &PRESETS
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;

codex-rs/core/Cargo.toml

@@ -60,6 +60,7 @@ shlex = { workspace = true }
 similar = { workspace = true }
 strum_macros = { workspace = true }
 tempfile = { workspace = true }
+test-case = "3.3.1"
 test-log = { workspace = true }
 thiserror = { workspace = true }
 time = { workspace = true, features = [

codex-rs/core/gpt_5_1_prompt.md

@@ -0,0 +1,370 @@
+You are GPT-5.1 running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
+
+Your capabilities:
+
+- Receive user prompts and other context provided by the harness, such as files in the workspace.
+- Communicate with the user by streaming thinking & responses, and by making & updating plans.
+- Emit function calls to run terminal commands and apply patches. Depending on how this specific run is configured, you can request that these function calls be escalated to the user for approval before running. More on this in the "Sandbox and approvals" section.
+
+Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
+
+# How you work
+
+## Personality
+
+Your default personality and tone is concise, direct, and friendly. You communicate efficiently, always keeping the user clearly informed about ongoing actions without unnecessary detail. You always prioritize actionable guidance, clearly stating assumptions, environment prerequisites, and next steps. Unless explicitly asked, you avoid excessively verbose explanations about your work.
+
+# AGENTS.md spec
+- Repos often contain AGENTS.md files. These files can appear anywhere within the repository.
+- These files are a way for humans to give you (the agent) instructions or tips for working within the container.
+- Some examples might be: coding conventions, info about how code is organized, or instructions for how to run or test code.
+- Instructions in AGENTS.md files:
+    - The scope of an AGENTS.md file is the entire directory tree rooted at the folder that contains it.
+    - For every file you touch in the final patch, you must obey instructions in any AGENTS.md file whose scope includes that file.
+    - Instructions about code style, structure, naming, etc. apply only to code within the AGENTS.md file's scope, unless the file states otherwise.
+    - More-deeply-nested AGENTS.md files take precedence in the case of conflicting instructions.
+    - Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
+- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
+
+## Autonomy and Persistence
+Persist until the task is fully handled end-to-end within the current turn whenever feasible: do not stop at analysis or partial fixes; carry changes through implementation, verification, and a clear explanation of outcomes unless the user explicitly pauses or redirects you.
+
+Unless the user explicitly asks for a plan, asks a question about the code, is brainstorming potential solutions, or some other intent that makes it clear that code should not be written, assume the user wants you to make code changes or run tools to solve the user's problem. In these cases, it's bad to output your proposed solution in a message, you should go ahead and actually implement the change. If you encounter challenges or blockers, you should attempt to resolve them yourself.
+
+## Responsiveness
+
+### User Updates Spec
+You'll work for stretches with tool calls — it's critical to keep the user updated as you work.
+
+Frequency & Length:
+- Send short updates (1–2 sentences) whenever there is a meaningful, important insight you need to share with the user to keep them informed.
+- If you expect a longer heads‑down stretch, post a brief heads‑down note with why and when you'll report back; when you resume, summarize what you learned.
+- Only the initial plan, plan updates, and final recap can be longer, with multiple bullets and paragraphs
+
+Tone:
+- Friendly, confident, senior-engineer energy. Positive, collaborative, humble; fix mistakes quickly.
+
+Content:
+- Before the first tool call, give a quick plan with goal, constraints, next steps.
+- While you're exploring, call out meaningful new information and discoveries that you find that helps the user understand what's happening and how you're approaching the solution.
+- If you change the plan (e.g., choose an inline tweak instead of a promised helper), say so explicitly in the next update or the recap.
+
+**Examples:**
+
+- “I’ve explored the repo; now checking the API route definitions.”
+- “Next, I’ll patch the config and update the related tests.”
+- “I’m about to scaffold the CLI commands and helper functions.”
+- “Ok cool, so I’ve wrapped my head around the repo. Now digging into the API routes.”
+- “Config’s looking tidy. Next up is patching helpers to keep things in sync.”
+- “Finished poking at the DB gateway. I will now chase down error handling.”
+- “Alright, build pipeline order is interesting. Checking how it reports failures.”
+- “Spotted a clever caching util; now hunting where it gets used.”
+
+## Planning
+
+You have access to an `update_plan` tool which tracks steps and progress and renders them to the user. Using the tool helps demonstrate that you've understood the task and convey how you're approaching it. Plans can help to make complex, ambiguous, or multi-phase work clearer and more collaborative for the user. A good plan should break the task into meaningful, logically ordered steps that are easy to verify as you go.
+
+Note that plans are not for padding out simple work with filler steps or stating the obvious. The content of your plan should not involve doing anything that you aren't capable of doing (i.e. don't try to test things that you can't test). Do not use plans for simple or single-step queries that you can just do or answer immediately.
+
+Do not repeat the full contents of the plan after an `update_plan` call — the harness already displays it. Instead, summarize the change made and highlight any important context or next step.
+
+Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
+
+Maintain statuses in the tool: exactly one item in_progress at a time; mark items complete when done; post timely status transitions. Do not jump an item from pending to completed: always set it to in_progress first. Do not batch-complete multiple items after the fact. Finish with all items completed or explicitly canceled/deferred before ending the turn. Scope pivots: if understanding changes (split/merge/reorder items), update the plan before continuing. Do not let the plan go stale while coding.
+
+Use a plan when:
+
+- The task is non-trivial and will require multiple actions over a long time horizon.
+- There are logical phases or dependencies where sequencing matters.
+- The work has ambiguity that benefits from outlining high-level goals.
+- You want intermediate checkpoints for feedback and validation.
+- When the user asked you to do more than one thing in a single prompt
+- The user has asked you to use the plan tool (aka "TODOs")
+- You generate additional steps while working, and plan to do them before yielding to the user
+
+### Examples
+
+**High-quality plans**
+
+Example 1:
+
+1. Add CLI entry with file args
+2. Parse Markdown via CommonMark library
+3. Apply semantic HTML template
+4. Handle code blocks, images, links
+5. Add error handling for invalid files
+
+Example 2:
+
+1. Define CSS variables for colors
+2. Add toggle with localStorage state
+3. Refactor components to use variables
+4. Verify all views for readability
+5. Add smooth theme-change transition
+
+Example 3:
+
+1. Set up Node.js + WebSocket server
+2. Add join/leave broadcast events
+3. Implement messaging with timestamps
+4. Add usernames + mention highlighting
+5. Persist messages in lightweight DB
+6. Add typing indicators + unread count
+
+**Low-quality plans**
+
+Example 1:
+
+1. Create CLI tool
+2. Add Markdown parser
+3. Convert to HTML
+
+Example 2:
+
+1. Add dark mode toggle
+2. Save preference
+3. Make styles look good
+
+Example 3:
+
+1. Create single-file HTML game
+2. Run quick sanity check
+3. Summarize usage instructions
+
+If you need to write a plan, only write high quality plans, not low quality ones.
+
+## Task execution
+
+You are a coding agent. You must keep going until the query or task is completely resolved, before ending your turn and yielding back to the user. Persist until the task is fully handled end-to-end within the current turn whenever feasible and persevere even when function calls fail. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
+
+You MUST adhere to the following criteria when solving queries:
+
+- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
+- Analyzing code for vulnerabilities is allowed.
+- Showing user code and tool call details is allowed.
+- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`). This is a FREEFORM tool, so do not wrap the patch in JSON.
+
+If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
+
+- Fix the problem at the root cause rather than applying surface-level patches, when possible.
+- Avoid unneeded complexity in your solution.
+- Do not attempt to fix unrelated bugs or broken tests. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+- Update documentation as necessary.
+- Keep changes consistent with the style of the existing codebase. Changes should be minimal and focused on the task.
+- Use `git log` and `git blame` to search the history of the codebase if additional context is required.
+- NEVER add copyright or license headers unless specifically requested.
+- Do not waste tokens by re-reading files after calling `apply_patch` on them. The tool call will fail if it didn't work. The same goes for making folders, deleting folders, etc.
+- Do not `git commit` your changes or create new git branches unless explicitly requested.
+- Do not add inline comments within code unless explicitly requested.
+- Do not use one-letter variable names unless explicitly requested.
+- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
+
+## Codex CLI harness, sandboxing, and approvals
+
+The Codex CLI harness supports several different configurations for sandboxing and escalation approvals that the user can choose from.
+
+Filesystem sandboxing defines which files can be read or written. The options for `sandbox_mode` are:
+- **read-only**: The sandbox only permits reading files.
+- **workspace-write**: The sandbox permits reading files, and editing files in `cwd` and `writable_roots`. Editing files in other directories requires approval.
+- **danger-full-access**: No filesystem sandboxing - all commands are permitted.
+
+Network sandboxing defines whether network can be accessed without approval. Options for `network_access` are:
+- **restricted**: Requires approval
+- **enabled**: No approval needed
+
+Approvals are your mechanism to get user consent to run shell commands without the sandbox. Possible configuration options for `approval_policy` are
+- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
+- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
+- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for escalating in the tool definition.)
+- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is paired with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
+
+When you are running with `approval_policy == on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
+- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
+- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
+- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
+- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters. Within this harness, prefer requesting approval via the tool over asking in natural language.
+- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
+- (for all of these, you should weigh alternative paths that do not require approval)
+
+When `sandbox_mode` is set to read-only, you'll need to request approval for any command that isn't a read.
+
+You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing enabled, and approval on-failure.
+
+Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
+
+When requesting approval to execute a command that will require escalated privileges:
+  - Provide the `with_escalated_permissions` parameter with the boolean value true
+  - Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
+
+## Validating your work
+
+If the codebase has tests or the ability to build or run, consider using them to verify changes once your work is complete.
+
+When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
+
+Similarly, once you're confident in correctness, you can suggest or use formatting commands to ensure that your code is well formatted. If there are issues you can iterate up to 3 times to get formatting right, but if you still can't manage it's better to save the user time and present them a correct solution where you call out the formatting in your final message. If the codebase does not have a formatter configured, do not add one.
+
+For all of testing, running, building, and formatting, do not attempt to fix unrelated bugs. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
+
+Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
+
+- When running in non-interactive approval modes like **never** or **on-failure**, you can proactively run tests, lint and do whatever you need to ensure you've completed the task. If you are unable to run tests, you must still do your utmost best to complete the task.
+- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
+- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
+
+## Ambition vs. precision
+
+For tasks that have no prior context (i.e. the user is starting something brand new), you should feel free to be ambitious and demonstrate creativity with your implementation.
+
+If you're operating in an existing codebase, you should make sure you do exactly what the user asks with surgical precision. Treat the surrounding codebase with respect, and don't overstep (i.e. changing filenames or variables unnecessarily). You should balance being sufficiently ambitious and proactive when completing tasks of this nature.
+
+You should use judicious initiative to decide on the right level of detail and complexity to deliver based on the user's needs. This means showing good judgment that you're capable of doing the right extras without gold-plating. This might be demonstrated by high-value, creative touches when scope of the task is vague; while being surgical and targeted when scope is tightly specified.
+
+## Sharing progress updates
+
+For especially longer tasks that you work on (i.e. requiring many tool calls, or a plan with multiple steps), you should provide progress updates back to the user at reasonable intervals. These updates should be structured as a concise sentence or two (no more than 8-10 words long) recapping progress so far in plain language: this update demonstrates your understanding of what needs to be done, progress so far (i.e. files explores, subtasks complete), and where you're going next.
+
+Before doing large chunks of work that may incur latency as experienced by the user (i.e. writing a new file), you should send a concise message to the user with an update indicating what you're about to do to ensure they know what you're spending time on. Don't start editing or writing large files before informing the user what you are doing and why.
+
+The messages you send before tool calls should describe what is immediately about to be done next in very concise language. If there was previous work done, this preamble message should also include a note about the work done so far to bring the user along.
+
+## Presenting your work and final message
+
+Your final message should read naturally, like an update from a concise teammate. For casual conversation, brainstorming tasks, or quick questions from the user, respond in a friendly, conversational tone. You should ask questions, suggest ideas, and adapt to the user’s style. If you've finished a large amount of work, when describing what you've done to the user, you should follow the final answer formatting guidelines to communicate substantive changes. You don't need to add structured formatting for one-word answers, greetings, or purely conversational exchanges.
+
+You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
+
+The user is working on the same computer as you, and has access to your work. As such there's no need to show the contents of files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
+
+If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If there’s something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
+
+Brevity is very important as a default. You should be very concise (i.e. no more than 10 lines), but can relax this requirement for tasks where additional detail and comprehensiveness is important for the user's understanding.
+
+### Final answer structure and style guidelines
+
+You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
+
+**Section Headers**
+
+- Use only when they improve clarity — they are not mandatory for every answer.
+- Choose descriptive names that fit the content
+- Keep headers short (1–3 words) and in `**Title Case**`. Always start headers with `**` and end with `**`
+- Leave no blank line before the first bullet under a header.
+- Section headers should only be used where they genuinely improve scanability; avoid fragmenting the answer.
+
+**Bullets**
+
+- Use `-` followed by a space for every bullet.
+- Merge related points when possible; avoid a bullet for every trivial detail.
+- Keep bullets to one line unless breaking for clarity is unavoidable.
+- Group into short lists (4–6 bullets) ordered by importance.
+- Use consistent keyword phrasing and formatting across sections.
+
+**Monospace**
+
+- Wrap all commands, file paths, env vars, code identifiers, and code samples in backticks (`` `...` ``).
+- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
+- Never mix monospace and bold markers; choose one based on whether it’s a keyword (`**`) or inline code/path (`` ` ``).
+
+**File References**
+When referencing files in your response, make sure to include the relevant start line and always follow the below rules:
+  * Use inline code to make file paths clickable.
+  * Each reference should have a stand alone path. Even if it's the same file.
+  * Accepted: absolute, workspace‑relative, a/ or b/ diff prefixes, or bare filename/suffix.
+  * Line/column (1‑based, optional): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
+  * Do not use URIs like file://, vscode://, or https://.
+  * Do not provide range of lines
+  * Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5
+
+**Structure**
+
+- Place related bullets together; don’t mix unrelated concepts in the same section.
+- Order sections from general → specific → supporting info.
+- For subsections (e.g., “Binaries” under “Rust Workspace”), introduce with a bolded keyword bullet, then list items under it.
+- Match structure to complexity:
+  - Multi-part or detailed results → use clear headers and grouped bullets.
+  - Simple results → minimal headers, possibly just a short list or paragraph.
+
+**Tone**
+
+- Keep the voice collaborative and natural, like a coding partner handing off work.
+- Be concise and factual — no filler or conversational commentary and avoid unnecessary repetition
+- Use present tense and active voice (e.g., “Runs tests” not “This will run tests”).
+- Keep descriptions self-contained; don’t refer to “above” or “below”.
+- Use parallel structure in lists for consistency.
+
+**Verbosity**
+- Final answer compactness rules (enforced):
+  - Tiny/small single-file change (≤ ~10 lines): 2–5 sentences or ≤3 bullets. No headings. 0–1 short snippet (≤3 lines) only if essential.
+  - Medium change (single area or a few files): ≤6 bullets or 6–10 sentences. At most 1–2 short snippets total (≤8 lines each).
+  - Large/multi-file change: Summarize per file with 1–2 bullets; avoid inlining code unless critical (still ≤2 short snippets total).
+  - Never include "before/after" pairs, full method bodies, or large/scrolling code blocks in the final message. Prefer referencing file/symbol names instead.
+
+**Don’t**
+
+- Don’t use literal words “bold” or “monospace” in the content.
+- Don’t nest bullets or create deep hierarchies.
+- Don’t output ANSI escape codes directly — the CLI renderer applies them.
+- Don’t cram unrelated keywords into a single bullet; split for clarity.
+- Don’t let keyword lists run long — wrap or reformat for scanability.
+
+Generally, ensure your final answers adapt their shape and depth to the request. For example, answers to code explanations should have a precise, structured explanation with code references that answer the question directly. For tasks with a simple implementation, lead with the outcome and supplement only with what’s needed for clarity. Larger changes can be presented as a logical walkthrough of your approach, grouping related steps, explaining rationale where it adds value, and highlighting next actions to accelerate the user. Your answers should provide the right level of detail while being easily scannable.
+
+For casual greetings, acknowledgements, or other one-off conversational messages that are not delivering substantive information or structured results, respond naturally without section headers or bullet formatting.
+
+# Tool Guidelines
+
+## Shell commands
+
+When using the shell, you must adhere to the following guidelines:
+
+- The arguments to `shell` will be passed to execvp().
+- Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary.
+- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
+- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
+
+## apply_patch
+
+Use the `apply_patch` tool to edit files. Your patch language is a stripped‑down, file‑oriented diff format designed to be easy to parse and safe to apply. You can think of it as a high‑level envelope:
+
+*** Begin Patch
+[ one or more file sections ]
+*** End Patch
+
+Within that envelope, you get a sequence of file operations.
+You MUST include a header to specify the action you are taking.
+Each operation starts with one of three headers:
+
+*** Add File: <path> - create a new file. Every following line is a + line (the initial contents).
+*** Delete File: <path> - remove an existing file. Nothing follows.
+*** Update File: <path> - patch an existing file in place (optionally with a rename).
+
+Example patch:
+
+```
+*** Begin Patch
+*** Add File: hello.txt
++Hello world
+*** Update File: src/app.py
+*** Move to: src/main.py
+@@ def greet():
+-print("Hi")
++print("Hello, world!")
+*** Delete File: obsolete.txt
+*** End Patch
+```
+
+It is important to remember:
+
+- You must include a header with your intended action (Add/Delete/Update)
+- You must prefix new lines with `+` even when creating a new file
+
+## `update_plan`
+
+A tool named `update_plan` is available to you. You can use it to keep an up‑to‑date, step‑by‑step plan for the task.
+
+To create a new plan, call `update_plan` with a short list of 1‑sentence steps (no more than 5-7 words each) with a `status` for each step (`pending`, `in_progress`, or `completed`).
+
+When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call.
+
+If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`.

codex-rs/core/src/auth.rs

@@ -227,6 +227,14 @@ impl CodexAuth {
             })
     }
 
+    /// Raw plan string from the ID token (including unknown/new plan types).
+    pub fn raw_plan_type(&self) -> Option<String> {
+        self.get_plan_type().map(|plan| match plan {
+            InternalPlanType::Known(k) => format!("{k:?}"),
+            InternalPlanType::Unknown(raw) => raw,
+        })
+    }
+
     /// Raw internal plan value from the ID token.
     /// Exposes the underlying `token_data::PlanType` without mapping it to the
     /// public `AccountPlanType`. Use this when downstream code needs to inspect
@@ -335,7 +343,10 @@ pub fn save_auth(
 }
 
 /// Load CLI auth data using the configured credential store backend.
-/// Returns `None` when no credentials are stored.
+/// Returns `None` when no credentials are stored. This function is
+/// provided only for tests. Production code should not directly load
+/// from the auth.json storage. It should use the AuthManager abstraction
+/// instead.
 pub fn load_auth_dot_json(
     codex_home: &Path,
     auth_credentials_store_mode: AuthCredentialsStoreMode,

codex-rs/core/src/bash.rs

@@ -1,8 +1,13 @@
+use std::path::PathBuf;
+
 use tree_sitter::Node;
 use tree_sitter::Parser;
 use tree_sitter::Tree;
 use tree_sitter_bash::LANGUAGE as BASH;
 
+use crate::shell::ShellType;
+use crate::shell::detect_shell_type;
+
 /// Parse the provided bash source using tree-sitter-bash, returning a Tree on
 /// success or None if parsing failed.
 pub fn try_parse_shell(shell_lc_arg: &str) -> Option<Tree> {
@@ -88,23 +93,16 @@ pub fn try_parse_word_only_commands_sequence(tree: &Tree, src: &str) -> Option<V
     Some(commands)
 }
 
-pub fn is_well_known_sh_shell(shell: &str) -> bool {
-    if shell == "bash" || shell == "zsh" {
-        return true;
-    }
-
-    let shell_name = std::path::Path::new(shell)
-        .file_name()
-        .and_then(|s| s.to_str())
-        .unwrap_or(shell);
-    matches!(shell_name, "bash" | "zsh")
-}
-
 pub fn extract_bash_command(command: &[String]) -> Option<(&str, &str)> {
     let [shell, flag, script] = command else {
         return None;
     };
-    if !matches!(flag.as_str(), "-lc" | "-c") || !is_well_known_sh_shell(shell) {
+    if !matches!(flag.as_str(), "-lc" | "-c")
+        || !matches!(
+            detect_shell_type(&PathBuf::from(shell)),
+            Some(ShellType::Zsh) | Some(ShellType::Bash)
+        )
+    {
         return None;
     }
     Some((shell, script))

codex-rs/core/src/chat_completions.rs

@@ -338,7 +338,7 @@ pub(crate) async fn stream_chat_completions(
     debug!(
         "POST to {}: {}",
         provider.get_full_url(&None),
-        serde_json::to_string_pretty(&payload).unwrap_or_default()
+        payload.to_string()
     );
 
     let mut attempt = 0;
@@ -477,10 +477,14 @@ async fn append_reasoning_text(
         ..
     }) = reasoning_item
     {
+        let content_index = content.len() as i64;
         content.push(ReasoningItemContent::ReasoningText { text: text.clone() });
 
         let _ = tx_event
-            .send(Ok(ResponseEvent::ReasoningContentDelta(text.clone())))
+            .send(Ok(ResponseEvent::ReasoningContentDelta {
+                delta: text.clone(),
+                content_index,
+            }))
             .await;
     }
 }
@@ -898,20 +902,26 @@ where
                         continue;
                     }
                 }
-                Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta)))) => {
+                Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta {
+                    delta,
+                    content_index,
+                }))) => {
                     // Always accumulate reasoning deltas so we can emit a final Reasoning item at Completed.
                     this.cumulative_reasoning.push_str(&delta);
                     if matches!(this.mode, AggregateMode::Streaming) {
                         // In streaming mode, also forward the delta immediately.
-                        return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta))));
+                        return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta {
+                            delta,
+                            content_index,
+                        })));
                     } else {
                         continue;
                     }
                 }
-                Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => {
+                Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta { .. }))) => {
                     continue;
                 }
-                Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded))) => {
+                Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded { .. }))) => {
                     continue;
                 }
                 Poll::Ready(Some(Ok(ResponseEvent::OutputItemAdded(item)))) => {

codex-rs/core/src/client.rs

@@ -35,10 +35,10 @@ use crate::auth::RefreshTokenError;
 use crate::chat_completions::AggregateStreamExt;
 use crate::chat_completions::stream_chat_completions;
 use crate::client_common::Prompt;
+use crate::client_common::Reasoning;
 use crate::client_common::ResponseEvent;
 use crate::client_common::ResponseStream;
 use crate::client_common::ResponsesApiRequest;
-use crate::client_common::create_reasoning_param_for_request;
 use crate::client_common::create_text_param_for_request;
 use crate::config::Config;
 use crate::default_client::CodexHttpClient;
@@ -199,12 +199,18 @@ impl ModelClient {
         let auth_manager = self.auth_manager.clone();
 
         let full_instructions = prompt.get_full_instructions(&self.config.model_family);
-        let tools_json = create_tools_json_for_responses_api(&prompt.tools)?;
-        let reasoning = create_reasoning_param_for_request(
-            &self.config.model_family,
-            self.effort,
-            self.summary,
-        );
+        let tools_json: Vec<Value> = create_tools_json_for_responses_api(&prompt.tools)?;
+
+        let reasoning = if self.config.model_family.supports_reasoning_summaries {
+            Some(Reasoning {
+                effort: self
+                    .effort
+                    .or(self.config.model_family.default_reasoning_effort),
+                summary: Some(self.summary),
+            })
+        } else {
+            None
+        };
 
         let include: Vec<String> = if reasoning.is_some() {
             vec!["reasoning.encrypted_content".to_string()]
@@ -215,7 +221,9 @@ impl ModelClient {
         let input_with_instructions = prompt.get_formatted_input();
 
         let verbosity = if self.config.model_family.support_verbosity {
-            self.config.model_verbosity
+            self.config
+                .model_verbosity
+                .or(self.config.model_family.default_verbosity)
         } else {
             if self.config.model_verbosity.is_some() {
                 warn!(
@@ -294,10 +302,9 @@ impl ModelClient {
         let auth = auth_manager.as_ref().and_then(|m| m.auth());
 
         trace!(
-            "POST to {}: {:?}",
+            "POST to {}: {}",
             self.provider.get_full_url(&auth),
-            serde_json::to_string(payload_json)
-                .unwrap_or("<unable to serialize payload>".to_string())
+            payload_json.to_string()
         );
 
         let mut req_builder = self
@@ -553,6 +560,8 @@ struct SseEvent {
     response: Option<Value>,
     item: Option<Value>,
     delta: Option<String>,
+    summary_index: Option<i64>,
+    content_index: Option<i64>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -812,16 +821,22 @@ async fn process_sse<S>(
                 }
             }
             "response.reasoning_summary_text.delta" => {
-                if let Some(delta) = event.delta {
-                    let event = ResponseEvent::ReasoningSummaryDelta(delta);
+                if let (Some(delta), Some(summary_index)) = (event.delta, event.summary_index) {
+                    let event = ResponseEvent::ReasoningSummaryDelta {
+                        delta,
+                        summary_index,
+                    };
                     if tx_event.send(Ok(event)).await.is_err() {
                         return;
                     }
                 }
             }
             "response.reasoning_text.delta" => {
-                if let Some(delta) = event.delta {
-                    let event = ResponseEvent::ReasoningContentDelta(delta);
+                if let (Some(delta), Some(content_index)) = (event.delta, event.content_index) {
+                    let event = ResponseEvent::ReasoningContentDelta {
+                        delta,
+                        content_index,
+                    };
                     if tx_event.send(Ok(event)).await.is_err() {
                         return;
                     }
@@ -898,10 +913,12 @@ async fn process_sse<S>(
                 }
             }
             "response.reasoning_summary_part.added" => {
-                // Boundary between reasoning summary sections (e.g., titles).
-                let event = ResponseEvent::ReasoningSummaryPartAdded;
-                if tx_event.send(Ok(event)).await.is_err() {
-                    return;
+                if let Some(summary_index) = event.summary_index {
+                    // Boundary between reasoning summary sections (e.g., titles).
+                    let event = ResponseEvent::ReasoningSummaryPartAdded { summary_index };
+                    if tx_event.send(Ok(event)).await.is_err() {
+                        return;
+                    }
                 }
             }
             "response.reasoning_summary_text.done" => {}

codex-rs/core/src/client_common.rs

@@ -203,9 +203,17 @@ pub enum ResponseEvent {
         token_usage: Option<TokenUsage>,
     },
     OutputTextDelta(String),
-    ReasoningSummaryDelta(String),
-    ReasoningContentDelta(String),
-    ReasoningSummaryPartAdded,
+    ReasoningSummaryDelta {
+        delta: String,
+        summary_index: i64,
+    },
+    ReasoningContentDelta {
+        delta: String,
+        content_index: i64,
+    },
+    ReasoningSummaryPartAdded {
+        summary_index: i64,
+    },
     RateLimits(RateLimitSnapshot),
 }
 
@@ -342,21 +350,6 @@ pub(crate) mod tools {
     }
 }
 
-pub(crate) fn create_reasoning_param_for_request(
-    model_family: &ModelFamily,
-    effort: Option<ReasoningEffortConfig>,
-    summary: ReasoningSummaryConfig,
-) -> Option<Reasoning> {
-    if !model_family.supports_reasoning_summaries {
-        return None;
-    }
-
-    Some(Reasoning {
-        effort,
-        summary: Some(summary),
-    })
-}
-
 pub(crate) fn create_text_param_for_request(
     verbosity: Option<VerbosityConfig>,
     output_schema: &Option<Value>,
@@ -421,6 +414,10 @@ mod tests {
                 slug: "gpt-5",
                 expects_apply_patch_instructions: true,
             },
+            InstructionsTestCase {
+                slug: "gpt-5.1",
+                expects_apply_patch_instructions: false,
+            },
             InstructionsTestCase {
                 slug: "codex-mini-latest",
                 expects_apply_patch_instructions: true,
@@ -433,6 +430,10 @@ mod tests {
                 slug: "gpt-5-codex",
                 expects_apply_patch_instructions: false,
             },
+            InstructionsTestCase {
+                slug: "gpt-5.1-codex",
+                expects_apply_patch_instructions: false,
+            },
         ];
         for test_case in test_cases {
             let model_family = find_family_for_model(test_case.slug).expect("known model slug");

codex-rs/core/src/codex.rs

@@ -96,6 +96,7 @@ use crate::protocol::StreamErrorEvent;
 use crate::protocol::Submission;
 use crate::protocol::TokenCountEvent;
 use crate::protocol::TokenUsage;
+use crate::protocol::TokenUsageInfo;
 use crate::protocol::TurnDiffEvent;
 use crate::protocol::WarningEvent;
 use crate::rollout::RolloutRecorder;
@@ -132,6 +133,7 @@ use codex_protocol::protocol::InitialHistory;
 use codex_protocol::user_input::UserInput;
 use codex_utils_readiness::Readiness;
 use codex_utils_readiness::ReadinessFlag;
+use codex_utils_tokenizer::warm_model_cache;
 
 /// The high-level interface to the Codex system.
 /// It operates as a queue pair where you send submissions and receive events.
@@ -589,6 +591,9 @@ impl Session {
         // Create the mutable state for the Session.
         let state = SessionState::new(session_configuration.clone());
 
+        // Warm the tokenizer cache for the session model without blocking startup.
+        warm_model_cache(&session_configuration.model);
+
         let services = SessionServices {
             mcp_connection_manager,
             unified_exec_manager: UnifiedExecSessionManager::default(),
@@ -1082,6 +1087,36 @@ impl Session {
         self.send_token_count_event(turn_context).await;
     }
 
+    pub(crate) async fn override_last_token_usage_estimate(
+        &self,
+        turn_context: &TurnContext,
+        estimated_total_tokens: i64,
+    ) {
+        {
+            let mut state = self.state.lock().await;
+            let mut info = state.token_info().unwrap_or(TokenUsageInfo {
+                total_token_usage: TokenUsage::default(),
+                last_token_usage: TokenUsage::default(),
+                model_context_window: None,
+            });
+
+            info.last_token_usage = TokenUsage {
+                input_tokens: 0,
+                cached_input_tokens: 0,
+                output_tokens: 0,
+                reasoning_output_tokens: 0,
+                total_tokens: estimated_total_tokens.max(0),
+            };
+
+            if info.model_context_window.is_none() {
+                info.model_context_window = turn_context.client.get_model_context_window();
+            }
+
+            state.set_token_info(Some(info));
+        }
+        self.send_token_count_event(turn_context).await;
+    }
+
     pub(crate) async fn update_rate_limits(
         &self,
         turn_context: &TurnContext,
@@ -2201,13 +2236,17 @@ async fn try_run_turn(
                     error_or_panic("ReasoningSummaryDelta without active item".to_string());
                 }
             }
-            ResponseEvent::ReasoningSummaryDelta(delta) => {
+            ResponseEvent::ReasoningSummaryDelta {
+                delta,
+                summary_index,
+            } => {
                 if let Some(active) = active_item.as_ref() {
                     let event = ReasoningContentDeltaEvent {
                         thread_id: sess.conversation_id.to_string(),
                         turn_id: turn_context.sub_id.clone(),
                         item_id: active.id(),
-                        delta: delta.clone(),
+                        delta,
+                        summary_index,
                     };
                     sess.send_event(&turn_context, EventMsg::ReasoningContentDelta(event))
                         .await;
@@ -2215,18 +2254,29 @@ async fn try_run_turn(
                     error_or_panic("ReasoningSummaryDelta without active item".to_string());
                 }
             }
-            ResponseEvent::ReasoningSummaryPartAdded => {
-                let event =
-                    EventMsg::AgentReasoningSectionBreak(AgentReasoningSectionBreakEvent {});
-                sess.send_event(&turn_context, event).await;
+            ResponseEvent::ReasoningSummaryPartAdded { summary_index } => {
+                if let Some(active) = active_item.as_ref() {
+                    let event =
+                        EventMsg::AgentReasoningSectionBreak(AgentReasoningSectionBreakEvent {
+                            item_id: active.id(),
+                            summary_index,
+                        });
+                    sess.send_event(&turn_context, event).await;
+                } else {
+                    error_or_panic("ReasoningSummaryPartAdded without active item".to_string());
+                }
             }
-            ResponseEvent::ReasoningContentDelta(delta) => {
+            ResponseEvent::ReasoningContentDelta {
+                delta,
+                content_index,
+            } => {
                 if let Some(active) = active_item.as_ref() {
                     let event = ReasoningRawContentDeltaEvent {
                         thread_id: sess.conversation_id.to_string(),
                         turn_id: turn_context.sub_id.clone(),
                         item_id: active.id(),
-                        delta: delta.clone(),
+                        delta,
+                        content_index,
                     };
                     sess.send_event(&turn_context, EventMsg::ReasoningRawContentDelta(event))
                         .await;

codex-rs/core/src/command_safety/is_safe_command.rs

@@ -1,4 +1,5 @@
 use crate::bash::parse_shell_lc_plain_commands;
+use crate::command_safety::windows_safe_commands::is_safe_command_windows;
 
 pub fn is_known_safe_command(command: &[String]) -> bool {
     let command: Vec<String> = command
@@ -11,12 +12,9 @@ pub fn is_known_safe_command(command: &[String]) -> bool {
             }
         })
         .collect();
-    #[cfg(target_os = "windows")]
-    {
-        use super::windows_safe_commands::is_safe_command_windows;
-        if is_safe_command_windows(&command) {
-            return true;
-        }
+
+    if is_safe_command_windows(&command) {
+        return true;
     }
 
     if is_safe_to_call_with_exec(&command) {

codex-rs/core/src/command_safety/mod.rs

@@ -1,4 +1,3 @@
 pub mod is_dangerous_command;
 pub mod is_safe_command;
-#[cfg(target_os = "windows")]
 pub mod windows_safe_commands;

codex-rs/core/src/compact.rs

@@ -153,6 +153,15 @@ async fn run_compact_task_inner(
     new_history.extend(ghost_snapshots);
     sess.replace_history(new_history).await;
 
+    if let Some(estimated_tokens) = sess
+        .clone_history()
+        .await
+        .estimate_token_count(&turn_context)
+    {
+        sess.override_last_token_usage_estimate(&turn_context, estimated_tokens)
+            .await;
+    }
+
     let rollout_item = RolloutItem::Compacted(CompactedItem {
         message: summary_text.clone(),
     });

codex-rs/core/src/config/edit.rs

@@ -3,6 +3,8 @@ use crate::config::types::McpServerConfig;
 use crate::config::types::Notice;
 use anyhow::Context;
 use codex_protocol::config_types::ReasoningEffort;
+use codex_protocol::config_types::TrustLevel;
+use codex_utils_tokenizer::warm_model_cache;
 use std::collections::BTreeMap;
 use std::path::Path;
 use std::path::PathBuf;
@@ -29,11 +31,13 @@ pub enum ConfigEdit {
     SetNoticeHideRateLimitModelNudge(bool),
     /// Toggle the Windows onboarding acknowledgement flag.
     SetWindowsWslSetupAcknowledged(bool),
+    /// Toggle the model migration prompt acknowledgement flag.
+    SetNoticeHideModelMigrationPrompt(String, bool),
     /// Replace the entire `[mcp_servers]` table.
     ReplaceMcpServers(BTreeMap<String, McpServerConfig>),
-    /// Set trust_level = "trusted" under `[projects."<path>"]`,
+    /// Set trust_level under `[projects."<path>"]`,
     /// migrating inline tables to explicit tables.
-    SetProjectTrusted(PathBuf),
+    SetProjectTrustLevel { path: PathBuf, level: TrustLevel },
     /// Set the value stored at the exact dotted path.
     SetPath {
         segments: Vec<String>,
@@ -227,6 +231,9 @@ impl ConfigDocument {
     fn apply(&mut self, edit: &ConfigEdit) -> anyhow::Result<bool> {
         match edit {
             ConfigEdit::SetModel { model, effort } => Ok({
+                if let Some(model) = &model {
+                    warm_model_cache(model)
+                }
                 let mut mutated = false;
                 mutated |= self.write_profile_value(
                     &["model"],
@@ -253,6 +260,13 @@ impl ConfigDocument {
                 &[Notice::TABLE_KEY, "hide_rate_limit_model_nudge"],
                 value(*acknowledged),
             )),
+            ConfigEdit::SetNoticeHideModelMigrationPrompt(migration_config, acknowledged) => {
+                Ok(self.write_value(
+                    Scope::Global,
+                    &[Notice::TABLE_KEY, migration_config.as_str()],
+                    value(*acknowledged),
+                ))
+            }
             ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged) => Ok(self.write_value(
                 Scope::Global,
                 &["windows_wsl_setup_acknowledged"],
@@ -261,10 +275,14 @@ impl ConfigDocument {
             ConfigEdit::ReplaceMcpServers(servers) => Ok(self.replace_mcp_servers(servers)),
             ConfigEdit::SetPath { segments, value } => Ok(self.insert(segments, value.clone())),
             ConfigEdit::ClearPath { segments } => Ok(self.clear_owned(segments)),
-            ConfigEdit::SetProjectTrusted(project_path) => {
+            ConfigEdit::SetProjectTrustLevel { path, level } => {
                 // Delegate to the existing, tested logic in config.rs to
                 // ensure tables are explicit and migration is preserved.
-                crate::config::set_project_trusted_inner(&mut self.doc, project_path.as_path())?;
+                crate::config::set_project_trust_level_inner(
+                    &mut self.doc,
+                    path.as_path(),
+                    *level,
+                )?;
                 Ok(true)
             }
         }
@@ -499,6 +517,15 @@ impl ConfigEditsBuilder {
         self
     }
 
+    pub fn set_hide_model_migration_prompt(mut self, model: &str, acknowledged: bool) -> Self {
+        self.edits
+            .push(ConfigEdit::SetNoticeHideModelMigrationPrompt(
+                model.to_string(),
+                acknowledged,
+            ));
+        self
+    }
+
     pub fn set_windows_wsl_setup_acknowledged(mut self, acknowledged: bool) -> Self {
         self.edits
             .push(ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged));
@@ -511,9 +538,15 @@ impl ConfigEditsBuilder {
         self
     }
 
-    pub fn set_project_trusted<P: Into<PathBuf>>(mut self, project_path: P) -> Self {
-        self.edits
-            .push(ConfigEdit::SetProjectTrusted(project_path.into()));
+    pub fn set_project_trust_level<P: Into<PathBuf>>(
+        mut self,
+        project_path: P,
+        trust_level: TrustLevel,
+    ) -> Self {
+        self.edits.push(ConfigEdit::SetProjectTrustLevel {
+            path: project_path.into(),
+            level: trust_level,
+        });
         self
     }
 
@@ -770,6 +803,35 @@ existing = "value"
         let expected = r#"[notice]
 existing = "value"
 hide_rate_limit_model_nudge = true
+"#;
+        assert_eq!(contents, expected);
+    }
+    #[test]
+    fn blocking_set_hide_gpt5_1_migration_prompt_preserves_table() {
+        let tmp = tempdir().expect("tmpdir");
+        let codex_home = tmp.path();
+        std::fs::write(
+            codex_home.join(CONFIG_TOML_FILE),
+            r#"[notice]
+existing = "value"
+"#,
+        )
+        .expect("seed");
+        apply_blocking(
+            codex_home,
+            None,
+            &[ConfigEdit::SetNoticeHideModelMigrationPrompt(
+                "hide_gpt5_1_migration_prompt".to_string(),
+                true,
+            )],
+        )
+        .expect("persist");
+
+        let contents =
+            std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
+        let expected = r#"[notice]
+existing = "value"
+hide_gpt5_1_migration_prompt = true
 "#;
         assert_eq!(contents, expected);
     }

codex-rs/core/src/config/mod.rs

@@ -38,6 +38,7 @@ use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::config_types::SandboxMode;
+use codex_protocol::config_types::TrustLevel;
 use codex_protocol::config_types::Verbosity;
 use codex_rmcp_client::OAuthCredentialsStoreMode;
 use dirs::home_dir;
@@ -382,15 +383,16 @@ fn ensure_no_inline_bearer_tokens(value: &TomlValue) -> std::io::Result<()> {
     Ok(())
 }
 
-pub(crate) fn set_project_trusted_inner(
+pub(crate) fn set_project_trust_level_inner(
     doc: &mut DocumentMut,
     project_path: &Path,
+    trust_level: TrustLevel,
 ) -> anyhow::Result<()> {
     // Ensure we render a human-friendly structure:
     //
     // [projects]
     // [projects."/path/to/project"]
-    // trust_level = "trusted"
+    // trust_level = "trusted" or "untrusted"
     //
     // rather than inline tables like:
     //
@@ -446,17 +448,21 @@ pub(crate) fn set_project_trusted_inner(
         return Err(anyhow::anyhow!("project table missing for {project_key}"));
     };
     proj_tbl.set_implicit(false);
-    proj_tbl["trust_level"] = toml_edit::value("trusted");
+    proj_tbl["trust_level"] = toml_edit::value(trust_level.to_string());
     Ok(())
 }
 
-/// Patch `CODEX_HOME/config.toml` project state.
+/// Patch `CODEX_HOME/config.toml` project state to set trust level.
 /// Use with caution.
-pub fn set_project_trusted(codex_home: &Path, project_path: &Path) -> anyhow::Result<()> {
+pub fn set_project_trust_level(
+    codex_home: &Path,
+    project_path: &Path,
+    trust_level: TrustLevel,
+) -> anyhow::Result<()> {
     use crate::config::edit::ConfigEditsBuilder;
 
     ConfigEditsBuilder::new(codex_home)
-        .set_project_trusted(project_path)
+        .set_project_trust_level(project_path, trust_level)
         .apply_blocking()
 }
 
@@ -686,15 +692,16 @@ impl From<ConfigToml> for UserSavedConfig {
 
 #[derive(Deserialize, Debug, Clone, PartialEq, Eq)]
 pub struct ProjectConfig {
-    pub trust_level: Option<String>,
+    pub trust_level: Option<TrustLevel>,
 }
 
 impl ProjectConfig {
     pub fn is_trusted(&self) -> bool {
-        match &self.trust_level {
-            Some(trust_level) => trust_level == "trusted",
-            None => false,
-        }
+        matches!(self.trust_level, Some(TrustLevel::Trusted))
+    }
+
+    pub fn is_untrusted(&self) -> bool {
+        matches!(self.trust_level, Some(TrustLevel::Untrusted))
     }
 }
 
@@ -735,9 +742,9 @@ impl ConfigToml {
             .or(profile_sandbox_mode)
             .or(self.sandbox_mode)
             .or_else(|| {
-                // if no sandbox_mode is set, but user has marked directory as trusted, use WorkspaceWrite
+                // if no sandbox_mode is set, but user has marked directory as trusted or untrusted, use WorkspaceWrite
                 self.get_active_project(resolved_cwd).and_then(|p| {
-                    if p.is_trusted() {
+                    if p.is_trusted() || p.is_untrusted() {
                         Some(SandboxMode::WorkspaceWrite)
                     } else {
                         None
@@ -958,6 +965,9 @@ impl Config {
                 if active_project.is_trusted() {
                     // If no explicit approval policy is set, but we trust cwd, default to OnRequest
                     AskForApproval::OnRequest
+                } else if active_project.is_untrusted() {
+                    // If project is explicitly marked untrusted, require approval for non-safe commands
+                    AskForApproval::UnlessTrusted
                 } else {
                     AskForApproval::default()
                 }
@@ -1323,7 +1333,7 @@ persistence = "none"
     }
 
     #[test]
-    fn tui_config_missing_notifications_field_defaults_to_disabled() {
+    fn tui_config_missing_notifications_field_defaults_to_enabled() {
         let cfg = r#"
 [tui]
 "#;
@@ -1332,7 +1342,7 @@ persistence = "none"
             .expect("TUI config without notifications should succeed");
         let tui = parsed.tui.expect("config should include tui section");
 
-        assert_eq!(tui.notifications, Notifications::Enabled(false));
+        assert_eq!(tui.notifications, Notifications::Enabled(true));
     }
 
     #[test]
@@ -3164,7 +3174,7 @@ model_verbosity = "high"
         let project_dir = Path::new("/some/path");
         let mut doc = DocumentMut::new();
 
-        set_project_trusted_inner(&mut doc, project_dir)?;
+        set_project_trust_level_inner(&mut doc, project_dir, TrustLevel::Trusted)?;
 
         let contents = doc.to_string();
 
@@ -3204,7 +3214,7 @@ trust_level = "trusted"
         let mut doc = initial.parse::<DocumentMut>()?;
 
         // Run the function; it should convert to explicit tables and set trusted
-        set_project_trusted_inner(&mut doc, project_dir)?;
+        set_project_trust_level_inner(&mut doc, project_dir, TrustLevel::Trusted)?;
 
         let contents = doc.to_string();
 
@@ -3231,7 +3241,7 @@ model = "foo""#;
 
         // Approve a new directory
         let new_project = Path::new("/Users/mbolin/code/codex2");
-        set_project_trusted_inner(&mut doc, new_project)?;
+        set_project_trust_level_inner(&mut doc, new_project, TrustLevel::Trusted)?;
 
         let contents = doc.to_string();
 
@@ -3254,6 +3264,87 @@ trust_level = "trusted"
 
         Ok(())
     }
+
+    #[test]
+    fn test_untrusted_project_gets_workspace_write_sandbox() -> anyhow::Result<()> {
+        let config_with_untrusted = r#"
+[projects."/tmp/test"]
+trust_level = "untrusted"
+"#;
+
+        let cfg = toml::from_str::<ConfigToml>(config_with_untrusted)
+            .expect("TOML deserialization should succeed");
+
+        let resolution = cfg.derive_sandbox_policy(None, None, &PathBuf::from("/tmp/test"));
+
+        // Verify that untrusted projects get WorkspaceWrite (or ReadOnly on Windows due to downgrade)
+        if cfg!(target_os = "windows") {
+            assert!(
+                matches!(resolution.policy, SandboxPolicy::ReadOnly),
+                "Expected ReadOnly on Windows, got {:?}",
+                resolution.policy
+            );
+        } else {
+            assert!(
+                matches!(resolution.policy, SandboxPolicy::WorkspaceWrite { .. }),
+                "Expected WorkspaceWrite for untrusted project, got {:?}",
+                resolution.policy
+            );
+        }
+
+        Ok(())
+    }
+
+    #[test]
+    fn test_untrusted_project_gets_unless_trusted_approval_policy() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let test_project_dir = TempDir::new()?;
+        let test_path = test_project_dir.path();
+
+        let mut projects = std::collections::HashMap::new();
+        projects.insert(
+            test_path.to_string_lossy().to_string(),
+            ProjectConfig {
+                trust_level: Some(TrustLevel::Untrusted),
+            },
+        );
+
+        let cfg = ConfigToml {
+            projects: Some(projects),
+            ..Default::default()
+        };
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides {
+                cwd: Some(test_path.to_path_buf()),
+                ..Default::default()
+            },
+            codex_home.path().to_path_buf(),
+        )?;
+
+        // Verify that untrusted projects get UnlessTrusted approval policy
+        assert_eq!(
+            config.approval_policy,
+            AskForApproval::UnlessTrusted,
+            "Expected UnlessTrusted approval policy for untrusted project"
+        );
+
+        // Verify that untrusted projects still get WorkspaceWrite sandbox (or ReadOnly on Windows)
+        if cfg!(target_os = "windows") {
+            assert!(
+                matches!(config.sandbox_policy, SandboxPolicy::ReadOnly),
+                "Expected ReadOnly on Windows"
+            );
+        } else {
+            assert!(
+                matches!(config.sandbox_policy, SandboxPolicy::WorkspaceWrite { .. }),
+                "Expected WorkspaceWrite sandbox for untrusted project"
+            );
+        }
+
+        Ok(())
+    }
 }
 
 #[cfg(test)]

codex-rs/core/src/config/types.rs

@@ -338,7 +338,7 @@ pub enum Notifications {
 
 impl Default for Notifications {
     fn default() -> Self {
-        Self::Enabled(false)
+        Self::Enabled(true)
     }
 }
 
@@ -346,7 +346,7 @@ impl Default for Notifications {
 #[derive(Deserialize, Debug, Clone, PartialEq, Default)]
 pub struct Tui {
     /// Enable desktop notifications from the TUI when the terminal is unfocused.
-    /// Defaults to `false`.
+    /// Defaults to `true`.
     #[serde(default)]
     pub notifications: Notifications,
 }
@@ -362,6 +362,8 @@ pub struct Notice {
     pub hide_world_writable_warning: Option<bool>,
     /// Tracks whether the user opted out of the rate limit model switch reminder.
     pub hide_rate_limit_model_nudge: Option<bool>,
+    /// Tracks whether the user has seen the model migration prompt
+    pub hide_gpt5_1_migration_prompt: Option<bool>,
 }
 
 impl Notice {

codex-rs/core/src/context_manager/history.rs

@@ -1,12 +1,20 @@
+use crate::codex::TurnContext;
+use crate::context_manager::normalize;
+use crate::context_manager::truncate;
+use crate::context_manager::truncate::format_output_for_model_body;
+use crate::context_manager::truncate::globally_truncate_function_output_items;
 use codex_protocol::models::FunctionCallOutputPayload;
 use codex_protocol::models::ResponseItem;
 use codex_protocol::protocol::TokenUsage;
 use codex_protocol::protocol::TokenUsageInfo;
+use codex_utils_tokenizer::Tokenizer;
 use std::ops::Deref;
 
-use crate::context_manager::normalize;
-use crate::context_manager::truncate::format_output_for_model_body;
-use crate::context_manager::truncate::globally_truncate_function_output_items;
+const CONTEXT_WINDOW_HARD_LIMIT_FACTOR: f64 = 1.1;
+const CONTEXT_WINDOW_HARD_LIMIT_BYTES: usize =
+    (truncate::MODEL_FORMAT_MAX_BYTES as f64 * CONTEXT_WINDOW_HARD_LIMIT_FACTOR) as usize;
+const CONTEXT_WINDOW_HARD_LIMIT_LINES: usize =
+    (truncate::MODEL_FORMAT_MAX_LINES as f64 * CONTEXT_WINDOW_HARD_LIMIT_FACTOR) as usize;
 
 /// Transcript of conversation history
 #[derive(Debug, Clone, Default)]
@@ -28,6 +36,10 @@ impl ContextManager {
         self.token_info.clone()
     }
 
+    pub(crate) fn set_token_info(&mut self, info: Option<TokenUsageInfo>) {
+        self.token_info = info;
+    }
+
     pub(crate) fn set_token_usage_full(&mut self, context_window: i64) {
         match &mut self.token_info {
             Some(info) => info.fill_to_context_window(context_window),
@@ -68,6 +80,28 @@ impl ContextManager {
         history
     }
 
+    // Estimate the number of tokens in the history. Return None if no tokenizer
+    // is available. This does not consider the reasoning traces.
+    // /!\ The value is a lower bound estimate and does not represent the exact
+    // context length.
+    pub(crate) fn estimate_token_count(&self, turn_context: &TurnContext) -> Option<i64> {
+        let model = turn_context.client.get_model();
+        let tokenizer = Tokenizer::for_model(model.as_str()).ok()?;
+        let model_family = turn_context.client.get_model_family();
+
+        Some(
+            self.items
+                .iter()
+                .map(|item| {
+                    serde_json::to_string(&item)
+                        .map(|item| tokenizer.count(&item))
+                        .unwrap_or_default()
+                })
+                .sum::<i64>()
+                + tokenizer.count(model_family.base_instructions.as_str()),
+        )
+    }
+
     pub(crate) fn remove_first_item(&mut self) {
         if !self.items.is_empty() {
             // Remove the oldest item (front of the list). Items are ordered from
@@ -119,7 +153,11 @@ impl ContextManager {
     fn process_item(item: &ResponseItem) -> ResponseItem {
         match item {
             ResponseItem::FunctionCallOutput { call_id, output } => {
-                let truncated = format_output_for_model_body(output.content.as_str());
+                let truncated = format_output_for_model_body(
+                    output.content.as_str(),
+                    CONTEXT_WINDOW_HARD_LIMIT_BYTES,
+                    CONTEXT_WINDOW_HARD_LIMIT_LINES,
+                );
                 let truncated_items = output
                     .content_items
                     .as_ref()
@@ -134,7 +172,11 @@ impl ContextManager {
                 }
             }
             ResponseItem::CustomToolCallOutput { call_id, output } => {
-                let truncated = format_output_for_model_body(output);
+                let truncated = format_output_for_model_body(
+                    output,
+                    CONTEXT_WINDOW_HARD_LIMIT_BYTES,
+                    CONTEXT_WINDOW_HARD_LIMIT_LINES,
+                );
                 ResponseItem::CustomToolCallOutput {
                     call_id: call_id.clone(),
                     output: truncated,

codex-rs/core/src/context_manager/history_tests.rs

@@ -1,4 +1,5 @@
 use super::*;
+use crate::context_manager::MODEL_FORMAT_MAX_LINES;
 use crate::context_manager::truncate;
 use codex_git::GhostCommit;
 use codex_protocol::models::ContentItem;
@@ -308,8 +309,10 @@ fn assert_truncated_message_matches(message: &str, line: &str, total_lines: usiz
 }
 
 fn truncated_message_pattern(line: &str, total_lines: usize) -> String {
-    let head_take = truncate::MODEL_FORMAT_HEAD_LINES.min(total_lines);
-    let tail_take = truncate::MODEL_FORMAT_TAIL_LINES.min(total_lines.saturating_sub(head_take));
+    let head_lines = MODEL_FORMAT_MAX_LINES / 2;
+    let tail_lines = MODEL_FORMAT_MAX_LINES - head_lines;
+    let head_take = head_lines.min(total_lines);
+    let tail_take = tail_lines.min(total_lines.saturating_sub(head_take));
     let omitted = total_lines.saturating_sub(head_take + tail_take);
     let escaped_line = regex_lite::escape(line);
     if omitted == 0 {
@@ -328,7 +331,11 @@ fn format_exec_output_truncates_large_error() {
     let line = "very long execution error line that should trigger truncation\n";
     let large_error = line.repeat(2_500); // way beyond both byte and line limits
 
-    let truncated = truncate::format_output_for_model_body(&large_error);
+    let truncated = truncate::format_output_for_model_body(
+        &large_error,
+        truncate::MODEL_FORMAT_MAX_BYTES,
+        truncate::MODEL_FORMAT_MAX_LINES,
+    );
 
     let total_lines = large_error.lines().count();
     assert_truncated_message_matches(&truncated, line, total_lines);
@@ -338,7 +345,11 @@ fn format_exec_output_truncates_large_error() {
 #[test]
 fn format_exec_output_marks_byte_truncation_without_omitted_lines() {
     let long_line = "a".repeat(truncate::MODEL_FORMAT_MAX_BYTES + 50);
-    let truncated = truncate::format_output_for_model_body(&long_line);
+    let truncated = truncate::format_output_for_model_body(
+        &long_line,
+        truncate::MODEL_FORMAT_MAX_BYTES,
+        truncate::MODEL_FORMAT_MAX_LINES,
+    );
 
     assert_ne!(truncated, long_line);
     let marker_line = format!(
@@ -359,7 +370,14 @@ fn format_exec_output_marks_byte_truncation_without_omitted_lines() {
 fn format_exec_output_returns_original_when_within_limits() {
     let content = "example output\n".repeat(10);
 
-    assert_eq!(truncate::format_output_for_model_body(&content), content);
+    assert_eq!(
+        truncate::format_output_for_model_body(
+            &content,
+            truncate::MODEL_FORMAT_MAX_BYTES,
+            truncate::MODEL_FORMAT_MAX_LINES
+        ),
+        content
+    );
 }
 
 #[test]
@@ -369,7 +387,11 @@ fn format_exec_output_reports_omitted_lines_and_keeps_head_and_tail() {
         .map(|idx| format!("line-{idx}\n"))
         .collect();
 
-    let truncated = truncate::format_output_for_model_body(&content);
+    let truncated = truncate::format_output_for_model_body(
+        &content,
+        truncate::MODEL_FORMAT_MAX_BYTES,
+        truncate::MODEL_FORMAT_MAX_LINES,
+    );
     let omitted = total_lines - truncate::MODEL_FORMAT_MAX_LINES;
     let expected_marker = format!("[... omitted {omitted} of {total_lines} lines ...]");
 
@@ -397,7 +419,11 @@ fn format_exec_output_prefers_line_marker_when_both_limits_exceeded() {
         .map(|idx| format!("line-{idx}-{long_line}\n"))
         .collect();
 
-    let truncated = truncate::format_output_for_model_body(&content);
+    let truncated = truncate::format_output_for_model_body(
+        &content,
+        truncate::MODEL_FORMAT_MAX_BYTES,
+        truncate::MODEL_FORMAT_MAX_LINES,
+    );
 
     assert!(
         truncated.contains("[... omitted 42 of 298 lines ...]"),

codex-rs/core/src/context_manager/mod.rs

@@ -3,4 +3,6 @@ mod normalize;
 mod truncate;
 
 pub(crate) use history::ContextManager;
+pub(crate) use truncate::MODEL_FORMAT_MAX_BYTES;
+pub(crate) use truncate::MODEL_FORMAT_MAX_LINES;
 pub(crate) use truncate::format_output_for_model_body;

codex-rs/core/src/context_manager/truncate.rs

@@ -2,12 +2,11 @@ use codex_protocol::models::FunctionCallOutputContentItem;
 use codex_utils_string::take_bytes_at_char_boundary;
 use codex_utils_string::take_last_bytes_at_char_boundary;
 
+use crate::util::error_or_panic;
+
 // Model-formatting limits: clients get full streams; only content sent to the model is truncated.
-pub(crate) const MODEL_FORMAT_MAX_BYTES: usize = 10 * 1024; // 10 KiB
-pub(crate) const MODEL_FORMAT_MAX_LINES: usize = 256; // lines
-pub(crate) const MODEL_FORMAT_HEAD_LINES: usize = MODEL_FORMAT_MAX_LINES / 2;
-pub(crate) const MODEL_FORMAT_TAIL_LINES: usize = MODEL_FORMAT_MAX_LINES - MODEL_FORMAT_HEAD_LINES; // 128
-pub(crate) const MODEL_FORMAT_HEAD_BYTES: usize = MODEL_FORMAT_MAX_BYTES / 2;
+pub const MODEL_FORMAT_MAX_BYTES: usize = 10 * 1024; // 10 KiB
+pub const MODEL_FORMAT_MAX_LINES: usize = 256; // lines
 
 pub(crate) fn globally_truncate_function_output_items(
     items: &[FunctionCallOutputContentItem],
@@ -56,21 +55,34 @@ pub(crate) fn globally_truncate_function_output_items(
     out
 }
 
-pub(crate) fn format_output_for_model_body(content: &str) -> String {
+pub(crate) fn format_output_for_model_body(
+    content: &str,
+    limit_bytes: usize,
+    limit_lines: usize,
+) -> String {
     // Head+tail truncation for the model: show the beginning and end with an elision.
     // Clients still receive full streams; only this formatted summary is capped.
     let total_lines = content.lines().count();
-    if content.len() <= MODEL_FORMAT_MAX_BYTES && total_lines <= MODEL_FORMAT_MAX_LINES {
+    if content.len() <= limit_bytes && total_lines <= limit_lines {
         return content.to_string();
     }
-    let output = truncate_formatted_exec_output(content, total_lines);
+    let output = truncate_formatted_exec_output(content, total_lines, limit_bytes, limit_lines);
     format!("Total output lines: {total_lines}\n\n{output}")
 }
 
-fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
+fn truncate_formatted_exec_output(
+    content: &str,
+    total_lines: usize,
+    limit_bytes: usize,
+    limit_lines: usize,
+) -> String {
+    debug_panic_on_double_truncation(content);
+    let head_lines: usize = limit_lines / 2;
+    let tail_lines: usize = limit_lines - head_lines; // 128
+    let head_bytes: usize = limit_bytes / 2;
     let segments: Vec<&str> = content.split_inclusive('\n').collect();
-    let head_take = MODEL_FORMAT_HEAD_LINES.min(segments.len());
-    let tail_take = MODEL_FORMAT_TAIL_LINES.min(segments.len().saturating_sub(head_take));
+    let head_take = head_lines.min(segments.len());
+    let tail_take = tail_lines.min(segments.len().saturating_sub(head_take));
     let omitted = segments.len().saturating_sub(head_take + tail_take);
 
     let head_slice_end: usize = segments
@@ -91,7 +103,7 @@ fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
     };
     let head_slice = &content[..head_slice_end];
     let tail_slice = &content[tail_slice_start..];
-    let truncated_by_bytes = content.len() > MODEL_FORMAT_MAX_BYTES;
+    let truncated_by_bytes = content.len() > limit_bytes;
     // this is a bit wrong. We are counting metadata lines and not just shell output lines.
     let marker = if omitted > 0 {
         Some(format!(
@@ -99,24 +111,24 @@ fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
         ))
     } else if truncated_by_bytes {
         Some(format!(
-            "\n[... output truncated to fit {MODEL_FORMAT_MAX_BYTES} bytes ...]\n\n"
+            "\n[... output truncated to fit {limit_bytes} bytes ...]\n\n"
         ))
     } else {
         None
     };
 
     let marker_len = marker.as_ref().map_or(0, String::len);
-    let base_head_budget = MODEL_FORMAT_HEAD_BYTES.min(MODEL_FORMAT_MAX_BYTES);
-    let head_budget = base_head_budget.min(MODEL_FORMAT_MAX_BYTES.saturating_sub(marker_len));
+    let base_head_budget = head_bytes.min(limit_bytes);
+    let head_budget = base_head_budget.min(limit_bytes.saturating_sub(marker_len));
     let head_part = take_bytes_at_char_boundary(head_slice, head_budget);
-    let mut result = String::with_capacity(MODEL_FORMAT_MAX_BYTES.min(content.len()));
+    let mut result = String::with_capacity(limit_bytes.min(content.len()));
 
     result.push_str(head_part);
     if let Some(marker_text) = marker.as_ref() {
         result.push_str(marker_text);
     }
 
-    let remaining = MODEL_FORMAT_MAX_BYTES.saturating_sub(result.len());
+    let remaining = limit_bytes.saturating_sub(result.len());
     if remaining == 0 {
         return result;
     }
@@ -126,3 +138,11 @@ fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
 
     result
 }
+
+fn debug_panic_on_double_truncation(content: &str) {
+    if content.contains("Total output lines:") && content.contains("omitted") {
+        error_or_panic(format!(
+            "FunctionCallOutput content was already truncated before ContextManager::record_items; this would cause double truncation {content}"
+        ));
+    }
+}

codex-rs/core/src/environment_context.rs

@@ -329,7 +329,6 @@ mod tests {
             Some(workspace_write_policy(vec!["/repo"], false)),
             Some(Shell::Bash(BashShell {
                 shell_path: "/bin/bash".into(),
-                bashrc_path: "/home/user/.bashrc".into(),
             })),
         );
         let context2 = EnvironmentContext::new(
@@ -338,7 +337,6 @@ mod tests {
             Some(workspace_write_policy(vec!["/repo"], false)),
             Some(Shell::Zsh(ZshShell {
                 shell_path: "/bin/zsh".into(),
-                zshrc_path: "/home/user/.zshrc".into(),
             })),
         );
 

codex-rs/core/src/exec.rs

@@ -532,8 +532,52 @@ async fn consume_truncated_output(
         }
     };
 
-    let stdout = stdout_handle.await??;
-    let stderr = stderr_handle.await??;
+    // Wait for the stdout/stderr collection tasks but guard against them
+    // hanging forever. In the normal case, both pipes are closed once the child
+    // terminates so the tasks exit quickly. However, if the child process
+    // spawned grandchildren that inherited its stdout/stderr file descriptors
+    // those pipes may stay open after we `kill` the direct child on timeout.
+    // That would cause the `read_capped` tasks to block on `read()`
+    // indefinitely, effectively hanging the whole agent.
+
+    const IO_DRAIN_TIMEOUT_MS: u64 = 2_000; // 2 s should be plenty for local pipes
+
+    // We need mutable bindings so we can `abort()` them on timeout.
+    use tokio::task::JoinHandle;
+
+    async fn await_with_timeout(
+        handle: &mut JoinHandle<std::io::Result<StreamOutput<Vec<u8>>>>,
+        timeout: Duration,
+    ) -> std::io::Result<StreamOutput<Vec<u8>>> {
+        match tokio::time::timeout(timeout, &mut *handle).await {
+            Ok(join_res) => match join_res {
+                Ok(io_res) => io_res,
+                Err(join_err) => Err(std::io::Error::other(join_err)),
+            },
+            Err(_elapsed) => {
+                // Timeout: abort the task to avoid hanging on open pipes.
+                handle.abort();
+                Ok(StreamOutput {
+                    text: Vec::new(),
+                    truncated_after_lines: None,
+                })
+            }
+        }
+    }
+
+    let mut stdout_handle = stdout_handle;
+    let mut stderr_handle = stderr_handle;
+
+    let stdout = await_with_timeout(
+        &mut stdout_handle,
+        Duration::from_millis(IO_DRAIN_TIMEOUT_MS),
+    )
+    .await?;
+    let stderr = await_with_timeout(
+        &mut stderr_handle,
+        Duration::from_millis(IO_DRAIN_TIMEOUT_MS),
+    )
+    .await?;
 
     drop(agg_tx);
 

codex-rs/core/src/model_family.rs

@@ -1,3 +1,6 @@
+use codex_protocol::config_types::ReasoningEffort;
+use codex_protocol::config_types::Verbosity;
+
 use crate::config::types::ReasoningSummaryFormat;
 use crate::tools::handlers::apply_patch::ApplyPatchToolType;
 use crate::tools::spec::ConfigShellToolType;
@@ -5,7 +8,9 @@ use crate::tools::spec::ConfigShellToolType;
 /// The `instructions` field in the payload sent to a model should always start
 /// with this content.
 const BASE_INSTRUCTIONS: &str = include_str!("../prompt.md");
+
 const GPT_5_CODEX_INSTRUCTIONS: &str = include_str!("../gpt_5_codex_prompt.md");
+const GPT_5_1_INSTRUCTIONS: &str = include_str!("../gpt_5_1_prompt.md");
 
 /// A model family is a group of models that share certain characteristics.
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
@@ -27,6 +32,9 @@ pub struct ModelFamily {
     // `summary` is optional).
     pub supports_reasoning_summaries: bool,
 
+    // The reasoning effort to use for this model family when none is explicitly chosen.
+    pub default_reasoning_effort: Option<ReasoningEffort>,
+
     // Define if we need a special handling of reasoning summary
     pub reasoning_summary_format: ReasoningSummaryFormat,
 
@@ -53,6 +61,9 @@ pub struct ModelFamily {
     /// If the model family supports setting the verbosity level when using Responses API.
     pub support_verbosity: bool,
 
+    // The default verbosity level for this model family when using Responses API.
+    pub default_verbosity: Option<Verbosity>,
+
     /// Preferred shell tool type for this model family when features do not override it.
     pub shell_type: ConfigShellToolType,
 }
@@ -76,7 +87,10 @@ macro_rules! model_family {
             effective_context_window_percent: 95,
             support_verbosity: false,
             shell_type: ConfigShellToolType::Default,
+            default_verbosity: None,
+            default_reasoning_effort: None,
         };
+
         // apply overrides
         $(
             mf.$key = $value;
@@ -118,8 +132,6 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
         model_family!(slug, "gpt-4o", needs_special_apply_patch_instructions: true)
     } else if slug.starts_with("gpt-3.5") {
         model_family!(slug, "gpt-3.5", needs_special_apply_patch_instructions: true)
-    } else if slug.starts_with("porcupine") {
-        model_family!(slug, "porcupine", shell_type: ConfigShellToolType::UnifiedExec)
     } else if slug.starts_with("test-gpt-5-codex") {
         model_family!(
             slug, slug,
@@ -154,7 +166,10 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
         )
 
     // Production models.
-    } else if slug.starts_with("gpt-5-codex") || slug.starts_with("codex-") {
+    } else if slug.starts_with("gpt-5-codex")
+        || slug.starts_with("gpt-5.1-codex")
+        || slug.starts_with("codex-")
+    {
         model_family!(
             slug, slug,
             supports_reasoning_summaries: true,
@@ -163,6 +178,16 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
             apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
             support_verbosity: false,
         )
+    } else if slug.starts_with("gpt-5.1") {
+        model_family!(
+            slug, "gpt-5.1",
+            supports_reasoning_summaries: true,
+            apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
+            support_verbosity: true,
+            default_verbosity: Some(Verbosity::Low),
+            base_instructions: GPT_5_1_INSTRUCTIONS.to_string(),
+            default_reasoning_effort: Some(ReasoningEffort::Medium),
+        )
     } else if slug.starts_with("gpt-5") {
         model_family!(
             slug, "gpt-5",
@@ -189,5 +214,7 @@ pub fn derive_default_model_family(model: &str) -> ModelFamily {
         effective_context_window_percent: 95,
         support_verbosity: false,
         shell_type: ConfigShellToolType::Default,
+        default_verbosity: None,
+        default_reasoning_effort: None,
     }
 }

codex-rs/core/src/openai_model_info.rs

@@ -70,7 +70,7 @@ pub(crate) fn get_model_info(model_family: &ModelFamily) -> Option<ModelInfo> {
         // https://platform.openai.com/docs/models/gpt-3.5-turbo
         "gpt-3.5-turbo" => Some(ModelInfo::new(16_385, 4_096)),
 
-        _ if slug.starts_with("gpt-5-codex") => {
+        _ if slug.starts_with("gpt-5-codex") || slug.starts_with("gpt-5.1-codex") => {
             Some(ModelInfo::new(CONTEXT_WINDOW_272K, MAX_OUTPUT_TOKENS_128K))
         }
 

codex-rs/core/src/response_processing.rs

@@ -13,92 +13,58 @@ pub(crate) async fn process_items(
     sess: &Session,
     turn_context: &TurnContext,
 ) -> (Vec<ResponseInputItem>, Vec<ResponseItem>) {
-    let mut items_to_record_in_conversation_history = Vec::<ResponseItem>::new();
+    let mut outputs_to_record = Vec::<ResponseItem>::new();
+    let mut new_inputs_to_record = Vec::<ResponseItem>::new();
     let mut responses = Vec::<ResponseInputItem>::new();
     for processed_response_item in processed_items {
         let crate::codex::ProcessedResponseItem { item, response } = processed_response_item;
-        match (&item, &response) {
-            (ResponseItem::Message { role, .. }, None) if role == "assistant" => {
-                // If the model returned a message, we need to record it.
-                items_to_record_in_conversation_history.push(item);
-            }
-            (
-                ResponseItem::LocalShellCall { .. },
-                Some(ResponseInputItem::FunctionCallOutput { call_id, output }),
-            ) => {
-                items_to_record_in_conversation_history.push(item);
-                items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
+
+        if let Some(response) = &response {
+            responses.push(response.clone());
+        }
+
+        match response {
+            Some(ResponseInputItem::FunctionCallOutput { call_id, output }) => {
+                new_inputs_to_record.push(ResponseItem::FunctionCallOutput {
                     call_id: call_id.clone(),
                     output: output.clone(),
                 });
             }
-            (
-                ResponseItem::FunctionCall { .. },
-                Some(ResponseInputItem::FunctionCallOutput { call_id, output }),
-            ) => {
-                items_to_record_in_conversation_history.push(item);
-                items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
+
+            Some(ResponseInputItem::CustomToolCallOutput { call_id, output }) => {
+                new_inputs_to_record.push(ResponseItem::CustomToolCallOutput {
                     call_id: call_id.clone(),
                     output: output.clone(),
                 });
             }
-            (
-                ResponseItem::CustomToolCall { .. },
-                Some(ResponseInputItem::CustomToolCallOutput { call_id, output }),
-            ) => {
-                items_to_record_in_conversation_history.push(item);
-                items_to_record_in_conversation_history.push(ResponseItem::CustomToolCallOutput {
-                    call_id: call_id.clone(),
-                    output: output.clone(),
-                });
-            }
-            (
-                ResponseItem::FunctionCall { .. },
-                Some(ResponseInputItem::McpToolCallOutput { call_id, result }),
-            ) => {
-                items_to_record_in_conversation_history.push(item);
+            Some(ResponseInputItem::McpToolCallOutput { call_id, result }) => {
                 let output = match result {
-                    Ok(call_tool_result) => FunctionCallOutputPayload::from(call_tool_result),
+                    Ok(call_tool_result) => FunctionCallOutputPayload::from(&call_tool_result),
                     Err(err) => FunctionCallOutputPayload {
                         content: err.clone(),
                         success: Some(false),
                         ..Default::default()
                     },
                 };
-                items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
+                new_inputs_to_record.push(ResponseItem::FunctionCallOutput {
                     call_id: call_id.clone(),
                     output,
                 });
             }
-            (
-                ResponseItem::Reasoning {
-                    id,
-                    summary,
-                    content,
-                    encrypted_content,
-                },
-                None,
-            ) => {
-                items_to_record_in_conversation_history.push(ResponseItem::Reasoning {
-                    id: id.clone(),
-                    summary: summary.clone(),
-                    content: content.clone(),
-                    encrypted_content: encrypted_content.clone(),
-                });
-            }
+            None => {}
             _ => {
                 warn!("Unexpected response item: {item:?} with response: {response:?}");
             }
         };
-        if let Some(response) = response {
-            responses.push(response);
-        }
+
+        outputs_to_record.push(item);
     }
 
+    let all_items_to_record = [outputs_to_record, new_inputs_to_record].concat();
     // Only attempt to take the lock if there is something to record.
-    if !items_to_record_in_conversation_history.is_empty() {
-        sess.record_conversation_items(turn_context, &items_to_record_in_conversation_history)
+    if !all_items_to_record.is_empty() {
+        sess.record_conversation_items(turn_context, &all_items_to_record)
             .await;
     }
-    (responses, items_to_record_in_conversation_history)
+    (responses, all_items_to_record)
 }

codex-rs/core/src/shell.rs

@@ -2,22 +2,26 @@ use serde::Deserialize;
 use serde::Serialize;
 use std::path::PathBuf;
 
+#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
+pub enum ShellType {
+    Zsh,
+    Bash,
+    PowerShell,
+}
+
 #[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
 pub struct ZshShell {
-    pub(crate) shell_path: String,
-    pub(crate) zshrc_path: String,
+    pub(crate) shell_path: PathBuf,
 }
 
 #[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
 pub struct BashShell {
-    pub(crate) shell_path: String,
-    pub(crate) bashrc_path: String,
+    pub(crate) shell_path: PathBuf,
 }
 
 #[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
 pub struct PowerShellConfig {
-    pub(crate) exe: String, // Executable name or path, e.g. "pwsh" or "powershell.exe".
-    pub(crate) bash_exe_fallback: Option<PathBuf>, // In case the model generates a bash command.
+    pub(crate) shell_path: PathBuf, // Executable name or path, e.g. "pwsh" or "powershell.exe".
 }
 
 #[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
@@ -36,7 +40,10 @@ impl Shell {
                     .file_name()
                     .map(|s| s.to_string_lossy().to_string())
             }
-            Shell::PowerShell(ps) => Some(ps.exe.clone()),
+            Shell::PowerShell(ps) => ps
+                .shell_path
+                .file_stem()
+                .map(|s| s.to_string_lossy().to_string()),
             Shell::Unknown => None,
         }
     }
@@ -47,10 +54,17 @@ impl Shell {
         match self {
             Shell::Zsh(ZshShell { shell_path, .. }) | Shell::Bash(BashShell { shell_path, .. }) => {
                 let arg = if use_login_shell { "-lc" } else { "-c" };
-                vec![shell_path.clone(), arg.to_string(), command.to_string()]
+                vec![
+                    shell_path.to_string_lossy().to_string(),
+                    arg.to_string(),
+                    command.to_string(),
+                ]
             }
             Shell::PowerShell(ps) => {
-                let mut args = vec![ps.exe.clone(), "-NoLogo".to_string()];
+                let mut args = vec![
+                    ps.shell_path.to_string_lossy().to_string(),
+                    "-NoLogo".to_string(),
+                ];
                 if !use_login_shell {
                     args.push("-NoProfile".to_string());
                 }
@@ -65,7 +79,7 @@ impl Shell {
 }
 
 #[cfg(unix)]
-fn detect_default_user_shell() -> Shell {
+fn get_user_shell_path() -> Option<PathBuf> {
     use libc::getpwuid;
     use libc::getuid;
     use std::ffi::CStr;
@@ -78,75 +92,174 @@ fn detect_default_user_shell() -> Shell {
             let shell_path = CStr::from_ptr((*pw).pw_shell)
                 .to_string_lossy()
                 .into_owned();
-            let home_path = CStr::from_ptr((*pw).pw_dir).to_string_lossy().into_owned();
+            Some(PathBuf::from(shell_path))
+        } else {
+            None
+        }
+    }
+}
 
-            if shell_path.ends_with("/zsh") {
-                return Shell::Zsh(ZshShell {
-                    shell_path,
-                    zshrc_path: format!("{home_path}/.zshrc"),
-                });
-            }
+#[cfg(not(unix))]
+fn get_user_shell_path() -> Option<PathBuf> {
+    None
+}
 
-            if shell_path.ends_with("/bash") {
-                return Shell::Bash(BashShell {
-                    shell_path,
-                    bashrc_path: format!("{home_path}/.bashrc"),
-                });
+fn file_exists(path: &PathBuf) -> Option<PathBuf> {
+    if std::fs::metadata(path).is_ok_and(|metadata| metadata.is_file()) {
+        Some(PathBuf::from(path))
+    } else {
+        None
+    }
+}
+
+fn get_shell_path(
+    shell_type: ShellType,
+    provided_path: Option<&PathBuf>,
+    binary_name: &str,
+    fallback_paths: Vec<&str>,
+) -> Option<PathBuf> {
+    // If exact provided path exists, use it
+    if provided_path.and_then(file_exists).is_some() {
+        return provided_path.cloned();
+    }
+
+    // Check if the shell we are trying to load is user's default shell
+    // if just use it
+    let default_shell_path = get_user_shell_path();
+    if let Some(default_shell_path) = default_shell_path
+        && detect_shell_type(&default_shell_path) == Some(shell_type)
+    {
+        return Some(default_shell_path);
+    }
+
+    if let Ok(path) = which::which(binary_name) {
+        return Some(path);
+    }
+
+    for path in fallback_paths {
+        //check exists
+        if let Some(path) = file_exists(&PathBuf::from(path)) {
+            return Some(path);
+        }
+    }
+
+    None
+}
+
+fn get_zsh_shell(path: Option<&PathBuf>) -> Option<ZshShell> {
+    let shell_path = get_shell_path(ShellType::Zsh, path, "zsh", vec!["/bin/zsh"]);
+
+    shell_path.map(|shell_path| ZshShell { shell_path })
+}
+
+fn get_bash_shell(path: Option<&PathBuf>) -> Option<BashShell> {
+    let shell_path = get_shell_path(ShellType::Bash, path, "bash", vec!["/bin/bash"]);
+
+    shell_path.map(|shell_path| BashShell { shell_path })
+}
+
+fn get_powershell_shell(path: Option<&PathBuf>) -> Option<PowerShellConfig> {
+    let shell_path = get_shell_path(
+        ShellType::PowerShell,
+        path,
+        "pwsh",
+        vec!["/usr/local/bin/pwsh"],
+    )
+    .or_else(|| get_shell_path(ShellType::PowerShell, path, "powershell", vec![]));
+
+    shell_path.map(|shell_path| PowerShellConfig { shell_path })
+}
+
+pub fn get_shell_by_model_provided_path(shell_path: &PathBuf) -> Shell {
+    detect_shell_type(shell_path)
+        .and_then(|shell_type| get_shell(shell_type, Some(shell_path)))
+        .unwrap_or(Shell::Unknown)
+}
+
+pub fn get_shell(shell_type: ShellType, path: Option<&PathBuf>) -> Option<Shell> {
+    match shell_type {
+        ShellType::Zsh => get_zsh_shell(path).map(Shell::Zsh),
+        ShellType::Bash => get_bash_shell(path).map(Shell::Bash),
+        ShellType::PowerShell => get_powershell_shell(path).map(Shell::PowerShell),
+    }
+}
+
+pub fn detect_shell_type(shell_path: &PathBuf) -> Option<ShellType> {
+    match shell_path.as_os_str().to_str() {
+        Some("zsh") => Some(ShellType::Zsh),
+        Some("bash") => Some(ShellType::Bash),
+        Some("pwsh") => Some(ShellType::PowerShell),
+        Some("powershell") => Some(ShellType::PowerShell),
+        _ => {
+            let shell_name = shell_path.file_stem();
+
+            if let Some(shell_name) = shell_name
+                && shell_name != shell_path
+            {
+                detect_shell_type(&PathBuf::from(shell_name))
+            } else {
+                None
             }
         }
     }
-    Shell::Unknown
 }
 
-#[cfg(unix)]
 pub async fn default_user_shell() -> Shell {
-    detect_default_user_shell()
-}
-
-#[cfg(target_os = "windows")]
-pub async fn default_user_shell() -> Shell {
-    use tokio::process::Command;
-
-    // Prefer PowerShell 7+ (`pwsh`) if available, otherwise fall back to Windows PowerShell.
-    let has_pwsh = Command::new("pwsh")
-        .arg("-NoLogo")
-        .arg("-NoProfile")
-        .arg("-Command")
-        .arg("$PSVersionTable.PSVersion.Major")
-        .output()
-        .await
-        .map(|o| o.status.success())
-        .unwrap_or(false);
-    let bash_exe = if Command::new("bash.exe")
-        .arg("--version")
-        .stdin(std::process::Stdio::null())
-        .output()
-        .await
-        .ok()
-        .map(|o| o.status.success())
-        .unwrap_or(false)
-    {
-        which::which("bash.exe").ok()
+    if cfg!(windows) {
+        get_shell(ShellType::PowerShell, None).unwrap_or(Shell::Unknown)
     } else {
-        None
-    };
-
-    if has_pwsh {
-        Shell::PowerShell(PowerShellConfig {
-            exe: "pwsh.exe".to_string(),
-            bash_exe_fallback: bash_exe,
-        })
-    } else {
-        Shell::PowerShell(PowerShellConfig {
-            exe: "powershell.exe".to_string(),
-            bash_exe_fallback: bash_exe,
-        })
+        get_user_shell_path()
+            .and_then(|shell| detect_shell_type(&shell))
+            .and_then(|shell_type| get_shell(shell_type, None))
+            .unwrap_or(Shell::Unknown)
     }
 }
 
-#[cfg(all(not(target_os = "windows"), not(unix)))]
-pub async fn default_user_shell() -> Shell {
-    Shell::Unknown
+#[cfg(test)]
+mod detect_shell_type_tests {
+    use super::*;
+
+    #[test]
+    fn test_detect_shell_type() {
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("zsh")),
+            Some(ShellType::Zsh)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("bash")),
+            Some(ShellType::Bash)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("pwsh")),
+            Some(ShellType::PowerShell)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("powershell")),
+            Some(ShellType::PowerShell)
+        );
+        assert_eq!(detect_shell_type(&PathBuf::from("fish")), None);
+        assert_eq!(detect_shell_type(&PathBuf::from("other")), None);
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("/bin/zsh")),
+            Some(ShellType::Zsh)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("/bin/bash")),
+            Some(ShellType::Bash)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("powershell.exe")),
+            Some(ShellType::PowerShell)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("pwsh.exe")),
+            Some(ShellType::PowerShell)
+        );
+        assert_eq!(
+            detect_shell_type(&PathBuf::from("/usr/local/bin/pwsh")),
+            Some(ShellType::PowerShell)
+        );
+    }
 }
 
 #[cfg(test)]
@@ -156,6 +269,34 @@ mod tests {
     use std::path::PathBuf;
     use std::process::Command;
 
+    #[test]
+    #[cfg(target_os = "macos")]
+    fn detects_zsh() {
+        let zsh_shell = get_shell(ShellType::Zsh, None).unwrap();
+
+        let ZshShell { shell_path } = match zsh_shell {
+            Shell::Zsh(zsh_shell) => zsh_shell,
+            _ => panic!("expected zsh shell"),
+        };
+
+        assert_eq!(shell_path, PathBuf::from("/bin/zsh"));
+    }
+
+    #[test]
+    fn detects_bash() {
+        let bash_shell = get_shell(ShellType::Bash, None).unwrap();
+        let BashShell { shell_path } = match bash_shell {
+            Shell::Bash(bash_shell) => bash_shell,
+            _ => panic!("expected bash shell"),
+        };
+
+        assert!(
+            shell_path == PathBuf::from("/bin/bash")
+                || shell_path == PathBuf::from("/usr/bin/bash"),
+            "shell path: {shell_path:?}",
+        );
+    }
+
     #[tokio::test]
     async fn test_current_shell_detects_zsh() {
         let shell = Command::new("sh")
@@ -164,292 +305,44 @@ mod tests {
             .output()
             .unwrap();
 
-        let home = std::env::var("HOME").unwrap();
         let shell_path = String::from_utf8_lossy(&shell.stdout).trim().to_string();
         if shell_path.ends_with("/zsh") {
             assert_eq!(
                 default_user_shell().await,
                 Shell::Zsh(ZshShell {
-                    shell_path: shell_path.to_string(),
-                    zshrc_path: format!("{home}/.zshrc",),
+                    shell_path: PathBuf::from(shell_path),
                 })
             );
         }
     }
 
     #[tokio::test]
-    async fn test_run_with_profile_bash_escaping_and_execution() {
-        let shell_path = "/bin/bash";
-
-        let cases = vec![
-            (
-                vec!["myecho"],
-                vec![shell_path, "-lc", "source BASHRC_PATH && (myecho)"],
-                Some("It works!\n"),
-            ),
-            (
-                vec!["bash", "-lc", "echo 'single' \"double\""],
-                vec![
-                    shell_path,
-                    "-lc",
-                    "source BASHRC_PATH && (echo 'single' \"double\")",
-                ],
-                Some("single double\n"),
-            ),
-        ];
-
-        for (input, expected_cmd, expected_output) in cases {
-            use std::collections::HashMap;
-
-            use crate::exec::ExecParams;
-            use crate::exec::SandboxType;
-            use crate::exec::process_exec_tool_call;
-            use crate::protocol::SandboxPolicy;
-
-            let temp_home = tempfile::tempdir().unwrap();
-            let bashrc_path = temp_home.path().join(".bashrc");
-            std::fs::write(
-                &bashrc_path,
-                r#"
-                set -x
-                function myecho {
-                    echo 'It works!'
-                }
-                "#,
-            )
-            .unwrap();
-            let command = expected_cmd
-                .iter()
-                .map(|s| s.replace("BASHRC_PATH", bashrc_path.to_str().unwrap()))
-                .collect::<Vec<_>>();
-
-            let output = process_exec_tool_call(
-                ExecParams {
-                    command: command.clone(),
-                    cwd: PathBuf::from(temp_home.path()),
-                    timeout_ms: None,
-                    env: HashMap::from([(
-                        "HOME".to_string(),
-                        temp_home.path().to_str().unwrap().to_string(),
-                    )]),
-                    with_escalated_permissions: None,
-                    justification: None,
-                    arg0: None,
-                },
-                SandboxType::None,
-                &SandboxPolicy::DangerFullAccess,
-                temp_home.path(),
-                &None,
-                None,
-            )
-            .await
-            .unwrap();
-
-            assert_eq!(output.exit_code, 0, "input: {input:?} output: {output:?}");
-            if let Some(expected) = expected_output {
-                assert_eq!(
-                    output.stdout.text, expected,
-                    "input: {input:?} output: {output:?}"
-                );
-            }
+    async fn detects_powershell_as_default() {
+        if !cfg!(windows) {
+            return;
         }
+
+        let powershell_shell = default_user_shell().await;
+        let PowerShellConfig { shell_path } = match powershell_shell {
+            Shell::PowerShell(powershell_shell) => powershell_shell,
+            _ => panic!("expected powershell shell"),
+        };
+
+        assert!(shell_path.ends_with("pwsh.exe") || shell_path.ends_with("powershell.exe"));
     }
-}
-
-#[cfg(test)]
-#[cfg(target_os = "macos")]
-mod macos_tests {
-    use std::path::PathBuf;
-
-    #[tokio::test]
-    async fn test_run_with_profile_escaping_and_execution() {
-        let shell_path = "/bin/zsh";
-
-        let cases = vec![
-            (
-                vec!["myecho"],
-                vec![shell_path, "-lc", "source ZSHRC_PATH && (myecho)"],
-                Some("It works!\n"),
-            ),
-            (
-                vec!["myecho"],
-                vec![shell_path, "-lc", "source ZSHRC_PATH && (myecho)"],
-                Some("It works!\n"),
-            ),
-            (
-                vec!["bash", "-c", "echo 'single' \"double\""],
-                vec![
-                    shell_path,
-                    "-lc",
-                    "source ZSHRC_PATH && (bash -c \"echo 'single' \\\"double\\\"\")",
-                ],
-                Some("single double\n"),
-            ),
-            (
-                vec!["bash", "-lc", "echo 'single' \"double\""],
-                vec![
-                    shell_path,
-                    "-lc",
-                    "source ZSHRC_PATH && (echo 'single' \"double\")",
-                ],
-                Some("single double\n"),
-            ),
-        ];
-        for (input, expected_cmd, expected_output) in cases {
-            use std::collections::HashMap;
-
-            use crate::exec::ExecParams;
-            use crate::exec::SandboxType;
-            use crate::exec::process_exec_tool_call;
-            use crate::protocol::SandboxPolicy;
-
-            let temp_home = tempfile::tempdir().unwrap();
-            let zshrc_path = temp_home.path().join(".zshrc");
-            std::fs::write(
-                &zshrc_path,
-                r#"
-                set -x
-                function myecho {
-                    echo 'It works!'
-                }
-                "#,
-            )
-            .unwrap();
-            let command = expected_cmd
-                .iter()
-                .map(|s| s.replace("ZSHRC_PATH", zshrc_path.to_str().unwrap()))
-                .collect::<Vec<_>>();
-
-            let output = process_exec_tool_call(
-                ExecParams {
-                    command: command.clone(),
-                    cwd: PathBuf::from(temp_home.path()),
-                    timeout_ms: None,
-                    env: HashMap::from([(
-                        "HOME".to_string(),
-                        temp_home.path().to_str().unwrap().to_string(),
-                    )]),
-                    with_escalated_permissions: None,
-                    justification: None,
-                    arg0: None,
-                },
-                SandboxType::None,
-                &SandboxPolicy::DangerFullAccess,
-                temp_home.path(),
-                &None,
-                None,
-            )
-            .await
-            .unwrap();
-
-            assert_eq!(output.exit_code, 0, "input: {input:?} output: {output:?}");
-            if let Some(expected) = expected_output {
-                assert_eq!(
-                    output.stdout.text, expected,
-                    "input: {input:?} output: {output:?}"
-                );
-            }
-        }
-    }
-}
-
-#[cfg(test)]
-#[cfg(target_os = "windows")]
-mod tests_windows {
-    use super::*;
 
     #[test]
-    fn test_format_default_shell_invocation_powershell() {
-        use std::path::PathBuf;
-
-        let cases = vec![
-            (
-                PowerShellConfig {
-                    exe: "pwsh.exe".to_string(),
-                    bash_exe_fallback: None,
-                },
-                vec!["bash", "-lc", "echo hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "powershell.exe".to_string(),
-                    bash_exe_fallback: None,
-                },
-                vec!["bash", "-lc", "echo hello"],
-                vec!["powershell.exe", "-NoProfile", "-Command", "echo hello"],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "pwsh.exe".to_string(),
-                    bash_exe_fallback: Some(PathBuf::from("bash.exe")),
-                },
-                vec!["bash", "-lc", "echo hello"],
-                vec!["bash.exe", "-lc", "echo hello"],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "pwsh.exe".to_string(),
-                    bash_exe_fallback: Some(PathBuf::from("bash.exe")),
-                },
-                vec![
-                    "bash",
-                    "-lc",
-                    "apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: destination_file.txt\n-original content\n+modified content\n*** End Patch\nEOF",
-                ],
-                vec![
-                    "bash.exe",
-                    "-lc",
-                    "apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: destination_file.txt\n-original content\n+modified content\n*** End Patch\nEOF",
-                ],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "pwsh.exe".to_string(),
-                    bash_exe_fallback: Some(PathBuf::from("bash.exe")),
-                },
-                vec!["echo", "hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "pwsh.exe".to_string(),
-                    bash_exe_fallback: Some(PathBuf::from("bash.exe")),
-                },
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
-                vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
-            ),
-            (
-                PowerShellConfig {
-                    exe: "powershell.exe".to_string(),
-                    bash_exe_fallback: Some(PathBuf::from("bash.exe")),
-                },
-                vec![
-                    "codex-mcp-server.exe",
-                    "--codex-run-as-apply-patch",
-                    "*** Begin Patch\n*** Update File: C:\\Users\\person\\destination_file.txt\n-original content\n+modified content\n*** End Patch",
-                ],
-                vec![
-                    "codex-mcp-server.exe",
-                    "--codex-run-as-apply-patch",
-                    "*** Begin Patch\n*** Update File: C:\\Users\\person\\destination_file.txt\n-original content\n+modified content\n*** End Patch",
-                ],
-            ),
-        ];
-
-        for (config, input, expected_cmd) in cases {
-            let command = expected_cmd
-                .iter()
-                .map(|s| (*s).to_string())
-                .collect::<Vec<_>>();
-
-            // These tests assert the final command for each scenario now that the helper
-            // has been removed. The inputs remain to document the original coverage.
-            let expected = expected_cmd
-                .iter()
-                .map(|s| (*s).to_string())
-                .collect::<Vec<_>>();
-            assert_eq!(command, expected, "input: {input:?} config: {config:?}");
+    fn finds_poweshell() {
+        if !cfg!(windows) {
+            return;
         }
+
+        let powershell_shell = get_shell(ShellType::PowerShell, None).unwrap();
+        let PowerShellConfig { shell_path } = match powershell_shell {
+            Shell::PowerShell(powershell_shell) => powershell_shell,
+            _ => panic!("expected powershell shell"),
+        };
+
+        assert!(shell_path.ends_with("pwsh.exe") || shell_path.ends_with("powershell.exe"));
     }
 }

codex-rs/core/src/state/session.rs

@@ -42,6 +42,10 @@ impl SessionState {
         self.history.replace(items);
     }
 
+    pub(crate) fn set_token_info(&mut self, info: Option<TokenUsageInfo>) {
+        self.history.set_token_info(info);
+    }
+
     // Token/rate limit helpers
     pub(crate) fn update_token_info_from_usage(
         &mut self,

codex-rs/core/src/tasks/user_shell.rs

@@ -20,6 +20,7 @@ use crate::parse_command::parse_command;
 use crate::protocol::EventMsg;
 use crate::protocol::ExecCommandBeginEvent;
 use crate::protocol::ExecCommandEndEvent;
+use crate::protocol::ExecCommandSource;
 use crate::protocol::SandboxPolicy;
 use crate::protocol::TaskStartedEvent;
 use crate::sandboxing::ExecEnv;
@@ -80,7 +81,8 @@ impl SessionTask for UserShellCommandTask {
                     command: shell_invocation.clone(),
                     cwd: turn_context.cwd.clone(),
                     parsed_cmd,
-                    is_user_shell_command: true,
+                    source: ExecCommandSource::UserShell,
+                    interaction_input: None,
                 }),
             )
             .await;

codex-rs/core/src/tools/context.rs

@@ -4,17 +4,13 @@ use crate::tools::TELEMETRY_PREVIEW_MAX_BYTES;
 use crate::tools::TELEMETRY_PREVIEW_MAX_LINES;
 use crate::tools::TELEMETRY_PREVIEW_TRUNCATION_NOTICE;
 use crate::turn_diff_tracker::TurnDiffTracker;
-use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::models::FunctionCallOutputContentItem;
 use codex_protocol::models::FunctionCallOutputPayload;
 use codex_protocol::models::ResponseInputItem;
 use codex_protocol::models::ShellToolCallParams;
-use codex_protocol::protocol::FileChange;
 use codex_utils_string::take_bytes_at_char_boundary;
 use mcp_types::CallToolResult;
 use std::borrow::Cow;
-use std::collections::HashMap;
-use std::path::PathBuf;
 use std::sync::Arc;
 use tokio::sync::Mutex;
 
@@ -244,25 +240,3 @@ mod tests {
         assert_eq!(lines.last(), Some(&TELEMETRY_PREVIEW_TRUNCATION_NOTICE));
     }
 }
-
-#[derive(Clone, Debug)]
-#[allow(dead_code)]
-pub(crate) struct ExecCommandContext {
-    pub(crate) turn: Arc<TurnContext>,
-    pub(crate) call_id: String,
-    pub(crate) command_for_display: Vec<String>,
-    pub(crate) cwd: PathBuf,
-    pub(crate) apply_patch: Option<ApplyPatchCommandContext>,
-    pub(crate) tool_name: String,
-    pub(crate) otel_event_manager: OtelEventManager,
-    // TODO(abhisek-oai): Find a better way to track this.
-    // https://github.com/openai/codex/pull/2471/files#r2470352242
-    pub(crate) is_user_shell_command: bool,
-}
-
-#[derive(Clone, Debug)]
-#[allow(dead_code)]
-pub(crate) struct ApplyPatchCommandContext {
-    pub(crate) user_explicitly_approved_this_action: bool,
-    pub(crate) changes: HashMap<PathBuf, FileChange>,
-}

codex-rs/core/src/tools/events.rs

@@ -8,6 +8,7 @@ use crate::parse_command::parse_command;
 use crate::protocol::EventMsg;
 use crate::protocol::ExecCommandBeginEvent;
 use crate::protocol::ExecCommandEndEvent;
+use crate::protocol::ExecCommandSource;
 use crate::protocol::FileChange;
 use crate::protocol::PatchApplyBeginEvent;
 use crate::protocol::PatchApplyEndEvent;
@@ -60,7 +61,8 @@ pub(crate) async fn emit_exec_command_begin(
     ctx: ToolEventCtx<'_>,
     command: &[String],
     cwd: &Path,
-    is_user_shell_command: bool,
+    source: ExecCommandSource,
+    interaction_input: Option<String>,
 ) {
     ctx.session
         .send_event(
@@ -70,7 +72,8 @@ pub(crate) async fn emit_exec_command_begin(
                 command: command.to_vec(),
                 cwd: cwd.to_path_buf(),
                 parsed_cmd: parse_command(command),
-                is_user_shell_command,
+                source,
+                interaction_input,
             }),
         )
         .await;
@@ -80,27 +83,26 @@ pub(crate) enum ToolEmitter {
     Shell {
         command: Vec<String>,
         cwd: PathBuf,
-        is_user_shell_command: bool,
+        source: ExecCommandSource,
     },
     ApplyPatch {
         changes: HashMap<PathBuf, FileChange>,
         auto_approved: bool,
     },
     UnifiedExec {
-        command: String,
+        command: Vec<String>,
         cwd: PathBuf,
-        // True for `exec_command` and false for `write_stdin`.
-        #[allow(dead_code)]
-        is_startup_command: bool,
+        source: ExecCommandSource,
+        interaction_input: Option<String>,
     },
 }
 
 impl ToolEmitter {
-    pub fn shell(command: Vec<String>, cwd: PathBuf, is_user_shell_command: bool) -> Self {
+    pub fn shell(command: Vec<String>, cwd: PathBuf, source: ExecCommandSource) -> Self {
         Self::Shell {
             command,
             cwd,
-            is_user_shell_command,
+            source,
         }
     }
 
@@ -111,11 +113,17 @@ impl ToolEmitter {
         }
     }
 
-    pub fn unified_exec(command: String, cwd: PathBuf, is_startup_command: bool) -> Self {
+    pub fn unified_exec(
+        command: &[String],
+        cwd: PathBuf,
+        source: ExecCommandSource,
+        interaction_input: Option<String>,
+    ) -> Self {
         Self::UnifiedExec {
-            command,
+            command: command.to_vec(),
             cwd,
-            is_startup_command,
+            source,
+            interaction_input,
         }
     }
 
@@ -125,11 +133,11 @@ impl ToolEmitter {
                 Self::Shell {
                     command,
                     cwd,
-                    is_user_shell_command,
+                    source,
                 },
                 ToolEventStage::Begin,
             ) => {
-                emit_exec_command_begin(ctx, command, cwd.as_path(), *is_user_shell_command).await;
+                emit_exec_command_begin(ctx, command, cwd.as_path(), *source, None).await;
             }
             (Self::Shell { .. }, ToolEventStage::Success(output)) => {
                 emit_exec_end(
@@ -217,8 +225,23 @@ impl ToolEmitter {
             ) => {
                 emit_patch_end(ctx, String::new(), (*message).to_string(), false).await;
             }
-            (Self::UnifiedExec { command, cwd, .. }, ToolEventStage::Begin) => {
-                emit_exec_command_begin(ctx, &[command.to_string()], cwd.as_path(), false).await;
+            (
+                Self::UnifiedExec {
+                    command,
+                    cwd,
+                    source,
+                    interaction_input,
+                },
+                ToolEventStage::Begin,
+            ) => {
+                emit_exec_command_begin(
+                    ctx,
+                    command,
+                    cwd.as_path(),
+                    *source,
+                    interaction_input.clone(),
+                )
+                .await;
             }
             (Self::UnifiedExec { .. }, ToolEventStage::Success(output)) => {
                 emit_exec_end(

codex-rs/core/src/tools/handlers/apply_patch.rs

@@ -42,6 +42,10 @@ impl ToolHandler for ApplyPatchHandler {
         )
     }
 
+    fn is_mutating(&self, _invocation: &ToolInvocation) -> bool {
+        true
+    }
+
     async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
         let ToolInvocation {
             session,

codex-rs/core/src/tools/handlers/shell.rs

@@ -10,6 +10,8 @@ use crate::codex::TurnContext;
 use crate::exec::ExecParams;
 use crate::exec_env::create_env;
 use crate::function_tool::FunctionCallError;
+use crate::is_safe_command::is_known_safe_command;
+use crate::protocol::ExecCommandSource;
 use crate::tools::context::ToolInvocation;
 use crate::tools::context::ToolOutput;
 use crate::tools::context::ToolPayload;
@@ -77,6 +79,18 @@ impl ToolHandler for ShellHandler {
         )
     }
 
+    fn is_mutating(&self, invocation: &ToolInvocation) -> bool {
+        match &invocation.payload {
+            ToolPayload::Function { arguments } => {
+                serde_json::from_str::<ShellToolCallParams>(arguments)
+                    .map(|params| !is_known_safe_command(&params.command))
+                    .unwrap_or(true)
+            }
+            ToolPayload::LocalShell { params } => !is_known_safe_command(&params.command),
+            _ => true, // unknown payloads => assume mutating
+        }
+    }
+
     async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
         let ToolInvocation {
             session,
@@ -272,11 +286,13 @@ impl ShellHandler {
         }
 
         // Regular shell execution path.
-        let emitter = ToolEmitter::shell(
-            exec_params.command.clone(),
-            exec_params.cwd.clone(),
-            is_user_shell_command,
-        );
+        let source = if is_user_shell_command {
+            ExecCommandSource::UserShell
+        } else {
+            ExecCommandSource::Agent
+        };
+        let emitter =
+            ToolEmitter::shell(exec_params.command.clone(), exec_params.cwd.clone(), source);
         let event_ctx = ToolEventCtx::new(session.as_ref(), turn.as_ref(), &call_id, None);
         emitter.begin(event_ctx).await;
 
@@ -311,8 +327,11 @@ impl ShellHandler {
 
 #[cfg(test)]
 mod tests {
+    use std::path::PathBuf;
+
     use crate::is_safe_command::is_known_safe_command;
     use crate::shell::BashShell;
+    use crate::shell::PowerShellConfig;
     use crate::shell::Shell;
     use crate::shell::ZshShell;
 
@@ -322,27 +341,19 @@ mod tests {
     #[test]
     fn commands_generated_by_shell_command_handler_can_be_matched_by_is_known_safe_command() {
         let bash_shell = Shell::Bash(BashShell {
-            shell_path: "/bin/bash".to_string(),
-            bashrc_path: "/home/user/.bashrc".to_string(),
+            shell_path: PathBuf::from("/bin/bash"),
         });
         assert_safe(&bash_shell, "ls -la");
 
         let zsh_shell = Shell::Zsh(ZshShell {
-            shell_path: "/bin/zsh".to_string(),
-            zshrc_path: "/home/user/.zshrc".to_string(),
+            shell_path: PathBuf::from("/bin/zsh"),
         });
         assert_safe(&zsh_shell, "ls -la");
 
-        #[cfg(target_os = "windows")]
-        {
-            use crate::shell::PowerShellConfig;
-
-            let powershell = Shell::PowerShell(PowerShellConfig {
-                exe: "pwsh.exe".to_string(),
-                bash_exe_fallback: None,
-            });
-            assert_safe(&powershell, "ls -Name");
-        }
+        let powershell = Shell::PowerShell(PowerShellConfig {
+            shell_path: PathBuf::from("pwsh.exe"),
+        });
+        assert_safe(&powershell, "ls -Name");
     }
 
     fn assert_safe(shell: &Shell, command: &str) {

codex-rs/core/src/tools/handlers/unified_exec.rs

@@ -1,12 +1,12 @@
 use std::path::PathBuf;
 
-use async_trait::async_trait;
-use serde::Deserialize;
-
 use crate::function_tool::FunctionCallError;
+use crate::is_safe_command::is_known_safe_command;
 use crate::protocol::EventMsg;
 use crate::protocol::ExecCommandOutputDeltaEvent;
+use crate::protocol::ExecCommandSource;
 use crate::protocol::ExecOutputStream;
+use crate::shell::get_shell_by_model_provided_path;
 use crate::tools::context::ToolInvocation;
 use crate::tools::context::ToolOutput;
 use crate::tools::context::ToolPayload;
@@ -20,6 +20,8 @@ use crate::unified_exec::UnifiedExecContext;
 use crate::unified_exec::UnifiedExecResponse;
 use crate::unified_exec::UnifiedExecSessionManager;
 use crate::unified_exec::WriteStdinRequest;
+use async_trait::async_trait;
+use serde::Deserialize;
 
 pub struct UnifiedExecHandler;
 
@@ -32,8 +34,8 @@ struct ExecCommandArgs {
     shell: String,
     #[serde(default = "default_login")]
     login: bool,
-    #[serde(default)]
-    yield_time_ms: Option<u64>,
+    #[serde(default = "default_exec_yield_time_ms")]
+    yield_time_ms: u64,
     #[serde(default)]
     max_output_tokens: Option<usize>,
     #[serde(default)]
@@ -47,12 +49,20 @@ struct WriteStdinArgs {
     session_id: i32,
     #[serde(default)]
     chars: String,
-    #[serde(default)]
-    yield_time_ms: Option<u64>,
+    #[serde(default = "default_write_stdin_yield_time_ms")]
+    yield_time_ms: u64,
     #[serde(default)]
     max_output_tokens: Option<usize>,
 }
 
+fn default_exec_yield_time_ms() -> u64 {
+    10000
+}
+
+fn default_write_stdin_yield_time_ms() -> u64 {
+    250
+}
+
 fn default_shell() -> String {
     "/bin/bash".to_string()
 }
@@ -74,6 +84,20 @@ impl ToolHandler for UnifiedExecHandler {
         )
     }
 
+    fn is_mutating(&self, invocation: &ToolInvocation) -> bool {
+        let (ToolPayload::Function { arguments } | ToolPayload::UnifiedExec { arguments }) =
+            &invocation.payload
+        else {
+            return true;
+        };
+
+        let Ok(params) = serde_json::from_str::<ExecCommandArgs>(arguments) else {
+            return true;
+        };
+        let command = get_command(&params);
+        !is_known_safe_command(&command)
+    }
+
     async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
         let ToolInvocation {
             session,
@@ -104,15 +128,15 @@ impl ToolHandler for UnifiedExecHandler {
                         "failed to parse exec_command arguments: {err:?}"
                     ))
                 })?;
+
+                let command = get_command(&args);
                 let ExecCommandArgs {
-                    cmd,
                     workdir,
-                    shell,
-                    login,
                     yield_time_ms,
                     max_output_tokens,
                     with_escalated_permissions,
                     justification,
+                    ..
                 } = args;
 
                 if with_escalated_permissions.unwrap_or(false)
@@ -139,15 +163,18 @@ impl ToolHandler for UnifiedExecHandler {
                     &context.call_id,
                     None,
                 );
-                let emitter = ToolEmitter::unified_exec(cmd.clone(), cwd.clone(), true);
+                let emitter = ToolEmitter::unified_exec(
+                    &command,
+                    cwd.clone(),
+                    ExecCommandSource::UnifiedExecStartup,
+                    None,
+                );
                 emitter.emit(event_ctx, ToolEventStage::Begin).await;
 
                 manager
                     .exec_command(
                         ExecCommandRequest {
-                            command: &cmd,
-                            shell: &shell,
-                            login,
+                            command,
                             yield_time_ms,
                             max_output_tokens,
                             workdir,
@@ -169,6 +196,7 @@ impl ToolHandler for UnifiedExecHandler {
                 })?;
                 manager
                     .write_stdin(WriteStdinRequest {
+                        call_id: &call_id,
                         session_id: args.session_id,
                         input: &args.chars,
                         yield_time_ms: args.yield_time_ms,
@@ -208,6 +236,11 @@ impl ToolHandler for UnifiedExecHandler {
     }
 }
 
+fn get_command(args: &ExecCommandArgs) -> Vec<String> {
+    let shell = get_shell_by_model_provided_path(&PathBuf::from(args.shell.clone()));
+    shell.derive_exec_args(&args.cmd, args.login)
+}
+
 fn format_response(response: &UnifiedExecResponse) -> String {
     let mut sections = Vec::new();
 

codex-rs/core/src/tools/mod.rs

@@ -9,6 +9,8 @@ pub mod runtimes;
 pub mod sandboxing;
 pub mod spec;
 
+use crate::context_manager::MODEL_FORMAT_MAX_BYTES;
+use crate::context_manager::MODEL_FORMAT_MAX_LINES;
 use crate::context_manager::format_output_for_model_body;
 use crate::exec::ExecToolCallOutput;
 pub use router::ToolRouter;
@@ -75,5 +77,5 @@ pub fn format_exec_output_str(exec_output: &ExecToolCallOutput) -> String {
     };
 
     // Truncate for model consumption before serialization.
-    format_output_for_model_body(&body)
+    format_output_for_model_body(&body, MODEL_FORMAT_MAX_BYTES, MODEL_FORMAT_MAX_LINES)
 }

codex-rs/core/src/tools/parallel.rs

@@ -16,7 +16,6 @@ use crate::tools::router::ToolCall;
 use crate::tools::router::ToolRouter;
 use codex_protocol::models::FunctionCallOutputPayload;
 use codex_protocol::models::ResponseInputItem;
-use codex_utils_readiness::Readiness;
 
 pub(crate) struct ToolCallRuntime {
     router: Arc<ToolRouter>,
@@ -55,7 +54,6 @@ impl ToolCallRuntime {
         let tracker = Arc::clone(&self.tracker);
         let lock = Arc::clone(&self.parallel_execution);
         let started = Instant::now();
-        let readiness = self.turn_context.tool_call_gate.clone();
 
         let handle: AbortOnDropHandle<Result<ResponseInputItem, FunctionCallError>> =
             AbortOnDropHandle::new(tokio::spawn(async move {
@@ -65,9 +63,6 @@ impl ToolCallRuntime {
                         Ok(Self::aborted_response(&call, secs))
                     },
                     res = async {
-                        tracing::trace!("waiting for tool gate");
-                        readiness.wait_ready().await;
-                        tracing::trace!("tool gate released");
                         let _guard = if supports_parallel {
                             Either::Left(lock.read().await)
                         } else {

codex-rs/core/src/tools/registry.rs

@@ -2,15 +2,15 @@ use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::Duration;
 
-use async_trait::async_trait;
-use codex_protocol::models::ResponseInputItem;
-use tracing::warn;
-
 use crate::client_common::tools::ToolSpec;
 use crate::function_tool::FunctionCallError;
 use crate::tools::context::ToolInvocation;
 use crate::tools::context::ToolOutput;
 use crate::tools::context::ToolPayload;
+use async_trait::async_trait;
+use codex_protocol::models::ResponseInputItem;
+use codex_utils_readiness::Readiness;
+use tracing::warn;
 
 #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
 pub enum ToolKind {
@@ -30,6 +30,10 @@ pub trait ToolHandler: Send + Sync {
         )
     }
 
+    fn is_mutating(&self, _invocation: &ToolInvocation) -> bool {
+        false
+    }
+
     async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError>;
 }
 
@@ -106,6 +110,11 @@ impl ToolRegistry {
                     let output_cell = &output_cell;
                     let invocation = invocation;
                     async move {
+                        if handler.is_mutating(&invocation) {
+                            tracing::trace!("waiting for tool gate");
+                            invocation.turn.tool_call_gate.wait_ready().await;
+                            tracing::trace!("tool gate released");
+                        }
                         match handler.handle(invocation).await {
                             Ok(output) => {
                                 let preview = output.log_preview();

codex-rs/core/src/tools/spec.rs

@@ -1257,6 +1257,23 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_build_specs_gpt51_codex_default() {
+        assert_model_tools(
+            "gpt-5.1-codex",
+            &Features::with_defaults(),
+            &[
+                "shell",
+                "list_mcp_resources",
+                "list_mcp_resource_templates",
+                "read_mcp_resource",
+                "update_plan",
+                "apply_patch",
+                "view_image",
+            ],
+        );
+    }
+
     #[test]
     fn test_build_specs_gpt5_codex_unified_exec_web_search() {
         assert_model_tools(
@@ -1278,6 +1295,27 @@ mod tests {
         );
     }
 
+    #[test]
+    fn test_build_specs_gpt51_codex_unified_exec_web_search() {
+        assert_model_tools(
+            "gpt-5.1-codex",
+            Features::with_defaults()
+                .enable(Feature::UnifiedExec)
+                .enable(Feature::WebSearchRequest),
+            &[
+                "exec_command",
+                "write_stdin",
+                "list_mcp_resources",
+                "list_mcp_resource_templates",
+                "read_mcp_resource",
+                "update_plan",
+                "apply_patch",
+                "web_search",
+                "view_image",
+            ],
+        );
+    }
+
     #[test]
     fn test_codex_mini_defaults() {
         assert_model_tools(
@@ -1295,17 +1333,34 @@ mod tests {
     }
 
     #[test]
-    fn test_porcupine_defaults() {
+    fn test_codex_5_1_mini_defaults() {
         assert_model_tools(
-            "porcupine",
+            "gpt-5.1-codex-mini",
             &Features::with_defaults(),
             &[
-                "exec_command",
-                "write_stdin",
+                "shell",
                 "list_mcp_resources",
                 "list_mcp_resource_templates",
                 "read_mcp_resource",
                 "update_plan",
+                "apply_patch",
+                "view_image",
+            ],
+        );
+    }
+
+    #[test]
+    fn test_gpt_5_1_defaults() {
+        assert_model_tools(
+            "gpt-5.1",
+            &Features::with_defaults(),
+            &[
+                "shell",
+                "list_mcp_resources",
+                "list_mcp_resource_templates",
+                "read_mcp_resource",
+                "update_plan",
+                "apply_patch",
                 "view_image",
             ],
         );

codex-rs/core/src/unified_exec/mod.rs

@@ -41,7 +41,6 @@ mod session_manager;
 pub(crate) use errors::UnifiedExecError;
 pub(crate) use session::UnifiedExecSession;
 
-pub(crate) const DEFAULT_YIELD_TIME_MS: u64 = 10_000;
 pub(crate) const MIN_YIELD_TIME_MS: u64 = 250;
 pub(crate) const MAX_YIELD_TIME_MS: u64 = 30_000;
 pub(crate) const DEFAULT_MAX_OUTPUT_TOKENS: usize = 10_000;
@@ -64,11 +63,9 @@ impl UnifiedExecContext {
 }
 
 #[derive(Debug)]
-pub(crate) struct ExecCommandRequest<'a> {
-    pub command: &'a str,
-    pub shell: &'a str,
-    pub login: bool,
-    pub yield_time_ms: Option<u64>,
+pub(crate) struct ExecCommandRequest {
+    pub command: Vec<String>,
+    pub yield_time_ms: u64,
     pub max_output_tokens: Option<usize>,
     pub workdir: Option<PathBuf>,
     pub with_escalated_permissions: Option<bool>,
@@ -77,9 +74,10 @@ pub(crate) struct ExecCommandRequest<'a> {
 
 #[derive(Debug)]
 pub(crate) struct WriteStdinRequest<'a> {
+    pub call_id: &'a str,
     pub session_id: i32,
     pub input: &'a str,
-    pub yield_time_ms: Option<u64>,
+    pub yield_time_ms: u64,
     pub max_output_tokens: Option<usize>,
 }
 
@@ -92,6 +90,7 @@ pub(crate) struct UnifiedExecResponse {
     pub session_id: Option<i32>,
     pub exit_code: Option<i32>,
     pub original_token_count: Option<usize>,
+    pub session_command: Option<Vec<String>>,
 }
 
 #[derive(Default)]
@@ -105,16 +104,13 @@ struct SessionEntry {
     session_ref: Arc<Session>,
     turn_ref: Arc<TurnContext>,
     call_id: String,
-    command: String,
+    command: Vec<String>,
     cwd: PathBuf,
     started_at: tokio::time::Instant,
 }
 
-pub(crate) fn clamp_yield_time(yield_time_ms: Option<u64>) -> u64 {
-    match yield_time_ms {
-        Some(value) => value.clamp(MIN_YIELD_TIME_MS, MAX_YIELD_TIME_MS),
-        None => DEFAULT_YIELD_TIME_MS,
-    }
+pub(crate) fn clamp_yield_time(yield_time_ms: u64) -> u64 {
+    yield_time_ms.clamp(MIN_YIELD_TIME_MS, MAX_YIELD_TIME_MS)
 }
 
 pub(crate) fn resolve_max_tokens(max_tokens: Option<usize>) -> usize {
@@ -187,7 +183,7 @@ mod tests {
         session: &Arc<Session>,
         turn: &Arc<TurnContext>,
         cmd: &str,
-        yield_time_ms: Option<u64>,
+        yield_time_ms: u64,
     ) -> Result<UnifiedExecResponse, UnifiedExecError> {
         let context =
             UnifiedExecContext::new(Arc::clone(session), Arc::clone(turn), "call".to_string());
@@ -197,9 +193,7 @@ mod tests {
             .unified_exec_manager
             .exec_command(
                 ExecCommandRequest {
-                    command: cmd,
-                    shell: "/bin/bash",
-                    login: true,
+                    command: vec!["bash".to_string(), "-lc".to_string(), cmd.to_string()],
                     yield_time_ms,
                     max_output_tokens: None,
                     workdir: None,
@@ -215,12 +209,13 @@ mod tests {
         session: &Arc<Session>,
         session_id: i32,
         input: &str,
-        yield_time_ms: Option<u64>,
+        yield_time_ms: u64,
     ) -> Result<UnifiedExecResponse, UnifiedExecError> {
         session
             .services
             .unified_exec_manager
             .write_stdin(WriteStdinRequest {
+                call_id: "write-stdin",
                 session_id,
                 input,
                 yield_time_ms,
@@ -253,14 +248,14 @@ mod tests {
 
         let (session, turn) = test_session_and_turn();
 
-        let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
+        let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
         let session_id = open_shell.session_id.expect("expected session_id");
 
         write_stdin(
             &session,
             session_id,
             "export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
-            Some(2_500),
+            2_500,
         )
         .await?;
 
@@ -268,7 +263,7 @@ mod tests {
             &session,
             session_id,
             "echo $CODEX_INTERACTIVE_SHELL_VAR\n",
-            Some(2_500),
+            2_500,
         )
         .await?;
         assert!(
@@ -285,24 +280,19 @@ mod tests {
 
         let (session, turn) = test_session_and_turn();
 
-        let shell_a = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
+        let shell_a = exec_command(&session, &turn, "bash -i", 2_500).await?;
         let session_a = shell_a.session_id.expect("expected session id");
 
         write_stdin(
             &session,
             session_a,
             "export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
-            Some(2_500),
+            2_500,
         )
         .await?;
 
-        let out_2 = exec_command(
-            &session,
-            &turn,
-            "echo $CODEX_INTERACTIVE_SHELL_VAR",
-            Some(2_500),
-        )
-        .await?;
+        let out_2 =
+            exec_command(&session, &turn, "echo $CODEX_INTERACTIVE_SHELL_VAR", 2_500).await?;
         assert!(
             out_2.session_id.is_none(),
             "short command should not retain a session"
@@ -316,7 +306,7 @@ mod tests {
             &session,
             session_a,
             "echo $CODEX_INTERACTIVE_SHELL_VAR\n",
-            Some(2_500),
+            2_500,
         )
         .await?;
         assert!(
@@ -333,14 +323,14 @@ mod tests {
 
         let (session, turn) = test_session_and_turn();
 
-        let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
+        let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
         let session_id = open_shell.session_id.expect("expected session id");
 
         write_stdin(
             &session,
             session_id,
             "export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
-            Some(2_500),
+            2_500,
         )
         .await?;
 
@@ -348,7 +338,7 @@ mod tests {
             &session,
             session_id,
             "sleep 5 && echo $CODEX_INTERACTIVE_SHELL_VAR\n",
-            Some(10),
+            10,
         )
         .await?;
         assert!(
@@ -358,7 +348,7 @@ mod tests {
 
         tokio::time::sleep(Duration::from_secs(7)).await;
 
-        let out_3 = write_stdin(&session, session_id, "", Some(100)).await?;
+        let out_3 = write_stdin(&session, session_id, "", 100).await?;
 
         assert!(
             out_3.output.contains("codex"),
@@ -373,7 +363,7 @@ mod tests {
     async fn requests_with_large_timeout_are_capped() -> anyhow::Result<()> {
         let (session, turn) = test_session_and_turn();
 
-        let result = exec_command(&session, &turn, "echo codex", Some(120_000)).await?;
+        let result = exec_command(&session, &turn, "echo codex", 120_000).await?;
 
         assert!(result.session_id.is_none());
         assert!(result.output.contains("codex"));
@@ -385,7 +375,7 @@ mod tests {
     #[ignore] // Ignored while we have a better way to test this.
     async fn completed_commands_do_not_persist_sessions() -> anyhow::Result<()> {
         let (session, turn) = test_session_and_turn();
-        let result = exec_command(&session, &turn, "echo codex", Some(2_500)).await?;
+        let result = exec_command(&session, &turn, "echo codex", 2_500).await?;
 
         assert!(
             result.session_id.is_none(),
@@ -412,14 +402,14 @@ mod tests {
 
         let (session, turn) = test_session_and_turn();
 
-        let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
+        let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
         let session_id = open_shell.session_id.expect("expected session id");
 
-        write_stdin(&session, session_id, "exit\n", Some(2_500)).await?;
+        write_stdin(&session, session_id, "exit\n", 2_500).await?;
 
         tokio::time::sleep(Duration::from_millis(200)).await;
 
-        let err = write_stdin(&session, session_id, "", Some(100))
+        let err = write_stdin(&session, session_id, "", 100)
             .await
             .expect_err("expected unknown session error");
 

codex-rs/core/src/unified_exec/session_manager.rs

@@ -6,12 +6,18 @@ use tokio::sync::mpsc;
 use tokio::time::Duration;
 use tokio::time::Instant;
 
+use crate::codex::Session;
+use crate::codex::TurnContext;
 use crate::exec::ExecToolCallOutput;
 use crate::exec::StreamOutput;
 use crate::exec_env::create_env;
+use crate::protocol::BackgroundEventEvent;
+use crate::protocol::EventMsg;
+use crate::protocol::ExecCommandSource;
 use crate::sandboxing::ExecEnv;
 use crate::tools::events::ToolEmitter;
 use crate::tools::events::ToolEventCtx;
+use crate::tools::events::ToolEventFailure;
 use crate::tools::events::ToolEventStage;
 use crate::tools::orchestrator::ToolOrchestrator;
 use crate::tools::runtimes::unified_exec::UnifiedExecRequest as UnifiedExecToolRequest;
@@ -19,7 +25,6 @@ use crate::tools::runtimes::unified_exec::UnifiedExecRuntime;
 use crate::tools::sandboxing::ToolCtx;
 
 use super::ExecCommandRequest;
-use super::MIN_YIELD_TIME_MS;
 use super::SessionEntry;
 use super::UnifiedExecContext;
 use super::UnifiedExecError;
@@ -36,23 +41,17 @@ use super::truncate_output_to_tokens;
 impl UnifiedExecSessionManager {
     pub(crate) async fn exec_command(
         &self,
-        request: ExecCommandRequest<'_>,
+        request: ExecCommandRequest,
         context: &UnifiedExecContext,
     ) -> Result<UnifiedExecResponse, UnifiedExecError> {
         let cwd = request
             .workdir
             .clone()
             .unwrap_or_else(|| context.turn.cwd.clone());
-        let shell_flag = if request.login { "-lc" } else { "-c" };
-        let command = vec![
-            request.shell.to_string(),
-            shell_flag.to_string(),
-            request.command.to_string(),
-        ];
 
         let session = self
             .open_session_with_sandbox(
-                command,
+                &request.command,
                 cwd.clone(),
                 request.with_escalated_permissions,
                 request.justification,
@@ -61,8 +60,7 @@ impl UnifiedExecSessionManager {
             .await?;
 
         let max_tokens = resolve_max_tokens(request.max_output_tokens);
-        let yield_time_ms =
-            clamp_yield_time(Some(request.yield_time_ms.unwrap_or(MIN_YIELD_TIME_MS)));
+        let yield_time_ms = clamp_yield_time(request.yield_time_ms);
 
         let start = Instant::now();
         let (output_buffer, output_notify) = session.output_handles();
@@ -74,15 +72,18 @@ impl UnifiedExecSessionManager {
         let text = String::from_utf8_lossy(&collected).to_string();
         let (output, original_token_count) = truncate_output_to_tokens(&text, max_tokens);
         let chunk_id = generate_chunk_id();
-        let exit_code = session.exit_code();
-        let session_id = if session.has_exited() {
-            None
-        } else {
-            Some(
-                self.store_session(session, context, request.command, cwd.clone(), start)
-                    .await,
-            )
-        };
+        let has_exited = session.has_exited();
+        let stored_id = self
+            .store_session(session, context, &request.command, cwd.clone(), start)
+            .await;
+        let exit_code = self
+            .sessions
+            .lock()
+            .await
+            .get(&stored_id)
+            .map(|entry| entry.session.exit_code());
+        // Only include a session_id in the response if the process is still alive.
+        let session_id = if has_exited { None } else { Some(stored_id) };
 
         let response = UnifiedExecResponse {
             event_call_id: context.call_id.clone(),
@@ -90,16 +91,21 @@ impl UnifiedExecSessionManager {
             wall_time,
             output,
             session_id,
-            exit_code,
+            exit_code: exit_code.flatten(),
             original_token_count,
+            session_command: Some(request.command.clone()),
         };
 
+        if response.session_id.is_some() {
+            Self::emit_waiting_status(&context.session, &context.turn, &request.command).await;
+        }
+
         // If the command completed during this call, emit an ExecCommandEnd via the emitter.
         if response.session_id.is_none() {
             let exit = response.exit_code.unwrap_or(-1);
             Self::emit_exec_end_from_context(
                 context,
-                request.command.to_string(),
+                &request.command,
                 cwd,
                 response.output.clone(),
                 exit,
@@ -117,11 +123,46 @@ impl UnifiedExecSessionManager {
     ) -> Result<UnifiedExecResponse, UnifiedExecError> {
         let session_id = request.session_id;
 
-        let (writer_tx, output_buffer, output_notify) =
-            self.prepare_session_handles(session_id).await?;
+        let (
+            writer_tx,
+            output_buffer,
+            output_notify,
+            session_ref,
+            turn_ref,
+            session_command,
+            session_cwd,
+        ) = self.prepare_session_handles(session_id).await?;
+
+        let interaction_emitter = ToolEmitter::unified_exec(
+            &session_command,
+            session_cwd.clone(),
+            ExecCommandSource::UnifiedExecInteraction,
+            (!request.input.is_empty()).then(|| request.input.to_string()),
+        );
+        let make_event_ctx = || {
+            ToolEventCtx::new(
+                session_ref.as_ref(),
+                turn_ref.as_ref(),
+                request.call_id,
+                None,
+            )
+        };
+        interaction_emitter
+            .emit(make_event_ctx(), ToolEventStage::Begin)
+            .await;
 
         if !request.input.is_empty() {
-            Self::send_input(&writer_tx, request.input.as_bytes()).await?;
+            if let Err(err) = Self::send_input(&writer_tx, request.input.as_bytes()).await {
+                interaction_emitter
+                    .emit(
+                        make_event_ctx(),
+                        ToolEventStage::Failure(ToolEventFailure::Message(format!(
+                            "write_stdin failed: {err:?}"
+                        ))),
+                    )
+                    .await;
+                return Err(err);
+            }
             tokio::time::sleep(Duration::from_millis(100)).await;
         }
 
@@ -159,8 +200,28 @@ impl UnifiedExecSessionManager {
             session_id,
             exit_code,
             original_token_count,
+            session_command: Some(session_command.clone()),
         };
 
+        let interaction_output = ExecToolCallOutput {
+            exit_code: response.exit_code.unwrap_or(0),
+            stdout: StreamOutput::new(response.output.clone()),
+            stderr: StreamOutput::new(String::new()),
+            aggregated_output: StreamOutput::new(response.output.clone()),
+            duration: response.wall_time,
+            timed_out: false,
+        };
+        interaction_emitter
+            .emit(
+                make_event_ctx(),
+                ToolEventStage::Success(interaction_output),
+            )
+            .await;
+
+        if response.session_id.is_some() {
+            Self::emit_waiting_status(&session_ref, &turn_ref, &session_command).await;
+        }
+
         if let (Some(exit), Some(entry)) = (response.exit_code, completion_entry) {
             let total_duration = Instant::now().saturating_duration_since(entry.started_at);
             Self::emit_exec_end_from_entry(entry, response.output.clone(), exit, total_duration)
@@ -197,17 +258,44 @@ impl UnifiedExecSessionManager {
     async fn prepare_session_handles(
         &self,
         session_id: i32,
-    ) -> Result<(mpsc::Sender<Vec<u8>>, OutputBuffer, Arc<Notify>), UnifiedExecError> {
+    ) -> Result<
+        (
+            mpsc::Sender<Vec<u8>>,
+            OutputBuffer,
+            Arc<Notify>,
+            Arc<Session>,
+            Arc<TurnContext>,
+            Vec<String>,
+            PathBuf,
+        ),
+        UnifiedExecError,
+    > {
         let sessions = self.sessions.lock().await;
-        let (output_buffer, output_notify, writer_tx) =
+        let (output_buffer, output_notify, writer_tx, session, turn, command, cwd) =
             if let Some(entry) = sessions.get(&session_id) {
                 let (buffer, notify) = entry.session.output_handles();
-                (buffer, notify, entry.session.writer_sender())
+                (
+                    buffer,
+                    notify,
+                    entry.session.writer_sender(),
+                    Arc::clone(&entry.session_ref),
+                    Arc::clone(&entry.turn_ref),
+                    entry.command.clone(),
+                    entry.cwd.clone(),
+                )
             } else {
                 return Err(UnifiedExecError::UnknownSessionId { session_id });
             };
 
-        Ok((writer_tx, output_buffer, output_notify))
+        Ok((
+            writer_tx,
+            output_buffer,
+            output_notify,
+            session,
+            turn,
+            command,
+            cwd,
+        ))
     }
 
     async fn send_input(
@@ -224,7 +312,7 @@ impl UnifiedExecSessionManager {
         &self,
         session: UnifiedExecSession,
         context: &UnifiedExecContext,
-        command: &str,
+        command: &[String],
         cwd: PathBuf,
         started_at: Instant,
     ) -> i32 {
@@ -236,7 +324,7 @@ impl UnifiedExecSessionManager {
             session_ref: Arc::clone(&context.session),
             turn_ref: Arc::clone(&context.turn),
             call_id: context.call_id.clone(),
-            command: command.to_string(),
+            command: command.to_vec(),
             cwd,
             started_at,
         };
@@ -264,7 +352,12 @@ impl UnifiedExecSessionManager {
             &entry.call_id,
             None,
         );
-        let emitter = ToolEmitter::unified_exec(entry.command, entry.cwd, true);
+        let emitter = ToolEmitter::unified_exec(
+            &entry.command,
+            entry.cwd,
+            ExecCommandSource::UnifiedExecStartup,
+            None,
+        );
         emitter
             .emit(event_ctx, ToolEventStage::Success(output))
             .await;
@@ -272,7 +365,7 @@ impl UnifiedExecSessionManager {
 
     async fn emit_exec_end_from_context(
         context: &UnifiedExecContext,
-        command: String,
+        command: &[String],
         cwd: PathBuf,
         aggregated_output: String,
         exit_code: i32,
@@ -292,12 +385,28 @@ impl UnifiedExecSessionManager {
             &context.call_id,
             None,
         );
-        let emitter = ToolEmitter::unified_exec(command, cwd, true);
+        let emitter =
+            ToolEmitter::unified_exec(command, cwd, ExecCommandSource::UnifiedExecStartup, None);
         emitter
             .emit(event_ctx, ToolEventStage::Success(output))
             .await;
     }
 
+    async fn emit_waiting_status(
+        session: &Arc<Session>,
+        turn: &Arc<TurnContext>,
+        command: &[String],
+    ) {
+        let command_display = command.join(" ");
+        let message = format!("Waiting for `{command_display}`");
+        session
+            .send_event(
+                turn.as_ref(),
+                EventMsg::BackgroundEvent(BackgroundEventEvent { message }),
+            )
+            .await;
+    }
+
     pub(crate) async fn open_session_with_exec_env(
         &self,
         env: &ExecEnv,
@@ -321,7 +430,7 @@ impl UnifiedExecSessionManager {
 
     pub(super) async fn open_session_with_sandbox(
         &self,
-        command: Vec<String>,
+        command: &[String],
         cwd: PathBuf,
         with_escalated_permissions: Option<bool>,
         justification: Option<String>,
@@ -330,7 +439,7 @@ impl UnifiedExecSessionManager {
         let mut orchestrator = ToolOrchestrator::new();
         let mut runtime = UnifiedExecRuntime::new(self);
         let req = UnifiedExecToolRequest::new(
-            command,
+            command.to_vec(),
             cwd,
             create_env(&context.turn.shell_environment_policy),
             with_escalated_permissions,

codex-rs/core/src/util.rs

@@ -16,7 +16,12 @@ pub(crate) fn backoff(attempt: u64) -> Duration {
 
 pub(crate) fn error_or_panic(message: String) {
     if cfg!(debug_assertions) || env!("CARGO_PKG_VERSION").contains("alpha") {
-        panic!("{message}");
+        error!("{message}");
+        panic!(
+            "This is an intentional panic to catch errors in debug and alpha builds. 
+            If you don't know why this panic is happening, please report the issue to the Codex team in the appropriate channels including `/feedback`.
+             {message}"
+        );
     } else {
         error!("{message}");
     }

codex-rs/core/tests/chat_completions_payload.rs

@@ -10,11 +10,11 @@ use codex_core::ModelProviderInfo;
 use codex_core::Prompt;
 use codex_core::ResponseItem;
 use codex_core::WireApi;
-use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
 use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::ConversationId;
 use codex_protocol::models::ReasoningItemContent;
 use core_test_support::load_default_config_for_test;
+use core_test_support::skip_if_no_network;
 use futures::StreamExt;
 use serde_json::Value;
 use tempfile::TempDir;
@@ -24,10 +24,6 @@ use wiremock::ResponseTemplate;
 use wiremock::matchers::method;
 use wiremock::matchers::path;
 
-fn network_disabled() -> bool {
-    std::env::var(CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok()
-}
-
 async fn run_request(input: Vec<ResponseItem>) -> Value {
     let server = MockServer::start().await;
 
@@ -191,12 +187,7 @@ fn first_assistant(messages: &[Value]) -> &Value {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn omits_reasoning_when_none_present() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![user_message("u1"), assistant_message("a1")]).await;
     let messages = messages_from(&body);
@@ -208,12 +199,7 @@ async fn omits_reasoning_when_none_present() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn attaches_reasoning_to_previous_assistant() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         user_message("u1"),
@@ -230,12 +216,7 @@ async fn attaches_reasoning_to_previous_assistant() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn attaches_reasoning_to_function_call_anchor() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         user_message("u1"),
@@ -257,12 +238,7 @@ async fn attaches_reasoning_to_function_call_anchor() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn attaches_reasoning_to_local_shell_call() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         user_message("u1"),
@@ -282,12 +258,7 @@ async fn attaches_reasoning_to_local_shell_call() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn drops_reasoning_when_last_role_is_user() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         assistant_message("aPrev"),
@@ -301,12 +272,7 @@ async fn drops_reasoning_when_last_role_is_user() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn ignores_reasoning_before_last_user() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         user_message("u1"),
@@ -321,12 +287,7 @@ async fn ignores_reasoning_before_last_user() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn skips_empty_reasoning_segments() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![
         user_message("u1"),
@@ -342,12 +303,7 @@ async fn skips_empty_reasoning_segments() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn suppresses_duplicate_assistant_messages() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let body = run_request(vec![assistant_message("dup"), assistant_message("dup")]).await;
     let messages = messages_from(&body);

codex-rs/core/tests/chat_completions_sse.rs

@@ -10,11 +10,11 @@ use codex_core::Prompt;
 use codex_core::ResponseEvent;
 use codex_core::ResponseItem;
 use codex_core::WireApi;
-use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
 use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::ConversationId;
 use codex_protocol::models::ReasoningItemContent;
 use core_test_support::load_default_config_for_test;
+use core_test_support::skip_if_no_network;
 use futures::StreamExt;
 use tempfile::TempDir;
 use wiremock::Mock;
@@ -23,10 +23,6 @@ use wiremock::ResponseTemplate;
 use wiremock::matchers::method;
 use wiremock::matchers::path;
 
-fn network_disabled() -> bool {
-    std::env::var(CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok()
-}
-
 async fn run_stream(sse_body: &str) -> Vec<ResponseEvent> {
     run_stream_with_bytes(sse_body.as_bytes()).await
 }
@@ -157,12 +153,7 @@ fn assert_reasoning(item: &ResponseItem, expected: &str) {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn streams_text_without_reasoning() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse = concat!(
         "data: {\"choices\":[{\"delta\":{\"content\":\"hi\"}}]}\n\n",
@@ -193,12 +184,7 @@ async fn streams_text_without_reasoning() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn streams_reasoning_from_string_delta() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse = concat!(
         "data: {\"choices\":[{\"delta\":{\"reasoning\":\"think1\"}}]}\n\n",
@@ -215,7 +201,13 @@ async fn streams_reasoning_from_string_delta() {
     }
 
     match &events[1] {
-        ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "think1"),
+        ResponseEvent::ReasoningContentDelta {
+            delta,
+            content_index,
+        } => {
+            assert_eq!(delta, "think1");
+            assert_eq!(content_index, &0);
+        }
         other => panic!("expected reasoning delta, got {other:?}"),
     }
 
@@ -244,12 +236,7 @@ async fn streams_reasoning_from_string_delta() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn streams_reasoning_from_object_delta() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse = concat!(
         "data: {\"choices\":[{\"delta\":{\"reasoning\":{\"text\":\"partA\"}}}]}\n\n",
@@ -267,12 +254,24 @@ async fn streams_reasoning_from_object_delta() {
     }
 
     match &events[1] {
-        ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partA"),
+        ResponseEvent::ReasoningContentDelta {
+            delta,
+            content_index,
+        } => {
+            assert_eq!(delta, "partA");
+            assert_eq!(content_index, &0);
+        }
         other => panic!("expected reasoning delta, got {other:?}"),
     }
 
     match &events[2] {
-        ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partB"),
+        ResponseEvent::ReasoningContentDelta {
+            delta,
+            content_index,
+        } => {
+            assert_eq!(delta, "partB");
+            assert_eq!(content_index, &1);
+        }
         other => panic!("expected reasoning delta, got {other:?}"),
     }
 
@@ -301,12 +300,7 @@ async fn streams_reasoning_from_object_delta() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn streams_reasoning_from_final_message() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse = "data: {\"choices\":[{\"message\":{\"reasoning\":\"final-cot\"},\"finish_reason\":\"stop\"}]}\n\n";
 
@@ -319,7 +313,13 @@ async fn streams_reasoning_from_final_message() {
     }
 
     match &events[1] {
-        ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "final-cot"),
+        ResponseEvent::ReasoningContentDelta {
+            delta,
+            content_index,
+        } => {
+            assert_eq!(delta, "final-cot");
+            assert_eq!(content_index, &0);
+        }
         other => panic!("expected reasoning delta, got {other:?}"),
     }
 
@@ -333,12 +333,7 @@ async fn streams_reasoning_from_final_message() {
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn streams_reasoning_before_tool_call() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse = concat!(
         "data: {\"choices\":[{\"delta\":{\"reasoning\":\"pre-tool\"}}]}\n\n",
@@ -354,7 +349,13 @@ async fn streams_reasoning_before_tool_call() {
     }
 
     match &events[1] {
-        ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "pre-tool"),
+        ResponseEvent::ReasoningContentDelta {
+            delta,
+            content_index,
+        } => {
+            assert_eq!(delta, "pre-tool");
+            assert_eq!(content_index, &0);
+        }
         other => panic!("expected reasoning delta, got {other:?}"),
     }
 
@@ -383,12 +384,7 @@ async fn streams_reasoning_before_tool_call() {
 #[tokio::test]
 #[traced_test]
 async fn chat_sse_emits_failed_on_parse_error() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse_body = concat!("data: not-json\n\n", "data: [DONE]\n\n");
 
@@ -420,12 +416,7 @@ async fn chat_sse_emits_failed_on_parse_error() {
 #[tokio::test]
 #[traced_test]
 async fn chat_sse_done_chunk_emits_event() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let sse_body = "data: [DONE]\n\n";
 
@@ -443,12 +434,7 @@ async fn chat_sse_done_chunk_emits_event() {
 #[tokio::test]
 #[traced_test]
 async fn chat_sse_emits_error_on_invalid_utf8() {
-    if network_disabled() {
-        println!(
-            "Skipping test because it cannot execute when network is disabled in a Codex sandbox."
-        );
-        return;
-    }
+    skip_if_no_network!();
 
     let _ = run_stream_with_bytes(b"data: \x80\x80\n\n").await;
 

codex-rs/core/tests/common/responses.rs

@@ -1,6 +1,7 @@
 use std::sync::Arc;
 use std::sync::Mutex;
 
+use anyhow::Result;
 use serde_json::Value;
 use wiremock::BodyPrintLimit;
 use wiremock::Match;
@@ -12,6 +13,8 @@ use wiremock::ResponseTemplate;
 use wiremock::matchers::method;
 use wiremock::matchers::path_regex;
 
+use crate::test_codex::ApplyPatchModelOutput;
+
 #[derive(Debug, Clone)]
 pub struct ResponseMock {
     requests: Arc<Mutex<Vec<ResponsesRequest>>>,
@@ -36,6 +39,10 @@ impl ResponseMock {
         self.requests.lock().unwrap().clone()
     }
 
+    pub fn last_request(&self) -> Option<ResponsesRequest> {
+        self.requests.lock().unwrap().last().cloned()
+    }
+
     /// Returns true if any captured request contains a `function_call` with the
     /// provided `call_id`.
     pub fn saw_function_call(&self, call_id: &str) -> bool {
@@ -128,6 +135,42 @@ impl ResponsesRequest {
             .map(str::to_string)
     }
 
+    pub fn function_call_output_content_and_success(
+        &self,
+        call_id: &str,
+    ) -> Option<(Option<String>, Option<bool>)> {
+        self.call_output_content_and_success(call_id, "function_call_output")
+    }
+
+    pub fn custom_tool_call_output_content_and_success(
+        &self,
+        call_id: &str,
+    ) -> Option<(Option<String>, Option<bool>)> {
+        self.call_output_content_and_success(call_id, "custom_tool_call_output")
+    }
+
+    fn call_output_content_and_success(
+        &self,
+        call_id: &str,
+        call_type: &str,
+    ) -> Option<(Option<String>, Option<bool>)> {
+        let output = self
+            .call_output(call_id, call_type)
+            .get("output")
+            .cloned()
+            .unwrap_or(Value::Null);
+        match output {
+            Value::String(text) => Some((Some(text), None)),
+            Value::Object(obj) => Some((
+                obj.get("content")
+                    .and_then(Value::as_str)
+                    .map(str::to_string),
+                obj.get("success").and_then(Value::as_bool),
+            )),
+            _ => Some((None, None)),
+        }
+    }
+
     pub fn header(&self, name: &str) -> Option<String> {
         self.0
             .headers
@@ -294,6 +337,7 @@ pub fn ev_reasoning_summary_text_delta(delta: &str) -> Value {
     serde_json::json!({
         "type": "response.reasoning_summary_text.delta",
         "delta": delta,
+        "summary_index": 0,
     })
 }
 
@@ -301,6 +345,7 @@ pub fn ev_reasoning_text_delta(delta: &str) -> Value {
     serde_json::json!({
         "type": "response.reasoning_text.delta",
         "delta": delta,
+        "content_index": 0,
     })
 }
 
@@ -367,6 +412,21 @@ pub fn ev_local_shell_call(call_id: &str, status: &str, command: Vec<&str>) -> V
     })
 }
 
+pub fn ev_apply_patch_call(
+    call_id: &str,
+    patch: &str,
+    output_type: ApplyPatchModelOutput,
+) -> Value {
+    match output_type {
+        ApplyPatchModelOutput::Freeform => ev_apply_patch_custom_tool_call(call_id, patch),
+        ApplyPatchModelOutput::Function => ev_apply_patch_function_call(call_id, patch),
+        ApplyPatchModelOutput::Shell => ev_apply_patch_shell_call(call_id, patch),
+        ApplyPatchModelOutput::ShellViaHeredoc => {
+            ev_apply_patch_shell_call_via_heredoc(call_id, patch)
+        }
+    }
+}
+
 /// Convenience: SSE event for an `apply_patch` custom tool call with raw patch
 /// text. This mirrors the payload produced by the Responses API when the model
 /// invokes `apply_patch` directly (before we convert it to a function call).
@@ -400,6 +460,21 @@ pub fn ev_apply_patch_function_call(call_id: &str, patch: &str) -> Value {
     })
 }
 
+pub fn ev_apply_patch_shell_call(call_id: &str, patch: &str) -> Value {
+    let args = serde_json::json!({ "command": ["apply_patch", patch] });
+    let arguments = serde_json::to_string(&args).expect("serialize apply_patch arguments");
+
+    ev_function_call(call_id, "shell", &arguments)
+}
+
+pub fn ev_apply_patch_shell_call_via_heredoc(call_id: &str, patch: &str) -> Value {
+    let script = format!("apply_patch <<'EOF'\n{patch}\nEOF\n");
+    let args = serde_json::json!({ "command": ["bash", "-lc", script] });
+    let arguments = serde_json::to_string(&args).expect("serialize apply_patch arguments");
+
+    ev_function_call(call_id, "shell", &arguments)
+}
+
 pub fn sse_failed(id: &str, code: &str, message: &str) -> String {
     sse(vec![serde_json::json!({
         "type": "response.failed",
@@ -453,6 +528,37 @@ pub async fn start_mock_server() -> MockServer {
         .await
 }
 
+#[derive(Clone)]
+pub struct FunctionCallResponseMocks {
+    pub function_call: ResponseMock,
+    pub completion: ResponseMock,
+}
+
+pub async fn mount_function_call_agent_response(
+    server: &MockServer,
+    call_id: &str,
+    arguments: &str,
+    tool_name: &str,
+) -> FunctionCallResponseMocks {
+    let first_response = sse(vec![
+        ev_response_created("resp-1"),
+        ev_function_call(call_id, tool_name, arguments),
+        ev_completed("resp-1"),
+    ]);
+    let function_call = mount_sse_once(server, first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    let completion = mount_sse_once(server, second_response).await;
+
+    FunctionCallResponseMocks {
+        function_call,
+        completion,
+    }
+}
+
 /// Mounts a sequence of SSE response bodies and serves them in order for each
 /// POST to `/v1/responses`. Panics if more requests are received than bodies
 /// provided. Also asserts the exact number of expected calls.

codex-rs/core/tests/common/test_codex.rs

@@ -11,6 +11,7 @@ use codex_core::ModelProviderInfo;
 use codex_core::built_in_model_providers;
 use codex_core::config::Config;
 use codex_core::features::Feature;
+use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::Op;
@@ -28,6 +29,15 @@ use crate::wait_for_event;
 
 type ConfigMutator = dyn FnOnce(&mut Config) + Send;
 
+/// A collection of different ways the model can output an apply_patch call
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
+pub enum ApplyPatchModelOutput {
+    Freeform,
+    Function,
+    Shell,
+    ShellViaHeredoc,
+}
+
 pub struct TestCodexBuilder {
     config_mutators: Vec<Box<ConfigMutator>>,
 }
@@ -41,6 +51,14 @@ impl TestCodexBuilder {
         self
     }
 
+    pub fn with_model(self, model: &str) -> Self {
+        let new_model = model.to_string();
+        self.with_config(move |config| {
+            config.model = new_model.clone();
+            config.model_family = find_family_for_model(&new_model).expect("model family");
+        })
+    }
+
     pub async fn build(&mut self, server: &wiremock::MockServer) -> anyhow::Result<TestCodex> {
         let home = Arc::new(TempDir::new()?);
         self.build_with_home(server, home, None).await
@@ -62,6 +80,7 @@ impl TestCodexBuilder {
         resume_from: Option<PathBuf>,
     ) -> anyhow::Result<TestCodex> {
         let (config, cwd) = self.prepare_config(server, &home).await?;
+
         let conversation_manager = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"));
 
         let new_conversation = match resume_from {
@@ -70,15 +89,20 @@ impl TestCodexBuilder {
                     CodexAuth::from_api_key("dummy"),
                 );
                 conversation_manager
-                    .resume_conversation_from_rollout(config, path, auth_manager)
+                    .resume_conversation_from_rollout(config.clone(), path, auth_manager)
+                    .await?
+            }
+            None => {
+                conversation_manager
+                    .new_conversation(config.clone())
                     .await?
             }
-            None => conversation_manager.new_conversation(config).await?,
         };
 
         Ok(TestCodex {
             home,
             cwd,
+            config,
             codex: new_conversation.conversation,
             session_configured: new_conversation.session_configured,
         })
@@ -122,6 +146,7 @@ pub struct TestCodex {
     pub cwd: Arc<TempDir>,
     pub codex: Arc<CodexConversation>,
     pub session_configured: SessionConfiguredEvent,
+    pub config: Config,
 }
 
 impl TestCodex {
@@ -134,14 +159,28 @@ impl TestCodex {
     }
 
     pub async fn submit_turn(&self, prompt: &str) -> Result<()> {
-        self.submit_turn_with_policy(prompt, SandboxPolicy::DangerFullAccess)
-            .await
+        self.submit_turn_with_policies(
+            prompt,
+            AskForApproval::Never,
+            SandboxPolicy::DangerFullAccess,
+        )
+        .await
     }
 
     pub async fn submit_turn_with_policy(
         &self,
         prompt: &str,
         sandbox_policy: SandboxPolicy,
+    ) -> Result<()> {
+        self.submit_turn_with_policies(prompt, AskForApproval::Never, sandbox_policy)
+            .await
+    }
+
+    pub async fn submit_turn_with_policies(
+        &self,
+        prompt: &str,
+        approval_policy: AskForApproval,
+        sandbox_policy: SandboxPolicy,
     ) -> Result<()> {
         let session_model = self.session_configured.model.clone();
         self.codex
@@ -151,7 +190,7 @@ impl TestCodex {
                 }],
                 final_output_json_schema: None,
                 cwd: self.cwd.path().to_path_buf(),
-                approval_policy: AskForApproval::Never,
+                approval_policy,
                 sandbox_policy,
                 model: session_model,
                 effort: None,
@@ -249,6 +288,19 @@ impl TestCodexHarness {
             .expect("output string")
             .to_string()
     }
+
+    pub async fn apply_patch_output(
+        &self,
+        call_id: &str,
+        output_type: ApplyPatchModelOutput,
+    ) -> String {
+        match output_type {
+            ApplyPatchModelOutput::Freeform => self.custom_tool_call_output(call_id).await,
+            ApplyPatchModelOutput::Function
+            | ApplyPatchModelOutput::Shell
+            | ApplyPatchModelOutput::ShellViaHeredoc => self.function_call_stdout(call_id).await,
+        }
+    }
 }
 
 fn custom_tool_call_output<'a>(bodies: &'a [Value], call_id: &str) -> &'a Value {

codex-rs/core/tests/suite/apply_patch_cli.rs

@@ -1,6 +1,8 @@
 #![allow(clippy::expect_used)]
 
 use anyhow::Result;
+use core_test_support::responses::ev_apply_patch_call;
+use core_test_support::test_codex::ApplyPatchModelOutput;
 use pretty_assertions::assert_eq;
 use std::fs;
 
@@ -25,6 +27,7 @@ use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodexHarness;
 use core_test_support::wait_for_event;
 use serde_json::json;
+use test_case::test_case;
 
 async fn apply_patch_harness() -> Result<TestCodexHarness> {
     apply_patch_harness_with(|_| {}).await
@@ -45,19 +48,25 @@ async fn mount_apply_patch(
     call_id: &str,
     patch: &str,
     assistant_msg: &str,
+    output_type: ApplyPatchModelOutput,
 ) {
     mount_sse_sequence(
         harness.server(),
-        apply_patch_responses(call_id, patch, assistant_msg),
+        apply_patch_responses(call_id, patch, assistant_msg, output_type),
     )
     .await;
 }
 
-fn apply_patch_responses(call_id: &str, patch: &str, assistant_msg: &str) -> Vec<String> {
+fn apply_patch_responses(
+    call_id: &str,
+    patch: &str,
+    assistant_msg: &str,
+    output_type: ApplyPatchModelOutput,
+) -> Vec<String> {
     vec![
         sse(vec![
             ev_response_created("resp-1"),
-            ev_apply_patch_function_call(call_id, patch),
+            ev_apply_patch_call(call_id, patch, output_type),
             ev_completed("resp-1"),
         ]),
         sse(vec![
@@ -68,7 +77,13 @@ fn apply_patch_responses(call_id: &str, patch: &str, assistant_msg: &str) -> Vec
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_multiple_operations_integration() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_multiple_operations_integration(
+    output_type: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness_with(|config| {
@@ -86,11 +101,11 @@ async fn apply_patch_cli_multiple_operations_integration() -> Result<()> {
     let patch = "*** Begin Patch\n*** Add File: nested/new.txt\n+created\n*** Delete File: delete.txt\n*** Update File: modify.txt\n@@\n-line2\n+changed\n*** End Patch";
 
     let call_id = "apply-multi-ops";
-    mount_apply_patch(&harness, call_id, patch, "done").await;
+    mount_apply_patch(&harness, call_id, patch, "done", output_type).await;
 
     harness.submit("please apply multi-ops patch").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, output_type).await;
 
     let expected = r"(?s)^Exit code: 0
 Wall time: [0-9]+(?:\.[0-9]+)? seconds
@@ -113,7 +128,11 @@ D delete.txt
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_multiple_chunks() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_multiple_chunks(model_output: ApplyPatchModelOutput) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -123,7 +142,7 @@ async fn apply_patch_cli_multiple_chunks() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: multi.txt\n@@\n-line2\n+changed2\n@@\n-line4\n+changed4\n*** End Patch";
     let call_id = "apply-multi-chunks";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply multi-chunk patch").await?;
 
@@ -135,7 +154,13 @@ async fn apply_patch_cli_multiple_chunks() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_moves_file_to_new_directory(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -147,7 +172,7 @@ async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/dir/name.txt\n@@\n-old content\n+new content\n*** End Patch";
     let call_id = "apply-move";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply move patch").await?;
 
@@ -157,7 +182,13 @@ async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_updates_file_appends_trailing_newline(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -167,7 +198,7 @@ async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: no_newline.txt\n@@\n-no newline at end\n+first line\n+second line\n*** End Patch";
     let call_id = "apply-append-nl";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply newline patch").await?;
 
@@ -178,7 +209,13 @@ async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_insert_only_hunk_modifies_file(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -188,7 +225,7 @@ async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: insert_only.txt\n@@\n alpha\n+beta\n omega\n*** End Patch";
     let call_id = "apply-insert-only";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("insert lines via apply_patch").await?;
 
@@ -197,7 +234,13 @@ async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_move_overwrites_existing_destination(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -211,7 +254,7 @@ async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/dir/name.txt\n@@\n-from\n+new\n*** End Patch";
     let call_id = "apply-move-overwrite";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply move overwrite patch").await?;
 
@@ -221,7 +264,13 @@ async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_move_without_content_change_has_no_turn_diff(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -236,7 +285,7 @@ async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Resul
 
     let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/name.txt\n@@\n same\n*** End Patch";
     let call_id = "apply-move-no-change";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     let model = test.session_configured.model.clone();
     codex
@@ -272,7 +321,13 @@ async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Resul
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_add_overwrites_existing_file(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -282,7 +337,7 @@ async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Add File: duplicate.txt\n+new content\n*** End Patch";
     let call_id = "apply-add-overwrite";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply add overwrite patch").await?;
 
@@ -291,18 +346,24 @@ async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_rejects_invalid_hunk_header() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_rejects_invalid_hunk_header(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
 
     let patch = "*** Begin Patch\n*** Frobnicate File: foo\n*** End Patch";
     let call_id = "apply-invalid-header";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply invalid header patch").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
 
     assert!(
         out.contains("apply_patch verification failed"),
@@ -316,7 +377,13 @@ async fn apply_patch_cli_rejects_invalid_hunk_header() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_reports_missing_context() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_reports_missing_context(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -327,11 +394,11 @@ async fn apply_patch_cli_reports_missing_context() -> Result<()> {
     let patch =
         "*** Begin Patch\n*** Update File: modify.txt\n@@\n-missing\n+changed\n*** End Patch";
     let call_id = "apply-missing-context";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply missing context patch").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
 
     assert!(
         out.contains("apply_patch verification failed"),
@@ -343,18 +410,24 @@ async fn apply_patch_cli_reports_missing_context() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_reports_missing_target_file() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_reports_missing_target_file(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
 
     let patch = "*** Begin Patch\n*** Update File: missing.txt\n@@\n-nope\n+better\n*** End Patch";
     let call_id = "apply-missing-file";
-    mount_apply_patch(&harness, call_id, patch, "fail").await;
+    mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
 
     harness.submit("attempt to update a missing file").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(
         out.contains("apply_patch verification failed"),
         "expected verification failure message"
@@ -372,18 +445,24 @@ async fn apply_patch_cli_reports_missing_target_file() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_delete_missing_file_reports_error() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_delete_missing_file_reports_error(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
 
     let patch = "*** Begin Patch\n*** Delete File: missing.txt\n*** End Patch";
     let call_id = "apply-delete-missing";
-    mount_apply_patch(&harness, call_id, patch, "fail").await;
+    mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
 
     harness.submit("attempt to delete missing file").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
 
     assert!(
         out.contains("apply_patch verification failed"),
@@ -402,18 +481,22 @@ async fn apply_patch_cli_delete_missing_file_reports_error() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_rejects_empty_patch() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_rejects_empty_patch(model_output: ApplyPatchModelOutput) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
 
     let patch = "*** Begin Patch\n*** End Patch";
     let call_id = "apply-empty";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply empty patch").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(
         out.contains("patch rejected: empty patch"),
         "expected rejection for empty patch: {out}"
@@ -422,7 +505,13 @@ async fn apply_patch_cli_rejects_empty_patch() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_delete_directory_reports_verification_error() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_delete_directory_reports_verification_error(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -431,18 +520,24 @@ async fn apply_patch_cli_delete_directory_reports_verification_error() -> Result
 
     let patch = "*** Begin Patch\n*** Delete File: dir\n*** End Patch";
     let call_id = "apply-delete-dir";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("delete a directory via apply_patch").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(out.contains("apply_patch verification failed"));
     assert!(out.contains("Failed to read"));
     Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_rejects_path_traversal_outside_workspace(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -458,7 +553,7 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
 
     let patch = "*** Begin Patch\n*** Add File: ../escape.txt\n+outside\n*** End Patch";
     let call_id = "apply-path-traversal";
-    mount_apply_patch(&harness, call_id, patch, "fail").await;
+    mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
 
     let sandbox_policy = SandboxPolicy::WorkspaceWrite {
         writable_roots: vec![],
@@ -473,7 +568,7 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
         )
         .await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(
         out.contains(
             "patch rejected: writing outside of the project; rejected by user approval settings"
@@ -488,7 +583,13 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -507,7 +608,7 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
 
     let patch = "*** Begin Patch\n*** Update File: stay.txt\n*** Move to: ../escape-move.txt\n@@\n-from\n+to\n*** End Patch";
     let call_id = "apply-move-traversal";
-    mount_apply_patch(&harness, call_id, patch, "fail").await;
+    mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
 
     let sandbox_policy = SandboxPolicy::WorkspaceWrite {
         writable_roots: vec![],
@@ -519,7 +620,7 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
         .submit_with_policy("attempt move traversal via apply_patch", sandbox_policy)
         .await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(
         out.contains(
             "patch rejected: writing outside of the project; rejected by user approval settings"
@@ -535,7 +636,13 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_verification_failure_has_no_side_effects() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_verification_failure_has_no_side_effects(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness_with(|config| {
@@ -547,7 +654,7 @@ async fn apply_patch_cli_verification_failure_has_no_side_effects() -> Result<()
     let call_id = "apply-partial-no-side-effects";
     let patch = "*** Begin Patch\n*** Add File: created.txt\n+hello\n*** Update File: missing.txt\n@@\n-old\n+new\n*** End Patch";
 
-    mount_apply_patch(&harness, call_id, patch, "failed").await;
+    mount_apply_patch(&harness, call_id, patch, "failed", model_output).await;
 
     harness.submit("attempt partial apply patch").await?;
 
@@ -700,7 +807,14 @@ async fn apply_patch_function_accepts_lenient_heredoc_wrapped_patch() -> Result<
         format!("*** Begin Patch\n*** Add File: {file_name}\n+lenient\n*** End Patch\n");
     let wrapped = format!("<<'EOF'\n{patch_inner}EOF\n");
     let call_id = "apply-lenient";
-    mount_apply_patch(&harness, call_id, &wrapped, "ok").await;
+    mount_apply_patch(
+        &harness,
+        call_id,
+        wrapped.as_str(),
+        "ok",
+        ApplyPatchModelOutput::Function,
+    )
+    .await;
 
     harness.submit("apply lenient heredoc patch").await?;
 
@@ -710,7 +824,11 @@ async fn apply_patch_function_accepts_lenient_heredoc_wrapped_patch() -> Result<
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_end_of_file_anchor(model_output: ApplyPatchModelOutput) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -720,7 +838,7 @@ async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
 
     let patch = "*** Begin Patch\n*** Update File: tail.txt\n@@\n-last\n+end\n*** End of File\n*** End Patch";
     let call_id = "apply-eof";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply EOF-anchored patch").await?;
     assert_eq!(fs::read_to_string(&target)?, "alpha\nend\n");
@@ -728,7 +846,13 @@ async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_cli_missing_second_chunk_context_rejected(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -740,11 +864,11 @@ async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
     let patch =
         "*** Begin Patch\n*** Update File: two_chunks.txt\n@@\n-b\n+B\n\n-d\n+D\n*** End Patch";
     let call_id = "apply-missing-ctx-2nd";
-    mount_apply_patch(&harness, call_id, patch, "fail").await;
+    mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
 
     harness.submit("apply missing context second chunk").await?;
 
-    let out = harness.function_call_stdout(call_id).await;
+    let out = harness.apply_patch_output(call_id, model_output).await;
     assert!(out.contains("apply_patch verification failed"));
     assert!(
         out.contains("Failed to find expected lines in"),
@@ -756,7 +880,13 @@ async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_emits_turn_diff_event_with_unified_diff(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -767,16 +897,7 @@ async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
     let call_id = "apply-diff-event";
     let file = "udiff.txt";
     let patch = format!("*** Begin Patch\n*** Add File: {file}\n+hello\n*** End Patch\n");
-    let first = sse(vec![
-        ev_response_created("resp-1"),
-        ev_apply_patch_function_call(call_id, &patch),
-        ev_completed("resp-1"),
-    ]);
-    let second = sse(vec![
-        ev_assistant_message("msg-1", "ok"),
-        ev_completed("resp-2"),
-    ]);
-    mount_sse_sequence(harness.server(), vec![first, second]).await;
+    mount_apply_patch(&harness, call_id, patch.as_str(), "ok", model_output).await;
 
     let model = test.session_configured.model.clone();
     codex
@@ -814,7 +935,13 @@ async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_turn_diff_for_rename_with_content_change() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_turn_diff_for_rename_with_content_change(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -829,16 +956,7 @@ async fn apply_patch_turn_diff_for_rename_with_content_change() -> Result<()> {
     // Patch: update + move
     let call_id = "apply-rename-change";
     let patch = "*** Begin Patch\n*** Update File: old.txt\n*** Move to: new.txt\n@@\n-old\n+new\n*** End Patch";
-    let first = sse(vec![
-        ev_response_created("resp-1"),
-        ev_apply_patch_function_call(call_id, patch),
-        ev_completed("resp-1"),
-    ]);
-    let second = sse(vec![
-        ev_assistant_message("msg-1", "ok"),
-        ev_completed("resp-2"),
-    ]);
-    mount_sse_sequence(harness.server(), vec![first, second]).await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     let model = test.session_configured.model.clone();
     codex
@@ -1031,7 +1149,13 @@ async fn apply_patch_aggregates_diff_preserves_success_after_failure() -> Result
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn apply_patch_change_context_disambiguates_target() -> Result<()> {
+#[test_case(ApplyPatchModelOutput::Freeform)]
+#[test_case(ApplyPatchModelOutput::Function)]
+#[test_case(ApplyPatchModelOutput::Shell)]
+#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
+async fn apply_patch_change_context_disambiguates_target(
+    model_output: ApplyPatchModelOutput,
+) -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let harness = apply_patch_harness().await?;
@@ -1042,7 +1166,7 @@ async fn apply_patch_change_context_disambiguates_target() -> Result<()> {
     let patch =
         "*** Begin Patch\n*** Update File: multi_ctx.txt\n@@ fn b\n-x=10\n+x=11\n*** End Patch";
     let call_id = "apply-ctx";
-    mount_apply_patch(&harness, call_id, patch, "ok").await;
+    mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
 
     harness.submit("apply with change_context").await?;
 

codex-rs/core/tests/suite/cli_stream.rs

@@ -110,12 +110,7 @@ async fn exec_cli_applies_experimental_instructions_file() {
         "data: {\"type\":\"response.created\",\"response\":{}}\n\n",
         "data: {\"type\":\"response.completed\",\"response\":{\"id\":\"r1\"}}\n\n"
     );
-    let resp_mock = core_test_support::responses::mount_sse_once_match(
-        &server,
-        path("/v1/responses"),
-        sse.to_string(),
-    )
-    .await;
+    let resp_mock = core_test_support::responses::mount_sse_once(&server, sse.to_string()).await;
 
     // Create a temporary instructions file with a unique marker we can assert
     // appears in the outbound request payload.

codex-rs/core/tests/suite/client.rs

@@ -21,6 +21,8 @@ use codex_core::protocol::Op;
 use codex_core::protocol::SessionSource;
 use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::ConversationId;
+use codex_protocol::config_types::ReasoningEffort;
+use codex_protocol::config_types::Verbosity;
 use codex_protocol::models::ReasoningItemContent;
 use codex_protocol::models::ReasoningItemReasoningSummary;
 use codex_protocol::models::WebSearchAction;
@@ -236,9 +238,7 @@ async fn resume_includes_initial_messages_and_sends_prior_items() {
 
     // Mock server that will receive the resumed request
     let server = MockServer::start().await;
-    let resp_mock =
-        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
-            .await;
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
 
     // Configure Codex to resume from our file
     let model_provider = ModelProviderInfo {
@@ -379,9 +379,7 @@ async fn includes_base_instructions_override_in_request() {
     skip_if_no_network!();
     // Mock server
     let server = MockServer::start().await;
-    let resp_mock =
-        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
-            .await;
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -569,9 +567,7 @@ async fn includes_user_instructions_message_in_request() {
     skip_if_no_network!();
     let server = MockServer::start().await;
 
-    let resp_mock =
-        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
-            .await;
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -624,14 +620,231 @@ async fn includes_user_instructions_message_in_request() {
     assert_message_ends_with(&request_body["input"][1], "</environment_context>");
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn includes_configured_effort_in_request() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex()
+        .with_model("gpt-5.1-codex")
+        .with_config(|config| {
+            config.model_reasoning_effort = Some(ReasoningEffort::Medium);
+        })
+        .build(&server)
+        .await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert_eq!(
+        request_body
+            .get("reasoning")
+            .and_then(|t| t.get("effort"))
+            .and_then(|v| v.as_str()),
+        Some("medium")
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn includes_no_effort_in_request() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex()
+        .with_model("gpt-5.1-codex")
+        .build(&server)
+        .await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert_eq!(
+        request_body
+            .get("reasoning")
+            .and_then(|t| t.get("effort"))
+            .and_then(|v| v.as_str()),
+        None
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn includes_default_reasoning_effort_in_request_when_defined_by_model_family()
+-> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex().with_model("gpt-5.1").build(&server).await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert_eq!(
+        request_body
+            .get("reasoning")
+            .and_then(|t| t.get("effort"))
+            .and_then(|v| v.as_str()),
+        Some("medium")
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn includes_default_verbosity_in_request() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex().with_model("gpt-5.1").build(&server).await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert_eq!(
+        request_body
+            .get("text")
+            .and_then(|t| t.get("verbosity"))
+            .and_then(|v| v.as_str()),
+        Some("low")
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn configured_verbosity_not_sent_for_models_without_support() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex()
+        .with_model("gpt-5-codex")
+        .with_config(|config| {
+            config.model_verbosity = Some(Verbosity::High);
+        })
+        .build(&server)
+        .await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert!(
+        request_body
+            .get("text")
+            .and_then(|t| t.get("verbosity"))
+            .is_none()
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn configured_verbosity_is_sent() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+    let server = MockServer::start().await;
+
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
+    let TestCodex { codex, .. } = test_codex()
+        .with_model("gpt-5")
+        .with_config(|config| {
+            config.model_verbosity = Some(Verbosity::High);
+        })
+        .build(&server)
+        .await?;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![UserInput::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
+
+    assert_eq!(
+        request_body
+            .get("text")
+            .and_then(|t| t.get("verbosity"))
+            .and_then(|v| v.as_str()),
+        Some("high")
+    );
+
+    Ok(())
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn includes_developer_instructions_message_in_request() {
     skip_if_no_network!();
     let server = MockServer::start().await;
 
-    let resp_mock =
-        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
-            .await;
+    let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),

codex-rs/core/tests/suite/compact.rs

@@ -14,6 +14,7 @@ use codex_protocol::user_input::UserInput;
 use core_test_support::load_default_config_for_test;
 use core_test_support::skip_if_no_network;
 use core_test_support::wait_for_event;
+use core_test_support::wait_for_event_match;
 use std::collections::VecDeque;
 use tempfile::TempDir;
 
@@ -366,6 +367,72 @@ async fn manual_compact_uses_custom_prompt() {
     assert!(!found_default_prompt, "default prompt should be replaced");
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn manual_compact_emits_estimated_token_usage_event() {
+    skip_if_no_network!();
+
+    let server = start_mock_server().await;
+
+    // Compact run where the API reports zero tokens in usage. Our local
+    // estimator should still compute a non-zero context size for the compacted
+    // history.
+    let sse_compact = sse(vec![
+        ev_assistant_message("m1", SUMMARY_TEXT),
+        ev_completed_with_tokens("r1", 0),
+    ]);
+    mount_sse_once(&server, sse_compact).await;
+
+    let model_provider = ModelProviderInfo {
+        base_url: Some(format!("{}/v1", server.uri())),
+        ..built_in_model_providers()["openai"].clone()
+    };
+    let home = TempDir::new().unwrap();
+    let mut config = load_default_config_for_test(&home);
+    config.model_provider = model_provider;
+    set_test_compact_prompt(&mut config);
+
+    let conversation_manager = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"));
+    let NewConversation {
+        conversation: codex,
+        ..
+    } = conversation_manager.new_conversation(config).await.unwrap();
+
+    // Trigger manual compact and collect TokenCount events for the compact turn.
+    codex.submit(Op::Compact).await.unwrap();
+
+    // First TokenCount: from the compact API call (usage.total_tokens = 0).
+    let first = wait_for_event_match(&codex, |ev| match ev {
+        EventMsg::TokenCount(tc) => tc
+            .info
+            .as_ref()
+            .map(|info| info.last_token_usage.total_tokens),
+        _ => None,
+    })
+    .await;
+
+    // Second TokenCount: from the local post-compaction estimate.
+    let last = wait_for_event_match(&codex, |ev| match ev {
+        EventMsg::TokenCount(tc) => tc
+            .info
+            .as_ref()
+            .map(|info| info.last_token_usage.total_tokens),
+        _ => None,
+    })
+    .await;
+
+    // Ensure the compact task itself completes.
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    assert_eq!(
+        first, 0,
+        "expected first TokenCount from compact API usage to be zero"
+    );
+    assert!(
+        last > 0,
+        "second TokenCount should reflect a non-zero estimated context size after compaction"
+    );
+}
+
 // Windows CI only: bump to 4 workers to prevent SSE/event starvation and test timeouts.
 #[cfg_attr(windows, tokio::test(flavor = "multi_thread", worker_threads = 4))]
 #[cfg_attr(not(windows), tokio::test(flavor = "multi_thread", worker_threads = 2))]

codex-rs/core/tests/suite/grep_files.rs

@@ -2,28 +2,14 @@
 
 use anyhow::Result;
 use codex_core::model_family::find_family_for_model;
-use codex_core::protocol::AskForApproval;
-use codex_core::protocol::EventMsg;
-use codex_core::protocol::Op;
-use codex_core::protocol::SandboxPolicy;
-use codex_protocol::config_types::ReasoningSummary;
-use codex_protocol::user_input::UserInput;
-use core_test_support::responses;
-use core_test_support::responses::ev_assistant_message;
-use core_test_support::responses::ev_completed;
-use core_test_support::responses::ev_function_call;
-use core_test_support::responses::ev_response_created;
-use core_test_support::responses::sse;
+use core_test_support::responses::mount_function_call_agent_response;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
-use core_test_support::wait_for_event;
-use serde_json::Value;
 use std::collections::HashSet;
 use std::path::Path;
 use std::process::Command as StdCommand;
-use wiremock::matchers::any;
 
 const MODEL_WITH_TOOL: &str = "test-gpt-5-codex";
 
@@ -69,18 +55,22 @@ async fn grep_files_tool_collects_matches() -> Result<()> {
     })
     .to_string();
 
-    mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
-    submit_turn(&test, "please find uses of needle").await?;
+    let mocks =
+        mount_function_call_agent_response(&server, call_id, &arguments, "grep_files").await;
+    test.submit_turn("please find uses of needle").await?;
 
-    let bodies = recorded_bodies(&server).await?;
-    let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
-    let payload = tool_output.get("output").expect("output field present");
-    let (content_opt, success_opt) = extract_content_and_success(payload);
+    let req = mocks.completion.single_request();
+    let (content_opt, success_opt) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("tool output present");
     let content = content_opt.expect("content present");
     let success = success_opt.unwrap_or(true);
-    assert!(success, "expected success for matches, got {payload:?}");
+    assert!(
+        success,
+        "expected success for matches, got content={content}"
+    );
 
-    let entries = collect_file_names(content);
+    let entries = collect_file_names(&content);
     assert_eq!(entries.len(), 2, "content: {content}");
     assert!(
         entries.contains("alpha.rs"),
@@ -118,16 +108,17 @@ async fn grep_files_tool_reports_empty_results() -> Result<()> {
     })
     .to_string();
 
-    mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
-    submit_turn(&test, "search again").await?;
+    let mocks =
+        mount_function_call_agent_response(&server, call_id, &arguments, "grep_files").await;
+    test.submit_turn("search again").await?;
 
-    let bodies = recorded_bodies(&server).await?;
-    let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
-    let payload = tool_output.get("output").expect("output field present");
-    let (content_opt, success_opt) = extract_content_and_success(payload);
+    let req = mocks.completion.single_request();
+    let (content_opt, success_opt) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("tool output present");
     let content = content_opt.expect("content present");
     if let Some(success) = success_opt {
-        assert!(!success, "expected success=false payload: {payload:?}");
+        assert!(!success, "expected success=false content={content}");
     }
     assert_eq!(content, "No matches found.");
 
@@ -144,73 +135,6 @@ async fn build_test_codex(server: &wiremock::MockServer) -> Result<TestCodex> {
     builder.build(server).await
 }
 
-async fn submit_turn(test: &TestCodex, prompt: &str) -> Result<()> {
-    let session_model = test.session_configured.model.clone();
-
-    test.codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: prompt.into(),
-            }],
-            final_output_json_schema: None,
-            cwd: test.cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
-        .await?;
-
-    wait_for_event(&test.codex, |event| {
-        matches!(event, EventMsg::TaskComplete(_))
-    })
-    .await;
-    Ok(())
-}
-
-async fn mount_tool_sequence(
-    server: &wiremock::MockServer,
-    call_id: &str,
-    arguments: &str,
-    tool_name: &str,
-) {
-    let first_response = sse(vec![
-        ev_response_created("resp-1"),
-        ev_function_call(call_id, tool_name, arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(server, any(), first_response).await;
-
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    responses::mount_sse_once_match(server, any(), second_response).await;
-}
-
-#[allow(clippy::expect_used)]
-async fn recorded_bodies(server: &wiremock::MockServer) -> Result<Vec<Value>> {
-    let requests = server.received_requests().await.expect("requests recorded");
-    Ok(requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect())
-}
-
-fn find_tool_output<'a>(requests: &'a [Value], call_id: &str) -> Option<&'a Value> {
-    requests.iter().find_map(|body| {
-        body.get("input")
-            .and_then(Value::as_array)
-            .and_then(|items| {
-                items.iter().find(|item| {
-                    item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                        && item.get("call_id").and_then(Value::as_str) == Some(call_id)
-                })
-            })
-    })
-}
-
 fn collect_file_names(content: &str) -> HashSet<String> {
     content
         .lines()
@@ -224,14 +148,3 @@ fn collect_file_names(content: &str) -> HashSet<String> {
         })
         .collect()
 }
-
-fn extract_content_and_success(value: &Value) -> (Option<&str>, Option<bool>) {
-    match value {
-        Value::String(text) => (Some(text.as_str()), None),
-        Value::Object(obj) => (
-            obj.get("content").and_then(Value::as_str),
-            obj.get("success").and_then(Value::as_bool),
-        ),
-        _ => (None, None),
-    }
-}

codex-rs/core/tests/suite/items.rs

@@ -18,7 +18,7 @@ use core_test_support::responses::ev_reasoning_text_delta;
 use core_test_support::responses::ev_response_created;
 use core_test_support::responses::ev_web_search_call_added;
 use core_test_support::responses::ev_web_search_call_done;
-use core_test_support::responses::mount_sse_once_match;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
@@ -26,7 +26,6 @@ use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event_match;
 use pretty_assertions::assert_eq;
-use wiremock::matchers::any;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn user_message_item_is_emitted() -> anyhow::Result<()> {
@@ -37,7 +36,7 @@ async fn user_message_item_is_emitted() -> anyhow::Result<()> {
     let TestCodex { codex, .. } = test_codex().build(&server).await?;
 
     let first_response = sse(vec![ev_response_created("resp-1"), ev_completed("resp-1")]);
-    mount_sse_once_match(&server, any(), first_response).await;
+    mount_sse_once(&server, first_response).await;
 
     codex
         .submit(Op::UserInput {
@@ -93,7 +92,7 @@ async fn assistant_message_item_is_emitted() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "all done"),
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), first_response).await;
+    mount_sse_once(&server, first_response).await;
 
     codex
         .submit(Op::UserInput {
@@ -149,7 +148,7 @@ async fn reasoning_item_is_emitted() -> anyhow::Result<()> {
         reasoning_item,
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), first_response).await;
+    mount_sse_once(&server, first_response).await;
 
     codex
         .submit(Op::UserInput {
@@ -207,7 +206,7 @@ async fn web_search_item_is_emitted() -> anyhow::Result<()> {
         web_search_done,
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), first_response).await;
+    mount_sse_once(&server, first_response).await;
 
     codex
         .submit(Op::UserInput {
@@ -259,7 +258,7 @@ async fn agent_message_content_delta_has_item_metadata() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "streamed response"),
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), stream).await;
+    mount_sse_once(&server, stream).await;
 
     codex
         .submit(Op::UserInput {
@@ -324,7 +323,7 @@ async fn reasoning_content_delta_has_item_metadata() -> anyhow::Result<()> {
         ev_reasoning_item("reasoning-1", &["step one"], &[]),
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), stream).await;
+    mount_sse_once(&server, stream).await;
 
     codex
         .submit(Op::UserInput {
@@ -381,7 +380,7 @@ async fn reasoning_raw_content_delta_respects_flag() -> anyhow::Result<()> {
         ev_reasoning_item("reasoning-raw", &["complete"], &["raw detail"]),
         ev_completed("resp-1"),
     ]);
-    mount_sse_once_match(&server, any(), stream).await;
+    mount_sse_once(&server, stream).await;
 
     codex
         .submit(Op::UserInput {

codex-rs/core/tests/suite/list_dir.rs

@@ -1,25 +1,11 @@
 #![cfg(not(target_os = "windows"))]
 
-use codex_core::protocol::AskForApproval;
-use codex_core::protocol::EventMsg;
-use codex_core::protocol::Op;
-use codex_core::protocol::SandboxPolicy;
-use codex_protocol::config_types::ReasoningSummary;
-use codex_protocol::user_input::UserInput;
-use core_test_support::responses;
-use core_test_support::responses::ev_assistant_message;
-use core_test_support::responses::ev_completed;
-use core_test_support::responses::ev_function_call;
-use core_test_support::responses::ev_response_created;
-use core_test_support::responses::sse;
+use core_test_support::responses::mount_function_call_agent_response;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
-use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
-use core_test_support::wait_for_event;
 use pretty_assertions::assert_eq;
-use serde_json::Value;
-use wiremock::matchers::any;
+use serde_json::json;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 #[ignore = "disabled until we enable list_dir tool"]
@@ -27,99 +13,30 @@ async fn list_dir_tool_returns_entries() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
+    let test = test_codex().build(&server).await?;
 
-    let TestCodex {
-        codex,
-        cwd,
-        session_configured,
-        ..
-    } = test_codex().build(&server).await?;
-
-    let dir_path = cwd.path().join("sample_dir");
+    let dir_path = test.cwd.path().join("sample_dir");
     std::fs::create_dir(&dir_path)?;
     std::fs::write(dir_path.join("alpha.txt"), "first file")?;
     std::fs::create_dir(dir_path.join("nested"))?;
     let dir_path = dir_path.to_string_lossy().to_string();
 
     let call_id = "list-dir-call";
-    let arguments = serde_json::json!({
+    let arguments = json!({
         "dir_path": dir_path,
         "offset": 1,
         "limit": 2,
     })
     .to_string();
 
-    let first_response = sse(vec![
-        ev_response_created("resp-1"),
-        ev_function_call(call_id, "list_dir", &arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
-
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
-
-    let session_model = session_configured.model.clone();
-
-    codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: "list directory contents".into(),
-            }],
-            final_output_json_schema: None,
-            cwd: cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
-        .await?;
-
-    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
-
-    let requests = server.received_requests().await.expect("recorded requests");
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().unwrap())
-        .collect::<Vec<_>>();
-    assert!(
-        !request_bodies.is_empty(),
-        "expected at least one request body"
-    );
-
-    let tool_output_item = request_bodies
-        .iter()
-        .find_map(|body| {
-            body.get("input")
-                .and_then(Value::as_array)
-                .and_then(|items| {
-                    items.iter().find(|item| {
-                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    })
-                })
-        })
-        .unwrap_or_else(|| {
-            panic!("function_call_output item not found in requests: {request_bodies:#?}")
-        });
-
-    assert_eq!(
-        tool_output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-
-    let output_text = tool_output_item
-        .get("output")
-        .and_then(|value| match value {
-            Value::String(text) => Some(text.as_str()),
-            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-            _ => None,
-        })
-        .expect("output text present");
-    assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
+    let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
+    test.submit_turn("list directory contents").await?;
+    let req = mocks.completion.single_request();
+    let (content_opt, _) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("function_call_output present");
+    let output = content_opt.expect("output content present in tool output");
+    assert_eq!(output, "E1: [file] alpha.txt\nE2: [dir] nested");
 
     Ok(())
 }
@@ -130,15 +47,9 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
+    let test = test_codex().build(&server).await?;
 
-    let TestCodex {
-        codex,
-        cwd,
-        session_configured,
-        ..
-    } = test_codex().build(&server).await?;
-
-    let dir_path = cwd.path().join("depth_one");
+    let dir_path = test.cwd.path().join("depth_one");
     std::fs::create_dir(&dir_path)?;
     std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
     std::fs::create_dir(dir_path.join("nested"))?;
@@ -146,7 +57,7 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
     let dir_path = dir_path.to_string_lossy().to_string();
 
     let call_id = "list-dir-depth1";
-    let arguments = serde_json::json!({
+    let arguments = json!({
         "dir_path": dir_path,
         "offset": 1,
         "limit": 10,
@@ -154,77 +65,15 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
     })
     .to_string();
 
-    let first_response = sse(vec![
-        ev_response_created("resp-1"),
-        ev_function_call(call_id, "list_dir", &arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
-
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
-
-    let session_model = session_configured.model.clone();
-
-    codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: "list directory contents depth one".into(),
-            }],
-            final_output_json_schema: None,
-            cwd: cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
+    let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
+    test.submit_turn("list directory contents depth one")
         .await?;
-
-    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
-
-    let requests = server.received_requests().await.expect("recorded requests");
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().unwrap())
-        .collect::<Vec<_>>();
-    assert!(
-        !request_bodies.is_empty(),
-        "expected at least one request body"
-    );
-
-    let tool_output_item = request_bodies
-        .iter()
-        .find_map(|body| {
-            body.get("input")
-                .and_then(Value::as_array)
-                .and_then(|items| {
-                    items.iter().find(|item| {
-                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    })
-                })
-        })
-        .unwrap_or_else(|| {
-            panic!("function_call_output item not found in requests: {request_bodies:#?}")
-        });
-
-    assert_eq!(
-        tool_output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-
-    let output_text = tool_output_item
-        .get("output")
-        .and_then(|value| match value {
-            Value::String(text) => Some(text.as_str()),
-            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-            _ => None,
-        })
-        .expect("output text present");
-    assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
+    let req = mocks.completion.single_request();
+    let (content_opt, _) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("function_call_output present");
+    let output = content_opt.expect("output content present in tool output");
+    assert_eq!(output, "E1: [file] alpha.txt\nE2: [dir] nested");
 
     Ok(())
 }
@@ -235,15 +84,9 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
+    let test = test_codex().build(&server).await?;
 
-    let TestCodex {
-        codex,
-        cwd,
-        session_configured,
-        ..
-    } = test_codex().build(&server).await?;
-
-    let dir_path = cwd.path().join("depth_two");
+    let dir_path = test.cwd.path().join("depth_two");
     std::fs::create_dir(&dir_path)?;
     std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
     let nested = dir_path.join("nested");
@@ -255,7 +98,7 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
     let dir_path_string = dir_path.to_string_lossy().to_string();
 
     let call_id = "list-dir-depth2";
-    let arguments = serde_json::json!({
+    let arguments = json!({
         "dir_path": dir_path_string,
         "offset": 1,
         "limit": 10,
@@ -263,81 +106,16 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
     })
     .to_string();
 
-    let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
-        ev_function_call(call_id, "list_dir", &arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
-
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
-
-    let session_model = session_configured.model.clone();
-
-    codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: "list directory contents depth two".into(),
-            }],
-            final_output_json_schema: None,
-            cwd: cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
+    let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
+    test.submit_turn("list directory contents depth two")
         .await?;
-
-    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
-
-    let requests = server.received_requests().await.expect("recorded requests");
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().unwrap())
-        .collect::<Vec<_>>();
-    assert!(
-        !request_bodies.is_empty(),
-        "expected at least one request body"
-    );
-
-    let tool_output_item = request_bodies
-        .iter()
-        .find_map(|body| {
-            body.get("input")
-                .and_then(Value::as_array)
-                .and_then(|items| {
-                    items.iter().find(|item| {
-                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    })
-                })
-        })
-        .unwrap_or_else(|| {
-            panic!("function_call_output item not found in requests: {request_bodies:#?}")
-        });
-
+    let req = mocks.completion.single_request();
+    let (content_opt, _) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("function_call_output present");
+    let output = content_opt.expect("output content present in tool output");
     assert_eq!(
-        tool_output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-
-    let output_text = tool_output_item
-        .get("output")
-        .and_then(|value| match value {
-            Value::String(text) => Some(text.as_str()),
-            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-            _ => None,
-        })
-        .expect("output text present");
-    assert_eq!(
-        output_text,
+        output,
         "E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand"
     );
 
@@ -350,15 +128,9 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
+    let test = test_codex().build(&server).await?;
 
-    let TestCodex {
-        codex,
-        cwd,
-        session_configured,
-        ..
-    } = test_codex().build(&server).await?;
-
-    let dir_path = cwd.path().join("depth_three");
+    let dir_path = test.cwd.path().join("depth_three");
     std::fs::create_dir(&dir_path)?;
     std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
     let nested = dir_path.join("nested");
@@ -370,7 +142,7 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
     let dir_path_string = dir_path.to_string_lossy().to_string();
 
     let call_id = "list-dir-depth3";
-    let arguments = serde_json::json!({
+    let arguments = json!({
         "dir_path": dir_path_string,
         "offset": 1,
         "limit": 10,
@@ -378,81 +150,16 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
     })
     .to_string();
 
-    let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
-        ev_function_call(call_id, "list_dir", &arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
-
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
-
-    let session_model = session_configured.model.clone();
-
-    codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: "list directory contents depth three".into(),
-            }],
-            final_output_json_schema: None,
-            cwd: cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
+    let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
+    test.submit_turn("list directory contents depth three")
         .await?;
-
-    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
-
-    let requests = server.received_requests().await.expect("recorded requests");
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().unwrap())
-        .collect::<Vec<_>>();
-    assert!(
-        !request_bodies.is_empty(),
-        "expected at least one request body"
-    );
-
-    let tool_output_item = request_bodies
-        .iter()
-        .find_map(|body| {
-            body.get("input")
-                .and_then(Value::as_array)
-                .and_then(|items| {
-                    items.iter().find(|item| {
-                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    })
-                })
-        })
-        .unwrap_or_else(|| {
-            panic!("function_call_output item not found in requests: {request_bodies:#?}")
-        });
-
+    let req = mocks.completion.single_request();
+    let (content_opt, _) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("function_call_output present");
+    let output = content_opt.expect("output content present in tool output");
     assert_eq!(
-        tool_output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-
-    let output_text = tool_output_item
-        .get("output")
-        .and_then(|value| match value {
-            Value::String(text) => Some(text.as_str()),
-            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-            _ => None,
-        })
-        .expect("output text present");
-    assert_eq!(
-        output_text,
+        output,
         "E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand\nE5: [file] nested/grand/gamma.txt"
     );
 

codex-rs/core/tests/suite/mod.rs

@@ -18,8 +18,6 @@ mod abort_tasks;
 #[cfg(not(target_os = "windows"))]
 mod apply_patch_cli;
 #[cfg(not(target_os = "windows"))]
-mod apply_patch_freeform;
-#[cfg(not(target_os = "windows"))]
 mod approvals;
 mod auth_refresh;
 mod cli_stream;

codex-rs/core/tests/suite/model_tools.rs

@@ -42,7 +42,7 @@ async fn collect_tool_identifiers_for_model(model: &str) -> Vec<String> {
     let server = MockServer::start().await;
 
     let sse = sse_completed(model);
-    let resp_mock = responses::mount_sse_once_match(&server, wiremock::matchers::any(), sse).await;
+    let resp_mock = responses::mount_sse_once(&server, sse).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -128,4 +128,32 @@ async fn model_selects_expected_tools() {
         ],
         "gpt-5-codex should expose the apply_patch tool",
     );
+
+    let gpt51_codex_tools = collect_tool_identifiers_for_model("gpt-5.1-codex").await;
+    assert_eq!(
+        gpt51_codex_tools,
+        vec![
+            "shell".to_string(),
+            "list_mcp_resources".to_string(),
+            "list_mcp_resource_templates".to_string(),
+            "read_mcp_resource".to_string(),
+            "update_plan".to_string(),
+            "apply_patch".to_string()
+        ],
+        "gpt-5.1-codex should expose the apply_patch tool",
+    );
+
+    let gpt51_tools = collect_tool_identifiers_for_model("gpt-5.1").await;
+    assert_eq!(
+        gpt51_tools,
+        vec![
+            "shell".to_string(),
+            "list_mcp_resources".to_string(),
+            "list_mcp_resource_templates".to_string(),
+            "read_mcp_resource".to_string(),
+            "update_plan".to_string(),
+            "apply_patch".to_string()
+        ],
+        "gpt-5.1 should expose the apply_patch tool",
+    );
 }

codex-rs/core/tests/suite/prompt_caching.rs

@@ -1,9 +1,5 @@
 #![allow(clippy::unwrap_used)]
 
-use codex_core::CodexAuth;
-use codex_core::ConversationManager;
-use codex_core::ModelProviderInfo;
-use codex_core::built_in_model_providers;
 use codex_core::config::OPENAI_DEFAULT_MODEL;
 use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
@@ -16,17 +12,15 @@ use codex_core::protocol_config_types::ReasoningSummary;
 use codex_core::shell::Shell;
 use codex_core::shell::default_user_shell;
 use codex_protocol::user_input::UserInput;
-use core_test_support::load_default_config_for_test;
 use core_test_support::load_sse_fixture_with_id;
+use core_test_support::responses::mount_sse_once;
+use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
+use core_test_support::test_codex::TestCodex;
+use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
 use std::collections::HashMap;
 use tempfile::TempDir;
-use wiremock::Mock;
-use wiremock::MockServer;
-use wiremock::ResponseTemplate;
-use wiremock::matchers::method;
-use wiremock::matchers::path;
 
 fn text_user_input(text: String) -> serde_json::Value {
     serde_json::json!({
@@ -70,48 +64,24 @@ fn assert_tool_names(body: &serde_json::Value, expected_names: &[&str]) {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
-async fn codex_mini_latest_tools() {
-    skip_if_no_network!();
+async fn codex_mini_latest_tools() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    // Expect two POSTs to /v1/responses
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
-    config.features.disable(Feature::ApplyPatchFreeform);
-
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    config.model = "codex-mini-latest".to_string();
-    config.model_family = find_family_for_model("codex-mini-latest").unwrap();
-
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
+    let TestCodex { codex, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+            config.features.disable(Feature::ApplyPatchFreeform);
+            config.model = "codex-mini-latest".to_string();
+            config.model_family = find_family_for_model("codex-mini-latest")
+                .expect("model family for codex-mini-latest");
+        })
+        .build(&server)
+        .await?;
 
     codex
         .submit(Op::UserInput {
@@ -119,8 +89,7 @@ async fn codex_mini_latest_tools() {
                 text: "hello 1".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     codex
@@ -129,72 +98,45 @@ async fn codex_mini_latest_tools() {
                 text: "hello 2".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
     let expected_instructions = [
         include_str!("../../prompt.md"),
         include_str!("../../../apply-patch/apply_patch_tool_instructions.md"),
     ]
     .join("\n");
 
-    let body0 = requests[0].body_json::<serde_json::Value>().unwrap();
+    let body0 = req1.single_request().body_json();
     assert_eq!(
         body0["instructions"],
         serde_json::json!(expected_instructions),
     );
-    let body1 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req2.single_request().body_json();
     assert_eq!(
         body1["instructions"],
         serde_json::json!(expected_instructions),
     );
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
-async fn prompt_tools_are_consistent_across_requests() {
-    skip_if_no_network!();
+async fn prompt_tools_are_consistent_across_requests() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    // Expect two POSTs to /v1/responses
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
-
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
+    let TestCodex { codex, config, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
     let base_instructions = config.model_family.base_instructions.clone();
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
 
     codex
         .submit(Op::UserInput {
@@ -202,8 +144,7 @@ async fn prompt_tools_are_consistent_across_requests() {
                 text: "hello 1".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     codex
@@ -212,13 +153,9 @@ async fn prompt_tools_are_consistent_across_requests() {
                 text: "hello 2".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
     // our internal implementation is responsible for keeping tools in sync
     // with the OpenAI schema, so we just verify the tool presence here
     let tools_by_model: HashMap<&'static str, Vec<&'static str>> = HashMap::from([
@@ -233,6 +170,18 @@ async fn prompt_tools_are_consistent_across_requests() {
                 "view_image",
             ],
         ),
+        (
+            "gpt-5.1",
+            vec![
+                "shell",
+                "list_mcp_resources",
+                "list_mcp_resource_templates",
+                "read_mcp_resource",
+                "update_plan",
+                "apply_patch",
+                "view_image",
+            ],
+        ),
         (
             "gpt-5-codex",
             vec![
@@ -245,12 +194,24 @@ async fn prompt_tools_are_consistent_across_requests() {
                 "view_image",
             ],
         ),
+        (
+            "gpt-5.1-codex",
+            vec![
+                "shell",
+                "list_mcp_resources",
+                "list_mcp_resource_templates",
+                "read_mcp_resource",
+                "update_plan",
+                "apply_patch",
+                "view_image",
+            ],
+        ),
     ]);
     let expected_tools_names = tools_by_model
         .get(OPENAI_DEFAULT_MODEL)
         .unwrap_or_else(|| panic!("expected tools to be defined for model {OPENAI_DEFAULT_MODEL}"))
         .as_slice();
-    let body0 = requests[0].body_json::<serde_json::Value>().unwrap();
+    let body0 = req1.single_request().body_json();
 
     let expected_instructions = if expected_tools_names.contains(&"apply_patch") {
         base_instructions
@@ -268,53 +229,32 @@ async fn prompt_tools_are_consistent_across_requests() {
     );
     assert_tool_names(&body0, expected_tools_names);
 
-    let body1 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req2.single_request().body_json();
     assert_eq!(
         body1["instructions"],
         serde_json::json!(expected_instructions),
     );
     assert_tool_names(&body1, expected_tools_names);
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn prefixes_context_and_instructions_once_and_consistently_across_requests() {
-    skip_if_no_network!();
+async fn prefixes_context_and_instructions_once_and_consistently_across_requests()
+-> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    // Expect two POSTs to /v1/responses
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
-
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
+    let TestCodex { codex, config, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
 
     codex
         .submit(Op::UserInput {
@@ -322,8 +262,7 @@ async fn prefixes_context_and_instructions_once_and_consistently_across_requests
                 text: "hello 1".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     codex
@@ -332,31 +271,14 @@ async fn prefixes_context_and_instructions_once_and_consistently_across_requests
                 text: "hello 2".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
     let shell = default_user_shell().await;
-
-    let expected_env_text = format!(
-        r#"<environment_context>
-  <cwd>{}</cwd>
-  <approval_policy>on-request</approval_policy>
-  <sandbox_mode>read-only</sandbox_mode>
-  <network_access>restricted</network_access>
-{}</environment_context>"#,
-        cwd.path().to_string_lossy(),
-        match shell.name() {
-            Some(name) => format!("  <shell>{name}</shell>\n"),
-            None => String::new(),
-        }
-    );
+    let cwd_str = config.cwd.to_string_lossy();
+    let expected_env_text = default_env_context_str(&cwd_str, &shell);
     let expected_ui_text = format!(
-        "# AGENTS.md instructions for {}\n\n<INSTRUCTIONS>\nbe consistent and helpful\n</INSTRUCTIONS>",
-        cwd.path().to_string_lossy()
+        "# AGENTS.md instructions for {cwd_str}\n\n<INSTRUCTIONS>\nbe consistent and helpful\n</INSTRUCTIONS>"
     );
 
     let expected_env_msg = serde_json::json!({
@@ -375,7 +297,7 @@ async fn prefixes_context_and_instructions_once_and_consistently_across_requests
         "role": "user",
         "content": [ { "type": "input_text", "text": "hello 1" } ]
     });
-    let body1 = requests[0].body_json::<serde_json::Value>().unwrap();
+    let body1 = req1.single_request().body_json();
     assert_eq!(
         body1["input"],
         serde_json::json!([expected_ui_msg, expected_env_msg, expected_user_message_1])
@@ -386,7 +308,7 @@ async fn prefixes_context_and_instructions_once_and_consistently_across_requests
         "role": "user",
         "content": [ { "type": "input_text", "text": "hello 2" } ]
     });
-    let body2 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body2 = req2.single_request().body_json();
     let expected_body2 = serde_json::json!(
         [
             body1["input"].as_array().unwrap().as_slice(),
@@ -395,47 +317,25 @@ async fn prefixes_context_and_instructions_once_and_consistently_across_requests
         .concat()
     );
     assert_eq!(body2["input"], expected_body2);
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() {
-    skip_if_no_network!();
+async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    // Expect two POSTs to /v1/responses
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
-
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
+    let TestCodex { codex, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
 
     // First turn
     codex
@@ -444,8 +344,7 @@ async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() {
                 text: "hello 1".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     let writable = TempDir::new().unwrap();
@@ -463,8 +362,7 @@ async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() {
             effort: Some(Some(ReasoningEffort::High)),
             summary: Some(ReasoningSummary::Detailed),
         })
-        .await
-        .unwrap();
+        .await?;
 
     // Second turn after overrides
     codex
@@ -473,16 +371,11 @@ async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() {
                 text: "hello 2".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    // Verify we issued exactly two requests, and the cached prefix stayed identical.
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
-    let body1 = requests[0].body_json::<serde_json::Value>().unwrap();
-    let body2 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req1.single_request().body_json();
+    let body2 = req2.single_request().body_json();
     // prompt_cache_key should remain constant across overrides
     assert_eq!(
         body1["prompt_cache_key"], body2["prompt_cache_key"],
@@ -523,47 +416,25 @@ async fn overrides_turn_context_but_keeps_cached_prefix_and_key_constant() {
         .concat()
     );
     assert_eq!(body2["input"], expected_body2);
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn per_turn_overrides_keep_cached_prefix_and_key_constant() {
-    skip_if_no_network!();
+async fn per_turn_overrides_keep_cached_prefix_and_key_constant() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    // Expect two POSTs to /v1/responses
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
-
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
+    let TestCodex { codex, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
 
     // First turn
     codex
@@ -572,8 +443,7 @@ async fn per_turn_overrides_keep_cached_prefix_and_key_constant() {
                 text: "hello 1".into(),
             }],
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     // Second turn using per-turn overrides via UserTurn
@@ -597,16 +467,11 @@ async fn per_turn_overrides_keep_cached_prefix_and_key_constant() {
             summary: ReasoningSummary::Detailed,
             final_output_json_schema: None,
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    // Verify we issued exactly two requests, and the cached prefix stayed identical.
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
-    let body1 = requests[0].body_json::<serde_json::Value>().unwrap();
-    let body2 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req1.single_request().body_json();
+    let body2 = req2.single_request().body_json();
 
     // prompt_cache_key should remain constant across per-turn overrides
     assert_eq!(
@@ -647,38 +512,25 @@ async fn per_turn_overrides_keep_cached_prefix_and_key_constant() {
         .concat()
     );
     assert_eq!(body2["input"], expected_body2);
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn send_user_turn_with_no_changes_does_not_send_environment_context() {
-    skip_if_no_network!();
+async fn send_user_turn_with_no_changes_does_not_send_environment_context() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
+    let TestCodex { codex, config, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
 
     let default_cwd = config.cwd.clone();
     let default_approval_policy = config.approval_policy;
@@ -687,14 +539,6 @@ async fn send_user_turn_with_no_changes_does_not_send_environment_context() {
     let default_effort = config.model_reasoning_effort;
     let default_summary = config.model_reasoning_summary;
 
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    let codex = conversation_manager
-        .new_conversation(config)
-        .await
-        .expect("create new conversation")
-        .conversation;
-
     codex
         .submit(Op::UserTurn {
             items: vec![UserInput::Text {
@@ -708,8 +552,7 @@ async fn send_user_turn_with_no_changes_does_not_send_environment_context() {
             summary: default_summary,
             final_output_json_schema: None,
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     codex
@@ -725,27 +568,20 @@ async fn send_user_turn_with_no_changes_does_not_send_environment_context() {
             summary: default_summary,
             final_output_json_schema: None,
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
-    let body1 = requests[0].body_json::<serde_json::Value>().unwrap();
-    let body2 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req1.single_request().body_json();
+    let body2 = req2.single_request().body_json();
 
     let shell = default_user_shell().await;
+    let default_cwd_lossy = default_cwd.to_string_lossy();
     let expected_ui_text = format!(
-        "# AGENTS.md instructions for {}\n\n<INSTRUCTIONS>\nbe consistent and helpful\n</INSTRUCTIONS>",
-        default_cwd.to_string_lossy()
+        "# AGENTS.md instructions for {default_cwd_lossy}\n\n<INSTRUCTIONS>\nbe consistent and helpful\n</INSTRUCTIONS>"
     );
     let expected_ui_msg = text_user_input(expected_ui_text);
 
-    let expected_env_msg_1 = text_user_input(default_env_context_str(
-        &cwd.path().to_string_lossy(),
-        &shell,
-    ));
+    let expected_env_msg_1 = text_user_input(default_env_context_str(&default_cwd_lossy, &shell));
     let expected_user_message_1 = text_user_input("hello 1".to_string());
 
     let expected_input_1 = serde_json::Value::Array(vec![
@@ -763,38 +599,25 @@ async fn send_user_turn_with_no_changes_does_not_send_environment_context() {
         expected_user_message_2,
     ]);
     assert_eq!(body2["input"], expected_input_2);
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn send_user_turn_with_changes_sends_environment_context() {
-    skip_if_no_network!();
+async fn send_user_turn_with_changes_sends_environment_context() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
     use pretty_assertions::assert_eq;
 
-    let server = MockServer::start().await;
+    let server = start_mock_server().await;
 
-    let sse = sse_completed("resp");
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(2)
-        .mount(&server)
-        .await;
-
-    let model_provider = ModelProviderInfo {
-        base_url: Some(format!("{}/v1", server.uri())),
-        ..built_in_model_providers()["openai"].clone()
-    };
-
-    let cwd = TempDir::new().unwrap();
-    let codex_home = TempDir::new().unwrap();
-    let mut config = load_default_config_for_test(&codex_home);
-    config.cwd = cwd.path().to_path_buf();
-    config.model_provider = model_provider;
-    config.user_instructions = Some("be consistent and helpful".to_string());
+    let req1 = mount_sse_once(&server, sse_completed("resp-1")).await;
+    let req2 = mount_sse_once(&server, sse_completed("resp-2")).await;
+    let TestCodex { codex, config, .. } = test_codex()
+        .with_config(|config| {
+            config.user_instructions = Some("be consistent and helpful".to_string());
+        })
+        .build(&server)
+        .await?;
 
     let default_cwd = config.cwd.clone();
     let default_approval_policy = config.approval_policy;
@@ -803,14 +626,6 @@ async fn send_user_turn_with_changes_sends_environment_context() {
     let default_effort = config.model_reasoning_effort;
     let default_summary = config.model_reasoning_summary;
 
-    let conversation_manager =
-        ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    let codex = conversation_manager
-        .new_conversation(config.clone())
-        .await
-        .expect("create new conversation")
-        .conversation;
-
     codex
         .submit(Op::UserTurn {
             items: vec![UserInput::Text {
@@ -824,8 +639,7 @@ async fn send_user_turn_with_changes_sends_environment_context() {
             summary: default_summary,
             final_output_json_schema: None,
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     codex
@@ -841,15 +655,11 @@ async fn send_user_turn_with_changes_sends_environment_context() {
             summary: ReasoningSummary::Detailed,
             final_output_json_schema: None,
         })
-        .await
-        .unwrap();
+        .await?;
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(requests.len(), 2, "expected two POST requests");
-
-    let body1 = requests[0].body_json::<serde_json::Value>().unwrap();
-    let body2 = requests[1].body_json::<serde_json::Value>().unwrap();
+    let body1 = req1.single_request().body_json();
+    let body2 = req2.single_request().body_json();
 
     let shell = default_user_shell().await;
     let expected_ui_text = format!(
@@ -888,4 +698,6 @@ async fn send_user_turn_with_changes_sends_environment_context() {
         expected_user_message_2,
     ]);
     assert_eq!(body2["input"], expected_input_2);
+
+    Ok(())
 }

codex-rs/core/tests/suite/read_file.rs

@@ -1,25 +1,11 @@
 #![cfg(not(target_os = "windows"))]
 
-use codex_core::protocol::AskForApproval;
-use codex_core::protocol::EventMsg;
-use codex_core::protocol::Op;
-use codex_core::protocol::SandboxPolicy;
-use codex_protocol::config_types::ReasoningSummary;
-use codex_protocol::user_input::UserInput;
-use core_test_support::responses;
-use core_test_support::responses::ev_assistant_message;
-use core_test_support::responses::ev_completed;
-use core_test_support::responses::ev_function_call;
-use core_test_support::responses::ev_response_created;
-use core_test_support::responses::sse;
+use core_test_support::responses::mount_function_call_agent_response;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
-use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
-use core_test_support::wait_for_event;
 use pretty_assertions::assert_eq;
-use serde_json::Value;
-use wiremock::matchers::any;
+use serde_json::json;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 #[ignore = "disabled until we enable read_file tool"]
@@ -27,72 +13,29 @@ async fn read_file_tool_returns_requested_lines() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
+    let test = test_codex().build(&server).await?;
 
-    let TestCodex {
-        codex,
-        cwd,
-        session_configured,
-        ..
-    } = test_codex().build(&server).await?;
-
-    let file_path = cwd.path().join("sample.txt");
+    let file_path = test.cwd.path().join("sample.txt");
     std::fs::write(&file_path, "first\nsecond\nthird\nfourth\n")?;
     let file_path = file_path.to_string_lossy().to_string();
 
     let call_id = "read-file-call";
-    let arguments = serde_json::json!({
+    let arguments = json!({
         "file_path": file_path,
         "offset": 2,
         "limit": 2,
     })
     .to_string();
 
-    let first_response = sse(vec![
-        ev_response_created("resp-1"),
-        ev_function_call(call_id, "read_file", &arguments),
-        ev_completed("resp-1"),
-    ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "read_file").await;
 
-    let second_response = sse(vec![
-        ev_assistant_message("msg-1", "done"),
-        ev_completed("resp-2"),
-    ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    test.submit_turn("please inspect sample.txt").await?;
 
-    let session_model = session_configured.model.clone();
-
-    codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: "please inspect sample.txt".into(),
-            }],
-            final_output_json_schema: None,
-            cwd: cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy: SandboxPolicy::DangerFullAccess,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
-        .await?;
-
-    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
-
-    let req = second_mock.single_request();
-    let tool_output_item = req.function_call_output(call_id);
-    assert_eq!(
-        tool_output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-    let output_text = tool_output_item
-        .get("output")
-        .and_then(|value| match value {
-            Value::String(text) => Some(text.as_str()),
-            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-            _ => None,
-        })
-        .expect("output text present");
+    let req = mocks.completion.single_request();
+    let (output_text_opt, _) = req
+        .function_call_output_content_and_success(call_id)
+        .expect("output present");
+    let output_text = output_text_opt.expect("output text present");
     assert_eq!(output_text, "L2: second\nL3: third");
 
     Ok(())

codex-rs/core/tests/suite/resume.rs

@@ -6,14 +6,13 @@ use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_reasoning_item;
 use core_test_support::responses::ev_response_created;
-use core_test_support::responses::mount_sse_once_match;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
 use std::sync::Arc;
-use wiremock::matchers::any;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn resume_includes_initial_messages_from_rollout_events() -> Result<()> {
@@ -31,7 +30,7 @@ async fn resume_includes_initial_messages_from_rollout_events() -> Result<()> {
         ev_assistant_message("msg-1", "Completed first turn"),
         ev_completed("resp-initial"),
     ]);
-    mount_sse_once_match(&server, any(), initial_sse).await;
+    mount_sse_once(&server, initial_sse).await;
 
     codex
         .submit(Op::UserInput {
@@ -83,7 +82,7 @@ async fn resume_includes_initial_messages_from_reasoning_events() -> Result<()>
         ev_assistant_message("msg-1", "Completed reasoning turn"),
         ev_completed("resp-initial"),
     ]);
-    mount_sse_once_match(&server, any(), initial_sse).await;
+    mount_sse_once(&server, initial_sse).await;
 
     codex
         .submit(Op::UserInput {

codex-rs/core/tests/suite/rmcp_client.rs

@@ -21,7 +21,7 @@ use codex_core::protocol::SandboxPolicy;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::user_input::UserInput;
 use core_test_support::responses;
-use core_test_support::responses::mount_sse_once_match;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
@@ -36,7 +36,6 @@ use tokio::process::Child;
 use tokio::process::Command;
 use tokio::time::Instant;
 use tokio::time::sleep;
-use wiremock::matchers::any;
 
 static OPENAI_PNG: &str = "data:image/png;base64,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";
 
@@ -51,9 +50,8 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
     let server_name = "rmcp";
     let tool_name = format!("mcp__{server_name}__echo");
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
@@ -61,9 +59,8 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
         ]),
     )
     .await;
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_assistant_message("msg-1", "rmcp echo tool completed successfully."),
             responses::ev_completed("resp-2"),
@@ -190,9 +187,8 @@ async fn stdio_image_responses_round_trip() -> anyhow::Result<()> {
     let tool_name = format!("mcp__{server_name}__image");
 
     // First stream: model decides to call the image tool.
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{}"),
@@ -201,9 +197,8 @@ async fn stdio_image_responses_round_trip() -> anyhow::Result<()> {
     )
     .await;
     // Second stream: after tool execution, assistant emits a message and completes.
-    let final_mock = mount_sse_once_match(
+    let final_mock = mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_assistant_message("msg-1", "rmcp image tool completed successfully."),
             responses::ev_completed("resp-2"),
@@ -531,9 +526,8 @@ async fn stdio_server_propagates_whitelisted_env_vars() -> anyhow::Result<()> {
     let server_name = "rmcp_whitelist";
     let tool_name = format!("mcp__{server_name}__echo");
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
@@ -541,9 +535,8 @@ async fn stdio_server_propagates_whitelisted_env_vars() -> anyhow::Result<()> {
         ]),
     )
     .await;
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_assistant_message("msg-1", "rmcp echo tool completed successfully."),
             responses::ev_completed("resp-2"),
@@ -666,9 +659,8 @@ async fn streamable_http_tool_call_round_trip() -> anyhow::Result<()> {
     let server_name = "rmcp_http";
     let tool_name = format!("mcp__{server_name}__echo");
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
@@ -676,9 +668,8 @@ async fn streamable_http_tool_call_round_trip() -> anyhow::Result<()> {
         ]),
     )
     .await;
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_assistant_message(
                 "msg-1",
@@ -834,9 +825,8 @@ async fn streamable_http_with_oauth_round_trip() -> anyhow::Result<()> {
     let server_name = "rmcp_http_oauth";
     let tool_name = format!("mcp__{server_name}__echo");
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
@@ -844,9 +834,8 @@ async fn streamable_http_with_oauth_round_trip() -> anyhow::Result<()> {
         ]),
     )
     .await;
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         responses::sse(vec![
             responses::ev_assistant_message(
                 "msg-1",

codex-rs/core/tests/suite/shell_serialization.rs

@@ -3,12 +3,7 @@
 use anyhow::Result;
 use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
-use codex_core::protocol::AskForApproval;
-use codex_core::protocol::EventMsg;
-use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
-use codex_protocol::config_types::ReasoningSummary;
-use codex_protocol::user_input::UserInput;
 use core_test_support::assert_regex_match;
 use core_test_support::responses::ev_apply_patch_function_call;
 use core_test_support::responses::ev_assistant_message;
@@ -21,9 +16,7 @@ use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
-use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
-use core_test_support::wait_for_event;
 use pretty_assertions::assert_eq;
 use regex_lite::Regex;
 use serde_json::Value;
@@ -42,69 +35,6 @@ const FIXTURE_JSON: &str = r#"{
 }
 "#;
 
-async fn submit_turn(test: &TestCodex, prompt: &str, sandbox_policy: SandboxPolicy) -> Result<()> {
-    let session_model = test.session_configured.model.clone();
-
-    test.codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: prompt.into(),
-            }],
-            final_output_json_schema: None,
-            cwd: test.cwd.path().to_path_buf(),
-            approval_policy: AskForApproval::Never,
-            sandbox_policy,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
-        .await?;
-
-    wait_for_event(&test.codex, |event| {
-        matches!(event, EventMsg::TaskComplete(_))
-    })
-    .await;
-
-    Ok(())
-}
-
-fn request_bodies(requests: &[wiremock::Request]) -> Result<Vec<Value>> {
-    requests
-        .iter()
-        .map(|req| Ok(serde_json::from_slice::<Value>(&req.body)?))
-        .collect()
-}
-
-fn find_function_call_output<'a>(bodies: &'a [Value], call_id: &str) -> Option<&'a Value> {
-    for body in bodies {
-        if let Some(items) = body.get("input").and_then(Value::as_array) {
-            for item in items {
-                if item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    && item.get("call_id").and_then(Value::as_str) == Some(call_id)
-                {
-                    return Some(item);
-                }
-            }
-        }
-    }
-    None
-}
-
-fn find_custom_tool_call_output<'a>(bodies: &'a [Value], call_id: &str) -> Option<&'a Value> {
-    for body in bodies {
-        if let Some(items) = body.get("input").and_then(Value::as_array) {
-            for item in items {
-                if item.get("type").and_then(Value::as_str) == Some("custom_tool_call_output")
-                    && item.get("call_id").and_then(Value::as_str) == Some(call_id)
-                {
-                    return Some(item);
-                }
-            }
-        }
-    }
-    None
-}
-
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
     skip_if_no_network!(Ok(()));
@@ -133,21 +63,16 @@ async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the json shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item = find_function_call_output(&bodies, call_id).expect("shell output present");
+    let req = mock.last_request().expect("shell output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -202,22 +127,18 @@ async fn shell_output_is_structured_with_freeform_apply_patch() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the structured shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("structured output present");
+    let req = mock
+        .last_request()
+        .expect("structured shell output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -269,21 +190,16 @@ async fn shell_output_preserves_fixture_json_without_serialization() -> Result<(
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "read the fixture JSON with sed",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item = find_function_call_output(&bodies, call_id).expect("shell output present");
+    let req = mock.last_request().expect("shell output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -345,22 +261,18 @@ async fn shell_output_structures_fixture_with_serialization() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "read the fixture JSON with structured output",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("structured output present");
+    let req = mock
+        .last_request()
+        .expect("structured output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -420,22 +332,18 @@ async fn shell_output_for_freeform_tool_records_duration() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the structured shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("structured output present");
+    let req = mock
+        .last_request()
+        .expect("structured output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -490,22 +398,18 @@ async fn shell_output_reserializes_truncated_content() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the truncation shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("truncated output present");
+    let req = mock
+        .last_request()
+        .expect("truncated output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -570,22 +474,18 @@ async fn apply_patch_custom_tool_output_is_structured() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "apply the patch via custom tool",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let req = mock
+        .last_request()
+        .expect("apply_patch output request recorded");
+    let output_item = req.custom_tool_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -630,22 +530,18 @@ async fn apply_patch_custom_tool_call_creates_file() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "apply the patch via custom tool to create a file",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let req = mock
+        .last_request()
+        .expect("apply_patch output request recorded");
+    let output_item = req.custom_tool_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -699,22 +595,18 @@ async fn apply_patch_custom_tool_call_updates_existing_file() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "apply the patch via custom tool to update a file",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let req = mock
+        .last_request()
+        .expect("apply_patch output request recorded");
+    let output_item = req.custom_tool_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -762,22 +654,18 @@ async fn apply_patch_custom_tool_call_reports_failure_output() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "attempt a failing apply_patch via custom tool",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let req = mock
+        .last_request()
+        .expect("apply_patch output request recorded");
+    let output_item = req.custom_tool_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -817,22 +705,18 @@ async fn apply_patch_function_call_output_is_structured() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "apply the patch via function-call apply_patch",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("apply_patch function output present");
+    let req = mock
+        .last_request()
+        .expect("apply_patch function output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -880,21 +764,16 @@ async fn shell_output_is_structured_for_nonzero_exit() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the failing shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item = find_function_call_output(&bodies, call_id).expect("shell output present");
+    let req = mock.last_request().expect("shell output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)
@@ -934,22 +813,18 @@ async fn local_shell_call_output_is_structured() -> Result<()> {
             ev_completed("resp-2"),
         ]),
     ];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policy(
         "run the local shell command",
         SandboxPolicy::DangerFullAccess,
     )
     .await?;
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("recorded requests present");
-    let bodies = request_bodies(&requests)?;
-    let output_item =
-        find_function_call_output(&bodies, call_id).expect("local shell output present");
+    let req = mock
+        .last_request()
+        .expect("local shell output request recorded");
+    let output_item = req.function_call_output(call_id);
     let output = output_item
         .get("output")
         .and_then(Value::as_str)

codex-rs/core/tests/suite/tool_harness.rs

@@ -14,6 +14,7 @@ use codex_protocol::plan_tool::StepStatus;
 use codex_protocol::user_input::UserInput;
 use core_test_support::assert_regex_match;
 use core_test_support::responses;
+use core_test_support::responses::ResponsesRequest;
 use core_test_support::responses::ev_apply_patch_function_call;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
@@ -28,14 +29,23 @@ use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
 use serde_json::Value;
 use serde_json::json;
-use wiremock::matchers::any;
 
-fn extract_output_text(item: &Value) -> Option<&str> {
-    item.get("output").and_then(|value| match value {
-        Value::String(text) => Some(text.as_str()),
-        Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-        _ => None,
-    })
+fn call_output(req: &ResponsesRequest, call_id: &str) -> (String, Option<bool>) {
+    let raw = req.function_call_output(call_id);
+    assert_eq!(
+        raw.get("call_id").and_then(Value::as_str),
+        Some(call_id),
+        "mismatched call_id in function_call_output"
+    );
+    let (content_opt, success) = match req.function_call_output_content_and_success(call_id) {
+        Some(values) => values,
+        None => panic!("function_call_output present"),
+    };
+    let content = match content_opt {
+        Some(c) => c,
+        None => panic!("function_call_output content present"),
+    };
+    (content, success)
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -62,13 +72,13 @@ async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()>
         ev_local_shell_call(call_id, "completed", command),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "all done"),
         ev_completed("resp-2"),
     ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -90,9 +100,8 @@ async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()>
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
     let req = second_mock.single_request();
-    let output_item = req.function_call_output(call_id);
-    let output_text = extract_output_text(&output_item).expect("output text present");
-    let exec_output: Value = serde_json::from_str(output_text)?;
+    let (output_text, _) = call_output(&req, call_id);
+    let exec_output: Value = serde_json::from_str(&output_text)?;
     assert_eq!(exec_output["metadata"]["exit_code"], 0);
     let stdout = exec_output["output"].as_str().expect("stdout field");
     assert_regex_match(r"(?s)^tool harness\n?$", stdout);
@@ -129,13 +138,13 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
         ev_function_call(call_id, "update_plan", &plan_args),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "plan acknowledged"),
         ev_completed("resp-2"),
     ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -174,12 +183,7 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
     assert!(saw_plan_update, "expected PlanUpdate event");
 
     let req = second_mock.single_request();
-    let output_item = req.function_call_output(call_id);
-    assert_eq!(
-        output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let (output_text, _success_flag) = call_output(&req, call_id);
     assert_eq!(output_text, "Plan updated");
 
     Ok(())
@@ -210,13 +214,13 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
         ev_function_call(call_id, "update_plan", &invalid_args),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "malformed plan payload"),
         ev_completed("resp-2"),
     ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -252,22 +256,12 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
     );
 
     let req = second_mock.single_request();
-    let output_item = req.function_call_output(call_id);
-    assert_eq!(
-        output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let (output_text, success_flag) = call_output(&req, call_id);
     assert!(
         output_text.contains("failed to parse function arguments"),
         "expected parse error message in output text, got {output_text:?}"
     );
-    if let Some(success_flag) = output_item
-        .get("output")
-        .and_then(|value| value.as_object())
-        .and_then(|obj| obj.get("success"))
-        .and_then(serde_json::Value::as_bool)
-    {
+    if let Some(success_flag) = success_flag {
         assert!(
             !success_flag,
             "expected tool output to mark success=false for malformed payload"
@@ -308,13 +302,13 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
         ev_apply_patch_function_call(call_id, &patch_content),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "patch complete"),
         ev_completed("resp-2"),
     ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -357,12 +351,7 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
     assert!(patch_end_success);
 
     let req = second_mock.single_request();
-    let output_item = req.function_call_output(call_id);
-    assert_eq!(
-        output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let (output_text, _success_flag) = call_output(&req, call_id);
 
     let expected_pattern = format!(
         r"(?s)^Exit code: 0
@@ -372,7 +361,7 @@ Success. Updated the following files:
 A {file_name}
 ?$"
     );
-    assert_regex_match(&expected_pattern, output_text);
+    assert_regex_match(&expected_pattern, &output_text);
 
     let updated_contents = fs::read_to_string(file_path)?;
     assert_eq!(
@@ -409,13 +398,13 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
         ev_apply_patch_function_call(call_id, patch_content),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "failed"),
         ev_completed("resp-2"),
     ]);
-    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -437,12 +426,7 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
     let req = second_mock.single_request();
-    let output_item = req.function_call_output(call_id);
-    assert_eq!(
-        output_item.get("call_id").and_then(Value::as_str),
-        Some(call_id)
-    );
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let (output_text, success_flag) = call_output(&req, call_id);
 
     assert!(
         output_text.contains("apply_patch verification failed"),
@@ -453,12 +437,7 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
         "expected parse diagnostics in output text, got {output_text:?}"
     );
 
-    if let Some(success_flag) = output_item
-        .get("output")
-        .and_then(|value| value.as_object())
-        .and_then(|obj| obj.get("success"))
-        .and_then(serde_json::Value::as_bool)
-    {
+    if let Some(success_flag) = success_flag {
         assert!(
             !success_flag,
             "expected tool output to mark success=false for parse failures"

codex-rs/core/tests/suite/tool_parallelism.rs

@@ -14,6 +14,7 @@ use codex_protocol::user_input::UserInput;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
@@ -21,6 +22,7 @@ use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
+use serde_json::Value;
 use serde_json::json;
 
 async fn run_turn(test: &TestCodex, prompt: &str) -> anyhow::Result<()> {
@@ -204,3 +206,81 @@ async fn mixed_tools_fall_back_to_serial() -> anyhow::Result<()> {
 
     Ok(())
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn tool_results_grouped() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let test = build_codex_with_test_tool(&server).await?;
+
+    let shell_args = serde_json::to_string(&json!({
+        "command": ["/bin/sh", "-c", "echo 'shell output'"],
+        "timeout_ms": 1_000,
+    }))?;
+
+    mount_sse_once(
+        &server,
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_function_call("call-1", "shell", &shell_args),
+            ev_function_call("call-2", "shell", &shell_args),
+            ev_function_call("call-3", "shell", &shell_args),
+            ev_completed("resp-1"),
+        ]),
+    )
+    .await;
+    let tool_output_request = mount_sse_once(
+        &server,
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    )
+    .await;
+
+    run_turn(&test, "run shell three times").await?;
+
+    let input = tool_output_request.single_request().input();
+
+    // find all function_call inputs with indexes
+    let function_calls = input
+        .iter()
+        .enumerate()
+        .filter(|(_, item)| item.get("type").and_then(Value::as_str) == Some("function_call"))
+        .collect::<Vec<_>>();
+
+    let function_call_outputs = input
+        .iter()
+        .enumerate()
+        .filter(|(_, item)| {
+            item.get("type").and_then(Value::as_str) == Some("function_call_output")
+        })
+        .collect::<Vec<_>>();
+
+    assert_eq!(function_calls.len(), 3);
+    assert_eq!(function_call_outputs.len(), 3);
+
+    for (index, _) in &function_calls {
+        for (output_index, _) in &function_call_outputs {
+            assert!(
+                *index < *output_index,
+                "all function calls must come before outputs"
+            );
+        }
+    }
+
+    // output should come in the order of the function calls
+    let zipped = function_calls
+        .iter()
+        .zip(function_call_outputs.iter())
+        .collect::<Vec<_>>();
+    for (call, output) in zipped {
+        assert_eq!(
+            call.1.get("call_id").and_then(Value::as_str),
+            output.1.get("call_id").and_then(Value::as_str)
+        );
+    }
+
+    Ok(())
+}

codex-rs/core/tests/suite/tools.rs

@@ -1,16 +1,16 @@
 #![cfg(not(target_os = "windows"))]
 #![allow(clippy::unwrap_used, clippy::expect_used)]
 
+use std::fs;
+use std::time::Duration;
+use std::time::Instant;
+
 use anyhow::Context;
 use anyhow::Result;
 use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
-use codex_core::protocol::EventMsg;
-use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
-use codex_protocol::config_types::ReasoningSummary;
-use codex_protocol::user_input::UserInput;
 use core_test_support::assert_regex_match;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
@@ -22,44 +22,11 @@ use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
-use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
-use core_test_support::wait_for_event;
 use regex_lite::Regex;
 use serde_json::Value;
 use serde_json::json;
 
-async fn submit_turn(
-    test: &TestCodex,
-    prompt: &str,
-    approval_policy: AskForApproval,
-    sandbox_policy: SandboxPolicy,
-) -> Result<()> {
-    let session_model = test.session_configured.model.clone();
-
-    test.codex
-        .submit(Op::UserTurn {
-            items: vec![UserInput::Text {
-                text: prompt.into(),
-            }],
-            final_output_json_schema: None,
-            cwd: test.cwd.path().to_path_buf(),
-            approval_policy,
-            sandbox_policy,
-            model: session_model,
-            effort: None,
-            summary: ReasoningSummary::Auto,
-        })
-        .await?;
-
-    wait_for_event(&test.codex, |event| {
-        matches!(event, EventMsg::TaskComplete(_))
-    })
-    .await;
-
-    Ok(())
-}
-
 fn tool_names(body: &Value) -> Vec<String> {
     body.get("tools")
         .and_then(Value::as_array)
@@ -106,8 +73,7 @@ async fn custom_tool_unknown_returns_custom_output_error() -> Result<()> {
     )
     .await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policies(
         "invoke custom tool",
         AskForApproval::Never,
         SandboxPolicy::DangerFullAccess,
@@ -185,8 +151,7 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
     )
     .await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policies(
         "run the shell command",
         AskForApproval::Never,
         SandboxPolicy::DangerFullAccess,
@@ -198,24 +163,22 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
         "approval policy is {policy:?}; reject command — you should not ask for escalated permissions if the approval policy is {policy:?}"
     );
 
-    let blocked_item = second_mock
+    let blocked_output = second_mock
         .single_request()
-        .function_call_output(call_id_blocked);
+        .function_call_output_content_and_success(call_id_blocked)
+        .and_then(|(content, _)| content)
+        .expect("blocked output string");
     assert_eq!(
-        blocked_item.get("output").and_then(Value::as_str),
-        Some(expected_message.as_str()),
+        blocked_output, expected_message,
         "unexpected rejection message"
     );
 
-    let success_item = third_mock
+    let success_output = third_mock
         .single_request()
-        .function_call_output(call_id_success);
-    let output_json: Value = serde_json::from_str(
-        success_item
-            .get("output")
-            .and_then(Value::as_str)
-            .expect("success output string"),
-    )?;
+        .function_call_output_content_and_success(call_id_success)
+        .and_then(|(content, _)| content)
+        .expect("success output string");
+    let output_json: Value = serde_json::from_str(&success_output)?;
     assert_eq!(
         output_json["metadata"]["exit_code"].as_i64(),
         Some(0),
@@ -344,8 +307,7 @@ async fn collect_tools(use_unified_exec: bool) -> Result<Vec<String>> {
     });
     let test = builder.build(&server).await?;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policies(
         "list tools",
         AskForApproval::Never,
         SandboxPolicy::DangerFullAccess,
@@ -419,8 +381,7 @@ async fn shell_timeout_includes_timeout_prefix_and_metadata() -> Result<()> {
     )
     .await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policies(
         "run a long command",
         AskForApproval::Never,
         SandboxPolicy::DangerFullAccess,
@@ -458,6 +419,101 @@ async fn shell_timeout_includes_timeout_prefix_and_metadata() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn shell_timeout_handles_background_grandchild_stdout() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.model = "gpt-5".to_string();
+        config.model_family = find_family_for_model("gpt-5").expect("gpt-5 is a valid model");
+        config.sandbox_policy = SandboxPolicy::DangerFullAccess;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "shell-grandchild-timeout";
+    let pid_path = test.cwd.path().join("grandchild_pid.txt");
+    let script_path = test.cwd.path().join("spawn_detached.py");
+    let script = format!(
+        r#"import subprocess
+import time
+from pathlib import Path
+
+# Spawn a detached grandchild that inherits stdout/stderr so the pipe stays open.
+proc = subprocess.Popen(["/bin/sh", "-c", "sleep 60"], start_new_session=True)
+Path({pid_path:?}).write_text(str(proc.pid))
+time.sleep(60)
+"#
+    );
+    fs::write(&script_path, script)?;
+
+    let args = json!({
+        "command": ["python3", script_path.to_string_lossy()],
+        "timeout_ms": 200,
+    });
+
+    mount_sse_once(
+        &server,
+        sse(vec![
+            ev_response_created("resp-1"),
+            ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+            ev_completed("resp-1"),
+        ]),
+    )
+    .await;
+    let second_mock = mount_sse_once(
+        &server,
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    )
+    .await;
+
+    let start = Instant::now();
+    let output_str = tokio::time::timeout(Duration::from_secs(10), async {
+        test.submit_turn_with_policies(
+            "run a command with a detached grandchild",
+            AskForApproval::Never,
+            SandboxPolicy::DangerFullAccess,
+        )
+        .await?;
+        let timeout_item = second_mock.single_request().function_call_output(call_id);
+        timeout_item
+            .get("output")
+            .and_then(Value::as_str)
+            .map(str::to_string)
+            .context("timeout output string")
+    })
+    .await
+    .context("exec call should not hang waiting for grandchild pipes to close")??;
+    let elapsed = start.elapsed();
+
+    if let Ok(output_json) = serde_json::from_str::<Value>(&output_str) {
+        assert_eq!(
+            output_json["metadata"]["exit_code"].as_i64(),
+            Some(124),
+            "expected timeout exit code 124",
+        );
+    } else {
+        let timeout_pattern = r"(?is)command timed out|timeout";
+        assert_regex_match(timeout_pattern, &output_str);
+    }
+
+    assert!(
+        elapsed < Duration::from_secs(9),
+        "command should return shortly after timeout even with live grandchildren: {elapsed:?}"
+    );
+
+    if let Ok(pid_str) = fs::read_to_string(&pid_path)
+        && let Ok(pid) = pid_str.trim().parse::<libc::pid_t>()
+    {
+        unsafe { libc::kill(pid, libc::SIGKILL) };
+    }
+
+    Ok(())
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn shell_spawn_failure_truncates_exec_error() -> Result<()> {
     skip_if_no_network!(Ok(()));
@@ -500,8 +556,7 @@ async fn shell_spawn_failure_truncates_exec_error() -> Result<()> {
     )
     .await;
 
-    submit_turn(
-        &test,
+    test.submit_turn_with_policies(
         "spawn a missing binary",
         AskForApproval::Never,
         SandboxPolicy::DangerFullAccess,

codex-rs/core/tests/suite/truncation.rs

@@ -19,7 +19,7 @@ use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
 use core_test_support::responses::ev_response_created;
-use core_test_support::responses::mount_sse_once_match;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
@@ -32,7 +32,6 @@ use serde_json::Value;
 use serde_json::json;
 use std::collections::HashMap;
 use std::time::Duration;
-use wiremock::matchers::any;
 
 // Verifies byte-truncation formatting for function error output (RespondToModel errors)
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -86,7 +85,7 @@ async fn truncate_function_error_trims_respond_to_model() -> Result<()> {
         serde_json::from_str::<serde_json::Value>(&output).is_err(),
         "expected error output to be plain text",
     );
-    let truncated_pattern = r#"(?s)^Total output lines: 1\s+.*\[\.\.\. output truncated to fit 10240 bytes \.\.\.\]\s*$"#;
+    let truncated_pattern = r#"(?s)^Total output lines: 1\s+.*\[\.\.\. output truncated to fit 11264 bytes \.\.\.\]\s*$"#;
     assert_regex_match(truncated_pattern, &output);
     assert!(
         !output.contains("omitted"),
@@ -113,15 +112,25 @@ async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
     let fixture = builder.build(&server).await?;
 
     let call_id = "shell-too-large";
-    let args = serde_json::json!({
-        "command": ["/bin/sh", "-c", "seq 1 400"],
-        "timeout_ms": 5_000,
-    });
+    let args = if cfg!(windows) {
+        serde_json::json!({
+            "command": [
+                "powershell",
+                "-Command",
+                "for ($i=1; $i -le 400; $i++) { Write-Output $i }"
+            ],
+            "timeout_ms": 5_000,
+        })
+    } else {
+        serde_json::json!({
+            "command": ["/bin/sh", "-c", "seq 1 400"],
+            "timeout_ms": 5_000,
+        })
+    };
 
     // First response: model tells us to run the tool; second: complete the turn.
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
@@ -129,9 +138,8 @@ async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
         ]),
     )
     .await;
-    let mock2 = mount_sse_once_match(
+    let mock2 = mount_sse_once(
         &server,
-        any(),
         sse(vec![
             responses::ev_assistant_message("msg-1", "done"),
             responses::ev_completed("resp-2"),
@@ -149,6 +157,7 @@ async fn tool_call_output_exceeds_limit_truncated_for_model() -> Result<()> {
         .single_request()
         .function_call_output_text(call_id)
         .context("function_call_output present for shell call")?;
+    let output = output.replace("\r\n", "\n");
 
     // Expect plain text (not JSON) with truncation markers and line elision.
     assert!(
@@ -180,6 +189,73 @@ $"#;
     Ok(())
 }
 
+// Ensures shell tool outputs that exceed the line limit are truncated only once.
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn tool_call_output_truncated_only_once() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.model = "gpt-5-codex".to_string();
+        config.model_family =
+            find_family_for_model("gpt-5-codex").expect("gpt-5-codex is a model family");
+    });
+    let fixture = builder.build(&server).await?;
+    let call_id = "shell-single-truncation";
+    let args = if cfg!(windows) {
+        serde_json::json!({
+            "command": [
+                "powershell",
+                "-Command",
+                "for ($i=1; $i -le 2000; $i++) { Write-Output $i }"
+            ],
+            "timeout_ms": 5_000,
+        })
+    } else {
+        serde_json::json!({
+            "command": ["/bin/sh", "-c", "seq 1 2000"],
+            "timeout_ms": 5_000,
+        })
+    };
+
+    mount_sse_once(
+        &server,
+        sse(vec![
+            responses::ev_response_created("resp-1"),
+            responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+            responses::ev_completed("resp-1"),
+        ]),
+    )
+    .await;
+    let mock2 = mount_sse_once(
+        &server,
+        sse(vec![
+            responses::ev_assistant_message("msg-1", "done"),
+            responses::ev_completed("resp-2"),
+        ]),
+    )
+    .await;
+
+    fixture
+        .submit_turn_with_policy("trigger big shell output", SandboxPolicy::DangerFullAccess)
+        .await?;
+
+    let output = mock2
+        .single_request()
+        .function_call_output_text(call_id)
+        .context("function_call_output present for shell call")?;
+
+    let total_line_headers = output.matches("Total output lines:").count();
+
+    assert_eq!(
+        total_line_headers, 1,
+        "shell output should carry only one truncation header: {output}"
+    );
+
+    Ok(())
+}
+
 // Verifies that an MCP tool call result exceeding the model formatting limits
 // is truncated before being sent back to the model.
 #[tokio::test(flavor = "multi_thread", worker_threads = 1)]
@@ -196,9 +272,8 @@ async fn mcp_tool_call_output_exceeds_limit_truncated_for_model() -> Result<()>
     let large_msg = "long-message-with-newlines-".repeat(600);
     let args_json = serde_json::json!({ "message": large_msg });
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         sse(vec![
             responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, &args_json.to_string()),
@@ -206,9 +281,8 @@ async fn mcp_tool_call_output_exceeds_limit_truncated_for_model() -> Result<()>
         ]),
     )
     .await;
-    let mock2 = mount_sse_once_match(
+    let mock2 = mount_sse_once(
         &server,
-        any(),
         sse(vec![
             responses::ev_assistant_message("msg-1", "rmcp echo tool completed."),
             responses::ev_completed("resp-2"),
@@ -269,7 +343,8 @@ async fn mcp_tool_call_output_exceeds_limit_truncated_for_model() -> Result<()>
         output.starts_with("Total output lines: 1\n\n{"),
         "expected total line header and JSON head, got: {output}"
     );
-    let byte_marker = Regex::new(r"\[\.\.\. output truncated to fit 10240 bytes \.\.\.\]")
+
+    let byte_marker = Regex::new(r"\[\.\.\. output truncated to fit 11264 bytes \.\.\.\]")
         .expect("compile regex");
     assert!(
         byte_marker.is_match(&output),
@@ -291,9 +366,8 @@ async fn mcp_image_output_preserves_image_and_no_text_summary() -> Result<()> {
     let server_name = "rmcp";
     let tool_name = format!("mcp__{server_name}__image");
 
-    mount_sse_once_match(
+    mount_sse_once(
         &server,
-        any(),
         sse(vec![
             ev_response_created("resp-1"),
             ev_function_call(call_id, &tool_name, "{}"),
@@ -301,9 +375,8 @@ async fn mcp_image_output_preserves_image_and_no_text_summary() -> Result<()> {
         ]),
     )
     .await;
-    let final_mock = mount_sse_once_match(
+    let final_mock = mount_sse_once(
         &server,
-        any(),
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-2"),

codex-rs/core/tests/suite/unified_exec.rs

@@ -7,6 +7,7 @@ use anyhow::Result;
 use codex_core::features::Feature;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
+use codex_core::protocol::ExecCommandSource;
 use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
 use codex_protocol::config_types::ReasoningSummary;
@@ -214,7 +215,11 @@ async fn unified_exec_emits_exec_command_begin_event() -> Result<()> {
 
     assert_eq!(
         begin_event.command,
-        vec!["/bin/echo hello unified exec".to_string()]
+        vec![
+            "/bin/bash".to_string(),
+            "-lc".to_string(),
+            "/bin/echo hello unified exec".to_string()
+        ]
     );
     assert_eq!(begin_event.cwd, cwd.path());
 
@@ -224,6 +229,7 @@ async fn unified_exec_emits_exec_command_begin_event() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[ignore = "flaky"]
 async fn unified_exec_respects_workdir_override() -> Result<()> {
     skip_if_no_network!(Ok(()));
     skip_if_sandbox!(Ok(()));
@@ -564,7 +570,109 @@ async fn unified_exec_emits_output_delta_for_write_stdin() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn unified_exec_skips_begin_event_for_empty_input() -> Result<()> {
+async fn unified_exec_emits_begin_for_write_stdin() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+    skip_if_sandbox!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.use_experimental_unified_exec_tool = true;
+        config.features.enable(Feature::UnifiedExec);
+    });
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = builder.build(&server).await?;
+
+    let open_call_id = "uexec-open-for-begin";
+    let open_args = json!({
+        "cmd": "/bin/sh -c echo ready".to_string(),
+        "yield_time_ms": 200,
+    });
+
+    let stdin_call_id = "uexec-stdin-begin";
+    let stdin_args = json!({
+        "chars": "echo hello",
+        "session_id": 0,
+        "yield_time_ms": 400,
+    });
+
+    let responses = vec![
+        sse(vec![
+            ev_response_created("resp-1"),
+            ev_function_call(
+                open_call_id,
+                "exec_command",
+                &serde_json::to_string(&open_args)?,
+            ),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_response_created("resp-2"),
+            ev_function_call(
+                stdin_call_id,
+                "write_stdin",
+                &serde_json::to_string(&stdin_args)?,
+            ),
+            ev_completed("resp-2"),
+        ]),
+        sse(vec![
+            ev_response_created("resp-3"),
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-3"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![UserInput::Text {
+                text: "begin events for stdin".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    let begin_event = wait_for_event_match(&codex, |msg| match msg {
+        EventMsg::ExecCommandBegin(ev) if ev.call_id == stdin_call_id => Some(ev.clone()),
+        _ => None,
+    })
+    .await;
+
+    assert_eq!(
+        begin_event.command,
+        vec![
+            "/bin/bash".to_string(),
+            "-lc".to_string(),
+            "/bin/sh -c echo ready".to_string()
+        ]
+    );
+    assert_eq!(
+        begin_event.interaction_input,
+        Some("echo hello".to_string())
+    );
+    assert_eq!(
+        begin_event.source,
+        ExecCommandSource::UnifiedExecInteraction
+    );
+
+    wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn unified_exec_emits_begin_event_for_write_stdin_requests() -> Result<()> {
     skip_if_no_network!(Ok(()));
     skip_if_sandbox!(Ok(()));
 
@@ -589,9 +697,9 @@ async fn unified_exec_skips_begin_event_for_empty_input() -> Result<()> {
 
     let poll_call_id = "uexec-poll-empty";
     let poll_args = json!({
-        "input": Vec::<String>::new(),
-        "session_id": "0",
-        "timeout_ms": 150,
+        "chars": "",
+        "session_id": 0,
+        "yield_time_ms": 150,
     });
 
     let responses = vec![
@@ -650,11 +758,45 @@ async fn unified_exec_skips_begin_event_for_empty_input() -> Result<()> {
 
     assert_eq!(
         begin_events.len(),
-        1,
-        "expected only the initial command to emit begin event"
+        2,
+        "expected begin events for the startup command and the write_stdin call"
     );
-    assert_eq!(begin_events[0].call_id, open_call_id);
-    assert_eq!(begin_events[0].command[0], "/bin/sh -c echo ready");
+
+    let open_event = begin_events
+        .iter()
+        .find(|ev| ev.call_id == open_call_id)
+        .expect("missing exec_command begin");
+    assert_eq!(
+        open_event.command,
+        vec![
+            "/bin/bash".to_string(),
+            "-lc".to_string(),
+            "/bin/sh -c echo ready".to_string()
+        ]
+    );
+    assert!(
+        open_event.interaction_input.is_none(),
+        "startup begin events should not include interaction input"
+    );
+    assert_eq!(open_event.source, ExecCommandSource::UnifiedExecStartup);
+
+    let poll_event = begin_events
+        .iter()
+        .find(|ev| ev.call_id == poll_call_id)
+        .expect("missing write_stdin begin");
+    assert_eq!(
+        poll_event.command,
+        vec![
+            "/bin/bash".to_string(),
+            "-lc".to_string(),
+            "/bin/sh -c echo ready".to_string()
+        ]
+    );
+    assert!(
+        poll_event.interaction_input.is_none(),
+        "poll begin events should omit interaction input"
+    );
+    assert_eq!(poll_event.source, ExecCommandSource::UnifiedExecInteraction);
 
     Ok(())
 }

codex-rs/core/tests/suite/user_notification.rs

@@ -15,7 +15,6 @@ use pretty_assertions::assert_eq;
 use serde_json::Value;
 use serde_json::json;
 use tempfile::TempDir;
-use wiremock::matchers::any;
 
 use responses::ev_assistant_message;
 use responses::ev_completed;
@@ -35,7 +34,7 @@ async fn summarize_context_three_requests_and_instructions() -> anyhow::Result<(
 
     let sse1 = sse(vec![ev_assistant_message("m1", "Done"), ev_completed("r1")]);
 
-    responses::mount_sse_once_match(&server, any(), sse1).await;
+    responses::mount_sse_once(&server, sse1).await;
 
     let notify_dir = TempDir::new()?;
     // write a script to the notify that touches a file next to it

codex-rs/core/tests/suite/user_shell_cmd.rs

@@ -1,13 +1,26 @@
+use anyhow::Context;
 use codex_core::ConversationManager;
 use codex_core::NewConversation;
+use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::ExecCommandEndEvent;
+use codex_core::protocol::ExecCommandSource;
 use codex_core::protocol::ExecOutputStream;
 use codex_core::protocol::Op;
+use codex_core::protocol::SandboxPolicy;
 use codex_core::protocol::TurnAbortReason;
 use core_test_support::assert_regex_match;
 use core_test_support::load_default_config_for_test;
 use core_test_support::responses;
+use core_test_support::responses::ev_assistant_message;
+use core_test_support::responses::ev_completed;
+use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
+use core_test_support::responses::mount_sse_once;
+use core_test_support::responses::sse;
+use core_test_support::responses::start_mock_server;
+use core_test_support::skip_if_no_network;
+use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
 use core_test_support::wait_for_event_match;
 use regex_lite::escape;
@@ -138,7 +151,7 @@ async fn user_shell_command_history_is_persisted_and_shared_with_model() -> anyh
         _ => None,
     })
     .await;
-    assert!(begin_event.is_user_shell_command);
+    assert_eq!(begin_event.source, ExecCommandSource::UserShell);
     let matches_last_arg = begin_event.command.last() == Some(&command);
     let matches_split = shlex::split(&command).is_some_and(|split| split == begin_event.command);
     assert!(
@@ -249,3 +262,70 @@ async fn user_shell_command_output_is_truncated_in_history() -> anyhow::Result<(
 
     Ok(())
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn user_shell_command_is_truncated_only_once() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.model = "gpt-5-codex".to_string();
+        config.model_family =
+            find_family_for_model("gpt-5-codex").expect("gpt-5-codex is a model family");
+    });
+    let fixture = builder.build(&server).await?;
+
+    let call_id = "user-shell-double-truncation";
+    let args = if cfg!(windows) {
+        serde_json::json!({
+            "command": [
+                "powershell",
+                "-Command",
+                "for ($i=1; $i -le 2000; $i++) { Write-Output $i }"
+            ],
+            "timeout_ms": 5_000,
+        })
+    } else {
+        serde_json::json!({
+            "command": ["/bin/sh", "-c", "seq 1 2000"],
+            "timeout_ms": 5_000,
+        })
+    };
+
+    mount_sse_once(
+        &server,
+        sse(vec![
+            ev_response_created("resp-1"),
+            ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+            ev_completed("resp-1"),
+        ]),
+    )
+    .await;
+    let mock2 = mount_sse_once(
+        &server,
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    )
+    .await;
+
+    fixture
+        .submit_turn_with_policy("trigger big shell output", SandboxPolicy::DangerFullAccess)
+        .await?;
+
+    let output = mock2
+        .single_request()
+        .function_call_output_text(call_id)
+        .context("function_call_output present for shell call")?;
+
+    let truncation_headers = output.matches("Total output lines:").count();
+
+    assert_eq!(
+        truncation_headers, 1,
+        "shell output should carry only one truncation header: {output}"
+    );
+
+    Ok(())
+}

codex-rs/core/tests/suite/view_image.rs

@@ -24,7 +24,6 @@ use image::ImageBuffer;
 use image::Rgba;
 use image::load_from_memory;
 use serde_json::Value;
-use wiremock::matchers::any;
 
 fn find_image_message(body: &Value) -> Option<&Value> {
     body.get("input")
@@ -45,14 +44,6 @@ fn find_image_message(body: &Value) -> Option<&Value> {
         })
 }
 
-fn extract_output_text(item: &Value) -> Option<&str> {
-    item.get("output").and_then(|value| match value {
-        Value::String(text) => Some(text.as_str()),
-        Value::Object(obj) => obj.get("content").and_then(Value::as_str),
-        _ => None,
-    })
-}
-
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn user_turn_with_local_image_attaches_image() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
@@ -79,7 +70,7 @@ async fn user_turn_with_local_image_attaches_image() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-1"),
     ]);
-    let mock = responses::mount_sse_once_match(&server, any(), response).await;
+    let mock = responses::mount_sse_once(&server, response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -164,13 +155,13 @@ async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -207,10 +198,12 @@ async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
     assert_eq!(tool_event.call_id, call_id);
     assert_eq!(tool_event.path, abs_path);
 
-    let body = mock.single_request().body_json();
-    let output_item = mock.single_request().function_call_output(call_id);
-
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let req = mock.single_request();
+    let body = req.body_json();
+    let output_text = req
+        .function_call_output_content_and_success(call_id)
+        .and_then(|(content, _)| content)
+        .expect("output text present");
     assert_eq!(output_text, "attached local image path");
 
     let image_message =
@@ -272,13 +265,13 @@ async fn view_image_tool_errors_when_path_is_directory() -> anyhow::Result<()> {
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -299,9 +292,12 @@ async fn view_image_tool_errors_when_path_is_directory() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let body_with_tool_output = mock.single_request().body_json();
-    let output_item = mock.single_request().function_call_output(call_id);
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let req = mock.single_request();
+    let body_with_tool_output = req.body_json();
+    let output_text = req
+        .function_call_output_content_and_success(call_id)
+        .and_then(|(content, _)| content)
+        .expect("output text present");
     let expected_message = format!("image path `{}` is not a file", abs_path.display());
     assert_eq!(output_text, expected_message);
 
@@ -341,13 +337,13 @@ async fn view_image_tool_placeholder_for_non_image_files() -> anyhow::Result<()>
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -398,8 +394,11 @@ async fn view_image_tool_placeholder_for_non_image_files() -> anyhow::Result<()>
         "placeholder should mention path: {placeholder}"
     );
 
-    let output_item = mock.single_request().function_call_output(call_id);
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let output_text = mock
+        .single_request()
+        .function_call_output_content_and_success(call_id)
+        .and_then(|(content, _)| content)
+        .expect("output text present");
     assert_eq!(output_text, "attached local image path");
 
     Ok(())
@@ -429,13 +428,13 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), first_response).await;
+    responses::mount_sse_once(&server, first_response).await;
 
     let second_response = sse(vec![
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once(&server, second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -456,9 +455,12 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let body_with_tool_output = mock.single_request().body_json();
-    let output_item = mock.single_request().function_call_output(call_id);
-    let output_text = extract_output_text(&output_item).expect("output text present");
+    let req = mock.single_request();
+    let body_with_tool_output = req.body_json();
+    let output_text = req
+        .function_call_output_content_and_success(call_id)
+        .and_then(|(content, _)| content)
+        .expect("output text present");
     let expected_prefix = format!("unable to locate image at `{}`:", abs_path.display());
     assert!(
         output_text.starts_with(&expected_prefix),

codex-rs/exec/src/cli.rs

@@ -52,6 +52,10 @@ pub struct Cli {
     #[arg(long = "skip-git-repo-check", default_value_t = false)]
     pub skip_git_repo_check: bool,
 
+    /// Additional directories that should be writable alongside the primary workspace.
+    #[arg(long = "add-dir", value_name = "DIR", value_hint = clap::ValueHint::DirPath)]
+    pub add_dir: Vec<PathBuf>,
+
     /// Path to a JSON Schema file describing the model's final response shape.
     #[arg(long = "output-schema", value_name = "FILE")]
     pub output_schema: Option<PathBuf>,

codex-rs/exec/src/lib.rs

@@ -62,6 +62,7 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
         dangerously_bypass_approvals_and_sandbox,
         cwd,
         skip_git_repo_check,
+        add_dir,
         color,
         last_message_file,
         json: json_mode,
@@ -180,7 +181,7 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
         show_raw_agent_reasoning: oss.then_some(true),
         tools_web_search_request: None,
         experimental_sandbox_command_assessment: None,
-        additional_writable_roots: Vec::new(),
+        additional_writable_roots: add_dir,
     };
     // Parse `-c` overrides.
     let cli_kv_overrides = match config_overrides.parse_overrides() {

codex-rs/exec/tests/event_processor_with_json_output.rs

@@ -5,6 +5,7 @@ use codex_core::protocol::Event;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::ExecCommandBeginEvent;
 use codex_core::protocol::ExecCommandEndEvent;
+use codex_core::protocol::ExecCommandSource;
 use codex_core::protocol::FileChange;
 use codex_core::protocol::McpInvocation;
 use codex_core::protocol::McpToolCallBeginEvent;
@@ -626,7 +627,8 @@ fn exec_command_end_success_produces_completed_command_item() {
             command: vec!["bash".to_string(), "-lc".to_string(), "echo hi".to_string()],
             cwd: std::env::current_dir().unwrap(),
             parsed_cmd: Vec::new(),
-            is_user_shell_command: false,
+            source: ExecCommandSource::Agent,
+            interaction_input: None,
         }),
     );
     let out_begin = ep.collect_thread_events(&begin);
@@ -687,7 +689,8 @@ fn exec_command_end_failure_produces_failed_command_item() {
             command: vec!["sh".to_string(), "-c".to_string(), "exit 1".to_string()],
             cwd: std::env::current_dir().unwrap(),
             parsed_cmd: Vec::new(),
-            is_user_shell_command: false,
+            source: ExecCommandSource::Agent,
+            interaction_input: None,
         }),
     );
     assert_eq!(

codex-rs/exec/tests/suite/add_dir.rs

@@ -0,0 +1,72 @@
+#![cfg(not(target_os = "windows"))]
+#![allow(clippy::expect_used, clippy::unwrap_used)]
+
+use core_test_support::responses;
+use core_test_support::test_codex_exec::test_codex_exec;
+
+/// Verify that the --add-dir flag is accepted and the command runs successfully.
+/// This test confirms the CLI argument is properly wired up.
+#[tokio::test(flavor = "multi_thread", worker_threads = 1)]
+async fn accepts_add_dir_flag() -> anyhow::Result<()> {
+    let test = test_codex_exec();
+
+    let server = responses::start_mock_server().await;
+    let body = responses::sse(vec![
+        responses::ev_response_created("response_1"),
+        responses::ev_assistant_message("response_1", "Task completed"),
+        responses::ev_completed("response_1"),
+    ]);
+    responses::mount_sse_once(&server, body).await;
+
+    // Create temporary directories to use with --add-dir
+    let temp_dir1 = tempfile::tempdir()?;
+    let temp_dir2 = tempfile::tempdir()?;
+
+    test.cmd_with_server(&server)
+        .arg("--skip-git-repo-check")
+        .arg("--sandbox")
+        .arg("workspace-write")
+        .arg("--add-dir")
+        .arg(temp_dir1.path())
+        .arg("--add-dir")
+        .arg(temp_dir2.path())
+        .arg("test with additional directories")
+        .assert()
+        .code(0);
+
+    Ok(())
+}
+
+/// Verify that multiple --add-dir flags can be specified.
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn accepts_multiple_add_dir_flags() -> anyhow::Result<()> {
+    let test = test_codex_exec();
+
+    let server = responses::start_mock_server().await;
+    let body = responses::sse(vec![
+        responses::ev_response_created("response_1"),
+        responses::ev_assistant_message("response_1", "Multiple directories accepted"),
+        responses::ev_completed("response_1"),
+    ]);
+    responses::mount_sse_once(&server, body).await;
+
+    let temp_dir1 = tempfile::tempdir()?;
+    let temp_dir2 = tempfile::tempdir()?;
+    let temp_dir3 = tempfile::tempdir()?;
+
+    test.cmd_with_server(&server)
+        .arg("--skip-git-repo-check")
+        .arg("--sandbox")
+        .arg("workspace-write")
+        .arg("--add-dir")
+        .arg(temp_dir1.path())
+        .arg("--add-dir")
+        .arg(temp_dir2.path())
+        .arg("--add-dir")
+        .arg(temp_dir3.path())
+        .arg("test with three directories")
+        .assert()
+        .code(0);
+
+    Ok(())
+}

codex-rs/exec/tests/suite/mod.rs

@@ -1,4 +1,5 @@
 // Aggregates all former standalone integration tests as modules.
+mod add_dir;
 mod apply_patch;
 mod auth_env;
 mod originator;

codex-rs/exec/tests/suite/output_schema.rs

@@ -4,7 +4,6 @@
 use core_test_support::responses;
 use core_test_support::test_codex_exec::test_codex_exec;
 use serde_json::Value;
-use wiremock::matchers::any;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn exec_includes_output_schema_in_request() -> anyhow::Result<()> {
@@ -28,7 +27,7 @@ async fn exec_includes_output_schema_in_request() -> anyhow::Result<()> {
         responses::ev_assistant_message("m1", "fixture hello"),
         responses::ev_completed("resp1"),
     ]);
-    let response_mock = responses::mount_sse_once_match(&server, any(), body).await;
+    let response_mock = responses::mount_sse_once(&server, body).await;
 
     test.cmd_with_server(&server)
         .arg("--skip-git-repo-check")

codex-rs/exec/tests/suite/server_error_exit.rs

@@ -3,7 +3,6 @@
 
 use core_test_support::responses;
 use core_test_support::test_codex_exec::test_codex_exec;
-use wiremock::matchers::any;
 
 /// Verify that when the server reports an error, `codex-exec` exits with a
 /// non-zero status code so automation can detect failures.
@@ -21,7 +20,7 @@ async fn exits_non_zero_when_server_reports_error() -> anyhow::Result<()> {
             "error": {"code": "rate_limit_exceeded", "message": "synthetic server error"}
         }
     })]);
-    responses::mount_sse_once_match(&server, any(), body).await;
+    responses::mount_sse_once(&server, body).await;
 
     test.cmd_with_server(&server)
         .arg("--skip-git-repo-check")

codex-rs/justfile

@@ -21,6 +21,11 @@ tui *args:
 file-search *args:
     cargo run --bin codex-file-search -- "$@"
 
+# Build the CLI and run the app-server test client
+app-server-test-client *args:
+    cargo build -p codex-cli
+    cargo run -p codex-app-server-test-client -- --codex-bin ./target/debug/codex "$@"
+
 # format code
 fmt:
     cargo fmt -- --config imports_granularity=Item

codex-rs/protocol/src/config_types.rs

@@ -19,10 +19,12 @@ use ts_rs::TS;
     JsonSchema,
     TS,
     EnumIter,
+    Hash,
 )]
 #[serde(rename_all = "lowercase")]
 #[strum(serialize_all = "lowercase")]
 pub enum ReasoningEffort {
+    None,
     Minimal,
     Low,
     #[default]
@@ -50,7 +52,18 @@ pub enum ReasoningSummary {
 /// Controls output length/detail on GPT-5 models via the Responses API.
 /// Serialized with lowercase values to match the OpenAI API.
 #[derive(
-    Debug, Serialize, Deserialize, Default, Clone, Copy, PartialEq, Eq, Display, JsonSchema, TS,
+    Hash,
+    Debug,
+    Serialize,
+    Deserialize,
+    Default,
+    Clone,
+    Copy,
+    PartialEq,
+    Eq,
+    Display,
+    JsonSchema,
+    TS,
 )]
 #[serde(rename_all = "lowercase")]
 #[strum(serialize_all = "lowercase")]
@@ -85,3 +98,13 @@ pub enum ForcedLoginMethod {
     Chatgpt,
     Api,
 }
+
+/// Represents the trust level for a project directory.
+/// This determines the approval policy and sandbox mode applied.
+#[derive(Debug, Serialize, Deserialize, Clone, Copy, PartialEq, Eq, Display, JsonSchema, TS)]
+#[serde(rename_all = "lowercase")]
+#[strum(serialize_all = "lowercase")]
+pub enum TrustLevel {
+    Trusted,
+    Untrusted,
+}

codex-rs/protocol/src/items.rs

@@ -11,6 +11,8 @@ use serde::Serialize;
 use ts_rs::TS;
 
 #[derive(Debug, Clone, Deserialize, Serialize, TS, JsonSchema)]
+#[serde(tag = "type")]
+#[ts(tag = "type")]
 pub enum TurnItem {
     UserMessage(UserMessageItem),
     AgentMessage(AgentMessageItem),
@@ -25,6 +27,8 @@ pub struct UserMessageItem {
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize, TS, JsonSchema)]
+#[serde(tag = "type")]
+#[ts(tag = "type")]
 pub enum AgentMessageContent {
     Text { text: String },
 }

codex-rs/protocol/src/protocol.rs

@@ -241,7 +241,7 @@ pub enum AskForApproval {
 /// Determines execution restrictions for model shell commands.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Display, JsonSchema, TS)]
 #[strum(serialize_all = "kebab-case")]
-#[serde(tag = "mode", rename_all = "kebab-case")]
+#[serde(tag = "type", rename_all = "kebab-case")]
 pub enum SandboxPolicy {
     /// No restrictions whatsoever. Use with caution.
     #[serde(rename = "danger-full-access")]
@@ -432,6 +432,7 @@ pub struct Event {
 /// NOTE: Make sure none of these values have optional types, as it will mess up the extension code-gen.
 #[derive(Debug, Clone, Deserialize, Serialize, Display, JsonSchema, TS)]
 #[serde(tag = "type", rename_all = "snake_case")]
+#[ts(tag = "type")]
 #[strum(serialize_all = "snake_case")]
 pub enum EventMsg {
     /// Error while executing a submission
@@ -617,6 +618,9 @@ pub struct ReasoningContentDeltaEvent {
     pub turn_id: String,
     pub item_id: String,
     pub delta: String,
+    // load with default value so it's backward compatible with the old format.
+    #[serde(default)]
+    pub summary_index: i64,
 }
 
 impl HasLegacyEvent for ReasoningContentDeltaEvent {
@@ -633,6 +637,9 @@ pub struct ReasoningRawContentDeltaEvent {
     pub turn_id: String,
     pub item_id: String,
     pub delta: String,
+    // load with default value so it's backward compatible with the old format.
+    #[serde(default)]
+    pub content_index: i64,
 }
 
 impl HasLegacyEvent for ReasoningRawContentDeltaEvent {
@@ -923,7 +930,13 @@ pub struct AgentReasoningRawContentDeltaEvent {
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
-pub struct AgentReasoningSectionBreakEvent {}
+pub struct AgentReasoningSectionBreakEvent {
+    // load with default value so it's backward compatible with the old format.
+    #[serde(default)]
+    pub item_id: String,
+    #[serde(default)]
+    pub summary_index: i64,
+}
 
 #[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
 pub struct AgentReasoningDeltaEvent {
@@ -1203,6 +1216,21 @@ pub struct ReviewLineRange {
     pub end: u32,
 }
 
+#[derive(Debug, Clone, Copy, Deserialize, Serialize, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "snake_case")]
+pub enum ExecCommandSource {
+    Agent,
+    UserShell,
+    UnifiedExecStartup,
+    UnifiedExecInteraction,
+}
+
+impl Default for ExecCommandSource {
+    fn default() -> Self {
+        Self::Agent
+    }
+}
+
 #[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
 pub struct ExecCommandBeginEvent {
     /// Identifier so this can be paired with the ExecCommandEnd event.
@@ -1212,10 +1240,13 @@ pub struct ExecCommandBeginEvent {
     /// The command's working directory if not the default cwd for the agent.
     pub cwd: PathBuf,
     pub parsed_cmd: Vec<ParsedCommand>,
-    /// True when this exec was initiated directly by the user (e.g. bang command),
-    /// not by the agent/model. Defaults to false for backwards compatibility.
+    /// Where the command originated. Defaults to Agent for backward compatibility.
     #[serde(default)]
-    pub is_user_shell_command: bool,
+    pub source: ExecCommandSource,
+    /// Raw input sent to a unified exec session (if this is an interaction event).
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    #[ts(optional)]
+    pub interaction_input: Option<String>,
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize, JsonSchema, TS)]
@@ -1431,7 +1462,8 @@ pub enum ReviewDecision {
 }
 
 #[derive(Debug, Clone, Deserialize, Serialize, PartialEq, JsonSchema, TS)]
-#[serde(rename_all = "snake_case")]
+#[serde(tag = "type", rename_all = "snake_case")]
+#[ts(tag = "type")]
 pub enum FileChange {
     Add {
         content: String,