mirror of
https://github.com/openai/codex.git
synced 2026-05-28 15:00:16 +00:00
Add user_input_requested_during_turn to MCP turn metadata (#22237)
## Why - Similar change as https://github.com/openai/codex/pull/21219 - Without change: MCP tool calls receive `_meta["x-codex-turn-metadata"]` with various key values. - Issue: MCP servers currently do not know if user input was requested during the turn (Ex: Model decides to prompt the user for approval mid-turn before making a possibly risky tool call). MCP servers may want to know this when tracking latency metrics because these instances are inflated. ## What Changed - With change: MCP turn metadata now includes `user_input_requested_during_turn` when a model-visible `request_user_input` call happened earlier in the turn, propagated in `_meta["x-codex-turn-metadata"]`. - `mark_turn_user_input_requested()` is called when user input is requested through either MCP elicitation (`mcp.rs`) or the `request_user_input` tool (`mod.rs`). - MCP tool call `_meta` is now built immediately before execution (`mcp_tool_call.rs`) so user input requested earlier in the same turn, including within the same tool call via elicitation, is reflected in the metadata. - Normal `/responses` turn metadata headers are unchanged. ## Verification - `codex-rs/core/src/session/mcp_tests.rs` - `codex-rs/core/src/tools/handlers/request_user_input_tests.rs` - `codex-rs/core/src/turn_metadata_tests.rs` - `codex-rs/core/tests/suite/search_tool.rs`
This commit is contained in:
@@ -192,12 +192,6 @@ pub(crate) async fn handle_mcp_tool_call(
|
||||
.unwrap_or_else(|| JsonValue::Object(serde_json::Map::new())),
|
||||
};
|
||||
}
|
||||
let request_meta = build_mcp_tool_call_request_meta(
|
||||
turn_context.as_ref(),
|
||||
&server,
|
||||
&call_id,
|
||||
metadata.as_ref(),
|
||||
);
|
||||
let connector_id = metadata
|
||||
.as_ref()
|
||||
.and_then(|metadata| metadata.connector_id.clone());
|
||||
@@ -235,7 +229,6 @@ pub(crate) async fn handle_mcp_tool_call(
|
||||
&call_id,
|
||||
invocation,
|
||||
metadata.as_ref(),
|
||||
request_meta,
|
||||
mcp_app_resource_uri,
|
||||
)
|
||||
.await;
|
||||
@@ -303,7 +296,6 @@ pub(crate) async fn handle_mcp_tool_call(
|
||||
&call_id,
|
||||
invocation,
|
||||
metadata.as_ref(),
|
||||
request_meta,
|
||||
mcp_app_resource_uri,
|
||||
)
|
||||
.await
|
||||
@@ -320,7 +312,6 @@ async fn handle_approved_mcp_tool_call(
|
||||
call_id: &str,
|
||||
invocation: McpInvocation,
|
||||
metadata: Option<&McpToolApprovalMetadata>,
|
||||
request_meta: Option<JsonValue>,
|
||||
mcp_app_resource_uri: Option<String>,
|
||||
) -> HandledMcpToolCall {
|
||||
let server = invocation.server.clone();
|
||||
@@ -353,6 +344,8 @@ async fn handle_approved_mcp_tool_call(
|
||||
};
|
||||
let result = async {
|
||||
let rewritten_arguments = rewrite?;
|
||||
let request_meta =
|
||||
build_mcp_tool_call_request_meta(turn_context, &server, call_id, metadata);
|
||||
let result = execute_mcp_tool_call(
|
||||
sess,
|
||||
turn_context,
|
||||
|
||||
@@ -154,6 +154,9 @@ impl Session {
|
||||
id,
|
||||
request,
|
||||
});
|
||||
turn_context
|
||||
.turn_metadata_state
|
||||
.mark_user_input_requested_during_turn();
|
||||
self.send_event(turn_context, event).await;
|
||||
rx_response.await.ok()
|
||||
}
|
||||
|
||||
@@ -2289,6 +2289,9 @@ impl Session {
|
||||
turn_id: turn_context.sub_id.clone(),
|
||||
questions: args.questions,
|
||||
});
|
||||
turn_context
|
||||
.turn_metadata_state
|
||||
.mark_user_input_requested_during_turn();
|
||||
self.send_event(turn_context, event).await;
|
||||
rx_response.await.ok()
|
||||
}
|
||||
|
||||
@@ -3,6 +3,8 @@ use std::collections::HashMap;
|
||||
use std::sync::Arc;
|
||||
use std::sync::Mutex;
|
||||
use std::sync::RwLock;
|
||||
use std::sync::atomic::AtomicBool;
|
||||
use std::sync::atomic::Ordering;
|
||||
|
||||
use codex_utils_string::to_ascii_json_string;
|
||||
use serde::Serialize;
|
||||
@@ -23,6 +25,7 @@ use codex_utils_absolute_path::AbsolutePathBuf;
|
||||
const MODEL_KEY: &str = "model";
|
||||
const REASONING_EFFORT_KEY: &str = "reasoning_effort";
|
||||
const TURN_STARTED_AT_UNIX_MS_KEY: &str = "turn_started_at_unix_ms";
|
||||
const USER_INPUT_REQUESTED_DURING_TURN_KEY: &str = "user_input_requested_during_turn";
|
||||
|
||||
pub(crate) struct McpTurnMetadataContext<'a> {
|
||||
pub(crate) model: &'a str,
|
||||
@@ -186,6 +189,7 @@ pub(crate) struct TurnMetadataState {
|
||||
enriched_header: Arc<RwLock<Option<String>>>,
|
||||
turn_started_at_unix_ms: Arc<RwLock<Option<i64>>>,
|
||||
responsesapi_client_metadata: Arc<RwLock<Option<HashMap<String, String>>>>,
|
||||
user_input_requested_during_turn: Arc<AtomicBool>,
|
||||
enrichment_task: Arc<Mutex<Option<JoinHandle<()>>>>,
|
||||
}
|
||||
|
||||
@@ -231,6 +235,7 @@ impl TurnMetadataState {
|
||||
enriched_header: Arc::new(RwLock::new(None)),
|
||||
turn_started_at_unix_ms: Arc::new(RwLock::new(None)),
|
||||
responsesapi_client_metadata: Arc::new(RwLock::new(None)),
|
||||
user_input_requested_during_turn: Arc::new(AtomicBool::new(false)),
|
||||
enrichment_task: Arc::new(Mutex::new(None)),
|
||||
}
|
||||
}
|
||||
@@ -285,9 +290,25 @@ impl TurnMetadataState {
|
||||
metadata.remove(REASONING_EFFORT_KEY);
|
||||
}
|
||||
}
|
||||
if self
|
||||
.user_input_requested_during_turn
|
||||
.load(Ordering::Relaxed)
|
||||
{
|
||||
metadata.insert(
|
||||
USER_INPUT_REQUESTED_DURING_TURN_KEY.to_string(),
|
||||
Value::Bool(true),
|
||||
);
|
||||
} else {
|
||||
metadata.remove(USER_INPUT_REQUESTED_DURING_TURN_KEY);
|
||||
}
|
||||
Some(Value::Object(metadata))
|
||||
}
|
||||
|
||||
pub(crate) fn mark_user_input_requested_during_turn(&self) {
|
||||
self.user_input_requested_during_turn
|
||||
.store(true, Ordering::Relaxed);
|
||||
}
|
||||
|
||||
pub(crate) fn set_responsesapi_client_metadata(
|
||||
&self,
|
||||
responsesapi_client_metadata: HashMap<String, String>,
|
||||
|
||||
@@ -213,6 +213,56 @@ fn turn_metadata_state_includes_model_and_reasoning_effort_only_in_request_meta(
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn turn_metadata_state_marks_user_input_requested_during_turn_only_for_mcp_request_meta() {
|
||||
let temp_dir = TempDir::new().expect("temp dir");
|
||||
let cwd = temp_dir.path().abs();
|
||||
let permission_profile = PermissionProfile::read_only();
|
||||
|
||||
let state = TurnMetadataState::new(
|
||||
"session-a".to_string(),
|
||||
"thread-a".to_string(),
|
||||
/*thread_source*/ None,
|
||||
"turn-a".to_string(),
|
||||
cwd,
|
||||
&permission_profile,
|
||||
WindowsSandboxLevel::Disabled,
|
||||
/*enforce_managed_network*/ false,
|
||||
);
|
||||
|
||||
let header = state.current_header_value().expect("header");
|
||||
let header_json: Value = serde_json::from_str(&header).expect("json");
|
||||
assert!(
|
||||
header_json
|
||||
.get(USER_INPUT_REQUESTED_DURING_TURN_KEY)
|
||||
.is_none()
|
||||
);
|
||||
|
||||
let meta = state
|
||||
.current_meta_value_for_mcp_request(test_mcp_turn_metadata_context())
|
||||
.expect("turn metadata should be present");
|
||||
assert!(meta.get(USER_INPUT_REQUESTED_DURING_TURN_KEY).is_none());
|
||||
|
||||
state.mark_user_input_requested_during_turn();
|
||||
|
||||
let header = state.current_header_value().expect("header");
|
||||
let header_json: Value = serde_json::from_str(&header).expect("json");
|
||||
assert!(
|
||||
header_json
|
||||
.get(USER_INPUT_REQUESTED_DURING_TURN_KEY)
|
||||
.is_none()
|
||||
);
|
||||
|
||||
let meta = state
|
||||
.current_meta_value_for_mcp_request(test_mcp_turn_metadata_context())
|
||||
.expect("turn metadata should be present");
|
||||
assert_eq!(
|
||||
meta.get(USER_INPUT_REQUESTED_DURING_TURN_KEY)
|
||||
.and_then(Value::as_bool),
|
||||
Some(true)
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn turn_metadata_state_ignores_client_turn_started_at_unix_ms_before_start() {
|
||||
let temp_dir = TempDir::new().expect("temp dir");
|
||||
|
||||
@@ -1,4 +1,10 @@
|
||||
use crate::test_codex::TestCodexBuilder;
|
||||
use crate::test_codex::test_codex;
|
||||
use anyhow::Result;
|
||||
use codex_core::config::Config;
|
||||
use codex_features::Feature;
|
||||
use codex_login::CodexAuth;
|
||||
use codex_models_manager::bundled_models_response;
|
||||
use serde_json::Value;
|
||||
use serde_json::json;
|
||||
use wiremock::Mock;
|
||||
@@ -15,10 +21,21 @@ const CONNECTOR_NAME: &str = "Calendar";
|
||||
const DISCOVERABLE_CALENDAR_ID: &str = "connector_2128aebfecb84f64a069897515042a44";
|
||||
const DISCOVERABLE_GMAIL_ID: &str = "connector_68df038e0ba48191908c8434991bbac2";
|
||||
const CONNECTOR_DESCRIPTION: &str = "Plan events and manage your calendar.";
|
||||
const CODEX_APPS_META_KEY: &str = "_codex_apps";
|
||||
const PROTOCOL_VERSION: &str = "2025-11-25";
|
||||
const SERVER_NAME: &str = "codex-apps-test";
|
||||
const SERVER_VERSION: &str = "1.0.0";
|
||||
const SEARCHABLE_TOOL_COUNT: usize = 100;
|
||||
const CALENDAR_CREATE_EVENT_TOOL_NAME: &str = "calendar_create_event";
|
||||
pub const CALENDAR_EXTRACT_TEXT_TOOL_NAME: &str = "calendar_extract_text";
|
||||
const CALENDAR_LIST_EVENTS_TOOL_NAME: &str = "calendar_list_events";
|
||||
pub const DIRECT_CALENDAR_CREATE_EVENT_TOOL: &str = "mcp__codex_apps__calendar_create_event";
|
||||
pub const DIRECT_CALENDAR_LIST_EVENTS_TOOL: &str = "mcp__codex_apps__calendar_list_events";
|
||||
pub const DIRECT_CALENDAR_EXTRACT_TEXT_TOOL: &str = "mcp__codex_apps__calendar_extract_text";
|
||||
pub const SEARCH_CALENDAR_NAMESPACE: &str = "mcp__codex_apps__calendar";
|
||||
pub const SEARCH_CALENDAR_CREATE_TOOL: &str = "_create_event";
|
||||
pub const SEARCH_CALENDAR_EXTRACT_TEXT_TOOL: &str = "_extract_text";
|
||||
pub const SEARCH_CALENDAR_LIST_TOOL: &str = "_list_events";
|
||||
pub const CALENDAR_CREATE_EVENT_RESOURCE_URI: &str =
|
||||
"connector://calendar/tools/calendar_create_event";
|
||||
pub const CALENDAR_CREATE_EVENT_MCP_APP_RESOURCE_URI: &str =
|
||||
@@ -71,6 +88,103 @@ impl AppsTestServer {
|
||||
}
|
||||
}
|
||||
|
||||
pub fn configure_search_capable_model(config: &mut Config) {
|
||||
let mut model_catalog = bundled_models_response()
|
||||
.unwrap_or_else(|err| panic!("bundled models.json should parse: {err}"));
|
||||
let model = model_catalog
|
||||
.models
|
||||
.iter_mut()
|
||||
.find(|model| model.slug == "gpt-5.4")
|
||||
.expect("gpt-5.4 exists in bundled models.json");
|
||||
config.model = Some("gpt-5.4".to_string());
|
||||
model.supports_search_tool = true;
|
||||
config.model_catalog = Some(model_catalog);
|
||||
}
|
||||
|
||||
fn configure_apps(config: &mut Config, apps_base_url: &str) {
|
||||
config
|
||||
.features
|
||||
.enable(Feature::Apps)
|
||||
.expect("test config should allow feature update");
|
||||
config.chatgpt_base_url = apps_base_url.to_string();
|
||||
}
|
||||
|
||||
pub fn configure_search_capable_apps(config: &mut Config, apps_base_url: &str) {
|
||||
configure_apps(config, apps_base_url);
|
||||
configure_search_capable_model(config);
|
||||
}
|
||||
|
||||
pub fn apps_enabled_builder(apps_base_url: impl Into<String>) -> TestCodexBuilder {
|
||||
let apps_base_url = apps_base_url.into();
|
||||
test_codex()
|
||||
.with_auth(CodexAuth::create_dummy_chatgpt_auth_for_testing())
|
||||
.with_config(move |config| configure_apps(config, apps_base_url.as_str()))
|
||||
}
|
||||
|
||||
pub fn search_capable_apps_builder(apps_base_url: impl Into<String>) -> TestCodexBuilder {
|
||||
let apps_base_url = apps_base_url.into();
|
||||
test_codex()
|
||||
.with_auth(CodexAuth::create_dummy_chatgpt_auth_for_testing())
|
||||
.with_config(move |config| configure_search_capable_apps(config, apps_base_url.as_str()))
|
||||
}
|
||||
|
||||
fn apps_tool_call_id(body: &Value) -> Option<&str> {
|
||||
body.get("params")?
|
||||
.get("_meta")?
|
||||
.get(CODEX_APPS_META_KEY)?
|
||||
.get("call_id")?
|
||||
.as_str()
|
||||
}
|
||||
|
||||
async fn recorded_apps_tool_calls(server: &MockServer) -> Vec<Value> {
|
||||
server
|
||||
.received_requests()
|
||||
.await
|
||||
.expect("mock server should capture requests")
|
||||
.into_iter()
|
||||
.filter_map(|request| {
|
||||
let body: Value = serde_json::from_slice(&request.body).ok()?;
|
||||
(request.url.path() == "/api/codex/apps"
|
||||
&& body.get("method").and_then(Value::as_str) == Some("tools/call"))
|
||||
.then_some(body)
|
||||
})
|
||||
.collect()
|
||||
}
|
||||
|
||||
pub async fn recorded_apps_tool_call_by_call_id(server: &MockServer, call_id: &str) -> Value {
|
||||
let matches = recorded_apps_tool_calls(server)
|
||||
.await
|
||||
.into_iter()
|
||||
.filter(|body| apps_tool_call_id(body) == Some(call_id))
|
||||
.collect::<Vec<_>>();
|
||||
assert_eq!(
|
||||
matches.len(),
|
||||
1,
|
||||
"expected exactly one apps tools/call request for call_id {call_id}"
|
||||
);
|
||||
matches
|
||||
.into_iter()
|
||||
.next()
|
||||
.expect("matching apps tools/call request should be recorded")
|
||||
}
|
||||
|
||||
pub async fn recorded_apps_tool_call_by_name(server: &MockServer, tool_name: &str) -> Value {
|
||||
let matches = recorded_apps_tool_calls(server)
|
||||
.await
|
||||
.into_iter()
|
||||
.filter(|body| body.pointer("/params/name").and_then(Value::as_str) == Some(tool_name))
|
||||
.collect::<Vec<_>>();
|
||||
assert_eq!(
|
||||
matches.len(),
|
||||
1,
|
||||
"expected exactly one apps tools/call request for tool {tool_name}"
|
||||
);
|
||||
matches
|
||||
.into_iter()
|
||||
.next()
|
||||
.expect("matching apps tools/call request should be recorded")
|
||||
}
|
||||
|
||||
async fn mount_oauth_metadata(server: &MockServer) {
|
||||
Mock::given(method("GET"))
|
||||
.and(path("/.well-known/oauth-authorization-server/mcp"))
|
||||
@@ -187,7 +301,7 @@ impl Respond for CodexAppsJsonRpcResponder {
|
||||
"result": {
|
||||
"tools": [
|
||||
{
|
||||
"name": "calendar_create_event",
|
||||
"name": CALENDAR_CREATE_EVENT_TOOL_NAME,
|
||||
"description": "Create a calendar event.",
|
||||
"annotations": {
|
||||
"readOnlyHint": false,
|
||||
@@ -217,7 +331,7 @@ impl Respond for CodexAppsJsonRpcResponder {
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "calendar_list_events",
|
||||
"name": CALENDAR_LIST_EVENTS_TOOL_NAME,
|
||||
"description": "List calendar events.",
|
||||
"annotations": {
|
||||
"readOnlyHint": true
|
||||
@@ -242,7 +356,7 @@ impl Respond for CodexAppsJsonRpcResponder {
|
||||
}
|
||||
},
|
||||
{
|
||||
"name": "calendar_extract_text",
|
||||
"name": CALENDAR_EXTRACT_TEXT_TOOL_NAME,
|
||||
"description": "Extract text from an uploaded document.",
|
||||
"annotations": {
|
||||
"readOnlyHint": false
|
||||
|
||||
312
codex-rs/core/tests/suite/mcp_turn_metadata.rs
Normal file
312
codex-rs/core/tests/suite/mcp_turn_metadata.rs
Normal file
@@ -0,0 +1,312 @@
|
||||
#![cfg(not(target_os = "windows"))]
|
||||
#![allow(clippy::unwrap_used, clippy::expect_used)]
|
||||
|
||||
use anyhow::Result;
|
||||
use codex_config::types::AppToolApproval;
|
||||
use codex_core::config::Config;
|
||||
use codex_features::Feature;
|
||||
use codex_protocol::config_types::CollaborationMode;
|
||||
use codex_protocol::config_types::ModeKind;
|
||||
use codex_protocol::config_types::Settings;
|
||||
use codex_protocol::models::PermissionProfile;
|
||||
use codex_protocol::protocol::AskForApproval;
|
||||
use codex_protocol::protocol::ElicitationAction;
|
||||
use codex_protocol::protocol::EventMsg;
|
||||
use codex_protocol::protocol::Op;
|
||||
use codex_protocol::request_user_input::RequestUserInputAnswer;
|
||||
use codex_protocol::request_user_input::RequestUserInputResponse;
|
||||
use codex_protocol::user_input::UserInput;
|
||||
use core_test_support::PathExt;
|
||||
use core_test_support::apps_test_server::AppsTestServer;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_CREATE_TOOL;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_NAMESPACE;
|
||||
use core_test_support::apps_test_server::recorded_apps_tool_call_by_call_id;
|
||||
use core_test_support::apps_test_server::search_capable_apps_builder;
|
||||
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_function_call_with_namespace;
|
||||
use core_test_support::responses::ev_response_created;
|
||||
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::turn_permission_fields;
|
||||
use core_test_support::wait_for_event;
|
||||
use core_test_support::wait_for_event_match;
|
||||
use pretty_assertions::assert_eq;
|
||||
use serde_json::json;
|
||||
use std::collections::HashMap;
|
||||
|
||||
fn set_calendar_approval_mode(config: &mut Config, approval_mode: AppToolApproval) {
|
||||
let approval_mode = match approval_mode {
|
||||
AppToolApproval::Auto => "auto",
|
||||
AppToolApproval::Prompt => "prompt",
|
||||
AppToolApproval::Approve => "approve",
|
||||
};
|
||||
let user_config_path = config.codex_home.join("config.toml").abs();
|
||||
let user_config = toml::from_str(&format!(
|
||||
r#"
|
||||
[apps.calendar]
|
||||
default_tools_approval_mode = "{approval_mode}"
|
||||
"#
|
||||
))
|
||||
.expect("apps config should parse");
|
||||
config.config_layer_stack = config
|
||||
.config_layer_stack
|
||||
.with_user_config(&user_config_path, user_config);
|
||||
}
|
||||
|
||||
async fn submit_user_turn(
|
||||
test: &TestCodex,
|
||||
text: &str,
|
||||
approval_policy: AskForApproval,
|
||||
collaboration_mode: Option<CollaborationMode>,
|
||||
) -> Result<()> {
|
||||
let (sandbox_policy, permission_profile) =
|
||||
turn_permission_fields(PermissionProfile::Disabled, test.cwd.path());
|
||||
test.codex
|
||||
.submit(Op::UserTurn {
|
||||
environments: None,
|
||||
items: vec![UserInput::Text {
|
||||
text: text.to_string(),
|
||||
text_elements: Vec::new(),
|
||||
}],
|
||||
final_output_json_schema: None,
|
||||
cwd: test.cwd.path().to_path_buf(),
|
||||
approval_policy,
|
||||
approvals_reviewer: None,
|
||||
sandbox_policy,
|
||||
permission_profile,
|
||||
model: test.session_configured.model.clone(),
|
||||
effort: None,
|
||||
summary: None,
|
||||
service_tier: None,
|
||||
collaboration_mode,
|
||||
personality: None,
|
||||
})
|
||||
.await?;
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
|
||||
async fn approved_mcp_tool_call_metadata_records_prior_user_input_request() -> Result<()> {
|
||||
skip_if_no_network!(Ok(()));
|
||||
|
||||
let server = start_mock_server().await;
|
||||
let apps_server = AppsTestServer::mount(&server).await?;
|
||||
let call_id = "calendar-call-approval";
|
||||
let calendar_args = serde_json::to_string(&json!({
|
||||
"title": "Lunch",
|
||||
"starts_at": "2026-03-10T12:00:00Z"
|
||||
}))?;
|
||||
let mock = mount_sse_sequence(
|
||||
&server,
|
||||
vec![
|
||||
sse(vec![
|
||||
ev_response_created("resp-1"),
|
||||
ev_function_call_with_namespace(
|
||||
call_id,
|
||||
SEARCH_CALENDAR_NAMESPACE,
|
||||
SEARCH_CALENDAR_CREATE_TOOL,
|
||||
&calendar_args,
|
||||
),
|
||||
ev_completed("resp-1"),
|
||||
]),
|
||||
sse(vec![
|
||||
ev_response_created("resp-2"),
|
||||
ev_assistant_message("msg-1", "done"),
|
||||
ev_completed("resp-2"),
|
||||
]),
|
||||
],
|
||||
)
|
||||
.await;
|
||||
|
||||
let mut builder = search_capable_apps_builder(apps_server.chatgpt_base_url.clone())
|
||||
.with_config(|config| {
|
||||
config
|
||||
.features
|
||||
.enable(Feature::ToolCallMcpElicitation)
|
||||
.expect("test config should allow feature update");
|
||||
set_calendar_approval_mode(config, AppToolApproval::Prompt);
|
||||
});
|
||||
let test = builder.build(&server).await?;
|
||||
|
||||
submit_user_turn(
|
||||
&test,
|
||||
"Use [$calendar](app://calendar) to create a calendar event.",
|
||||
AskForApproval::OnRequest,
|
||||
/*collaboration_mode*/ None,
|
||||
)
|
||||
.await?;
|
||||
|
||||
let EventMsg::McpToolCallBegin(begin) = wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::McpToolCallBegin(_))
|
||||
})
|
||||
.await
|
||||
else {
|
||||
unreachable!("event guard guarantees McpToolCallBegin");
|
||||
};
|
||||
assert_eq!(begin.call_id, call_id);
|
||||
|
||||
let EventMsg::ElicitationRequest(request) = wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::ElicitationRequest(_))
|
||||
})
|
||||
.await
|
||||
else {
|
||||
unreachable!("event guard guarantees ElicitationRequest");
|
||||
};
|
||||
|
||||
test.codex
|
||||
.submit(Op::ResolveElicitation {
|
||||
server_name: request.server_name,
|
||||
request_id: request.id,
|
||||
decision: ElicitationAction::Accept,
|
||||
content: None,
|
||||
meta: None,
|
||||
})
|
||||
.await?;
|
||||
|
||||
wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::TurnComplete(_))
|
||||
})
|
||||
.await;
|
||||
|
||||
assert_eq!(mock.requests().len(), 2);
|
||||
let apps_tool_call = recorded_apps_tool_call_by_call_id(&server, call_id).await;
|
||||
|
||||
assert_eq!(
|
||||
apps_tool_call
|
||||
.pointer("/params/_meta/x-codex-turn-metadata/user_input_requested_during_turn"),
|
||||
Some(&json!(true))
|
||||
);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
|
||||
async fn mcp_tool_call_metadata_records_prior_request_user_input_tool() -> Result<()> {
|
||||
skip_if_no_network!(Ok(()));
|
||||
|
||||
let server = start_mock_server().await;
|
||||
let apps_server = AppsTestServer::mount(&server).await?;
|
||||
let request_user_input_call_id = "user-input-call";
|
||||
let calendar_call_id = "calendar-call-after-user-input";
|
||||
let request_user_input_args = json!({
|
||||
"questions": [{
|
||||
"id": "confirm_path",
|
||||
"header": "Confirm",
|
||||
"question": "Proceed with the plan?",
|
||||
"options": [{
|
||||
"label": "Yes (Recommended)",
|
||||
"description": "Continue the current plan."
|
||||
}, {
|
||||
"label": "No",
|
||||
"description": "Stop and revisit the approach."
|
||||
}]
|
||||
}]
|
||||
})
|
||||
.to_string();
|
||||
let calendar_args = serde_json::to_string(&json!({
|
||||
"title": "Lunch",
|
||||
"starts_at": "2026-03-10T12:00:00Z"
|
||||
}))?;
|
||||
let mock = mount_sse_sequence(
|
||||
&server,
|
||||
vec![
|
||||
sse(vec![
|
||||
ev_response_created("resp-1"),
|
||||
ev_function_call(
|
||||
request_user_input_call_id,
|
||||
"request_user_input",
|
||||
&request_user_input_args,
|
||||
),
|
||||
ev_completed("resp-1"),
|
||||
]),
|
||||
sse(vec![
|
||||
ev_response_created("resp-2"),
|
||||
ev_function_call_with_namespace(
|
||||
calendar_call_id,
|
||||
SEARCH_CALENDAR_NAMESPACE,
|
||||
SEARCH_CALENDAR_CREATE_TOOL,
|
||||
&calendar_args,
|
||||
),
|
||||
ev_completed("resp-2"),
|
||||
]),
|
||||
sse(vec![
|
||||
ev_response_created("resp-3"),
|
||||
ev_assistant_message("msg-1", "done"),
|
||||
ev_completed("resp-3"),
|
||||
]),
|
||||
],
|
||||
)
|
||||
.await;
|
||||
|
||||
let mut builder = search_capable_apps_builder(apps_server.chatgpt_base_url.clone())
|
||||
.with_config(|config| {
|
||||
set_calendar_approval_mode(config, AppToolApproval::Approve);
|
||||
});
|
||||
let test = builder.build(&server).await?;
|
||||
|
||||
submit_user_turn(
|
||||
&test,
|
||||
"Ask for confirmation, then create a calendar event.",
|
||||
AskForApproval::Never,
|
||||
Some(CollaborationMode {
|
||||
mode: ModeKind::Plan,
|
||||
settings: Settings {
|
||||
model: test.session_configured.model.clone(),
|
||||
reasoning_effort: None,
|
||||
developer_instructions: None,
|
||||
},
|
||||
}),
|
||||
)
|
||||
.await?;
|
||||
|
||||
let request = wait_for_event_match(&test.codex, |event| match event {
|
||||
EventMsg::RequestUserInput(request) => Some(request.clone()),
|
||||
_ => None,
|
||||
})
|
||||
.await;
|
||||
assert_eq!(request.call_id, request_user_input_call_id);
|
||||
|
||||
test.codex
|
||||
.submit(Op::UserInputAnswer {
|
||||
id: request.turn_id,
|
||||
response: RequestUserInputResponse {
|
||||
answers: HashMap::from([(
|
||||
"confirm_path".to_string(),
|
||||
RequestUserInputAnswer {
|
||||
answers: vec!["Yes (Recommended)".to_string()],
|
||||
},
|
||||
)]),
|
||||
},
|
||||
})
|
||||
.await?;
|
||||
|
||||
let EventMsg::McpToolCallBegin(begin) = wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::McpToolCallBegin(_))
|
||||
})
|
||||
.await
|
||||
else {
|
||||
unreachable!("event guard guarantees McpToolCallBegin");
|
||||
};
|
||||
assert_eq!(begin.call_id, calendar_call_id);
|
||||
|
||||
wait_for_event(&test.codex, |event| {
|
||||
matches!(event, EventMsg::TurnComplete(_))
|
||||
})
|
||||
.await;
|
||||
|
||||
assert_eq!(mock.requests().len(), 3);
|
||||
let apps_tool_call = recorded_apps_tool_call_by_call_id(&server, calendar_call_id).await;
|
||||
|
||||
assert_eq!(
|
||||
apps_tool_call
|
||||
.pointer("/params/_meta/x-codex-turn-metadata/user_input_requested_during_turn"),
|
||||
Some(&json!(true))
|
||||
);
|
||||
|
||||
Ok(())
|
||||
}
|
||||
@@ -57,6 +57,7 @@ mod image_rollout;
|
||||
mod items;
|
||||
mod json_result;
|
||||
mod live_cli;
|
||||
mod mcp_turn_metadata;
|
||||
mod model_overrides;
|
||||
mod model_switching;
|
||||
mod model_visible_layout;
|
||||
|
||||
@@ -5,13 +5,16 @@ use std::path::Path;
|
||||
|
||||
use anyhow::Context;
|
||||
use anyhow::Result;
|
||||
use codex_core::config::Config;
|
||||
use codex_features::Feature;
|
||||
use codex_login::CodexAuth;
|
||||
use codex_protocol::models::PermissionProfile;
|
||||
use codex_protocol::protocol::AskForApproval;
|
||||
use core_test_support::apps_test_server::AppsTestServer;
|
||||
use core_test_support::apps_test_server::CALENDAR_EXTRACT_TEXT_TOOL_NAME;
|
||||
use core_test_support::apps_test_server::DIRECT_CALENDAR_EXTRACT_TEXT_TOOL as DOCUMENT_EXTRACT_HOOK_MATCHER;
|
||||
use core_test_support::apps_test_server::DOCUMENT_EXTRACT_TEXT_RESOURCE_URI;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_EXTRACT_TEXT_TOOL as DOCUMENT_EXTRACT_TOOL;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_NAMESPACE as DOCUMENT_EXTRACT_NAMESPACE;
|
||||
use core_test_support::apps_test_server::apps_enabled_builder;
|
||||
use core_test_support::apps_test_server::recorded_apps_tool_call_by_name;
|
||||
use core_test_support::hooks::trust_discovered_hooks;
|
||||
use core_test_support::responses::ev_assistant_message;
|
||||
use core_test_support::responses::ev_completed;
|
||||
@@ -20,7 +23,6 @@ use core_test_support::responses::ev_response_created;
|
||||
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::test_codex::test_codex;
|
||||
use pretty_assertions::assert_eq;
|
||||
use serde_json::Value;
|
||||
use serde_json::json;
|
||||
@@ -31,17 +33,6 @@ use wiremock::matchers::header;
|
||||
use wiremock::matchers::method;
|
||||
use wiremock::matchers::path;
|
||||
|
||||
const DOCUMENT_EXTRACT_NAMESPACE: &str = "mcp__codex_apps__calendar";
|
||||
const DOCUMENT_EXTRACT_TOOL: &str = "_extract_text";
|
||||
const DOCUMENT_EXTRACT_HOOK_MATCHER: &str = "mcp__codex_apps__calendar_extract_text";
|
||||
|
||||
fn configure_apps(config: &mut Config, chatgpt_base_url: &str) {
|
||||
if let Err(err) = config.features.enable(Feature::Apps) {
|
||||
panic!("test config should allow feature update: {err}");
|
||||
}
|
||||
config.chatgpt_base_url = chatgpt_base_url.to_string();
|
||||
}
|
||||
|
||||
fn write_post_tool_use_hook(home: &Path) -> Result<()> {
|
||||
let script_path = home.join("post_tool_use_hook.py");
|
||||
let log_path = home.join("post_tool_use_hook_log.jsonl");
|
||||
@@ -154,15 +145,13 @@ async fn codex_apps_file_params_upload_local_paths_before_mcp_tool_call() -> Res
|
||||
)
|
||||
.await;
|
||||
|
||||
let mut builder = test_codex()
|
||||
.with_auth(CodexAuth::create_dummy_chatgpt_auth_for_testing())
|
||||
let mut builder = apps_enabled_builder(apps_server.chatgpt_base_url.clone())
|
||||
.with_pre_build_hook(move |home| {
|
||||
if let Err(error) = write_post_tool_use_hook(home) {
|
||||
panic!("failed to write apps file post tool use hook fixture: {error}");
|
||||
}
|
||||
})
|
||||
.with_config(move |config| {
|
||||
configure_apps(config, apps_server.chatgpt_base_url.as_str());
|
||||
trust_discovered_hooks(config);
|
||||
});
|
||||
let test = builder.build(&server).await?;
|
||||
@@ -192,20 +181,8 @@ async fn codex_apps_file_params_upload_local_paths_before_mcp_tool_call() -> Res
|
||||
}))
|
||||
);
|
||||
|
||||
let apps_tool_call = server
|
||||
.received_requests()
|
||||
.await
|
||||
.unwrap_or_default()
|
||||
.into_iter()
|
||||
.find_map(|request| {
|
||||
let body: Value = serde_json::from_slice(&request.body).ok()?;
|
||||
(request.url.path() == "/api/codex/apps"
|
||||
&& body.get("method").and_then(Value::as_str) == Some("tools/call")
|
||||
&& body.pointer("/params/name").and_then(Value::as_str)
|
||||
== Some("calendar_extract_text"))
|
||||
.then_some(body)
|
||||
})
|
||||
.expect("apps calendar_extract_text tools/call request should be recorded");
|
||||
let apps_tool_call =
|
||||
recorded_apps_tool_call_by_name(&server, CALENDAR_EXTRACT_TEXT_TOOL_NAME).await;
|
||||
|
||||
assert_eq!(
|
||||
apps_tool_call.pointer("/params/arguments/file"),
|
||||
|
||||
@@ -7,7 +7,6 @@ use codex_config::types::McpServerTransportConfig;
|
||||
use codex_core::config::Config;
|
||||
use codex_features::Feature;
|
||||
use codex_login::CodexAuth;
|
||||
use codex_models_manager::bundled_models_response;
|
||||
use codex_protocol::dynamic_tools::DynamicToolCallOutputContentItem;
|
||||
use codex_protocol::dynamic_tools::DynamicToolResponse;
|
||||
use codex_protocol::dynamic_tools::DynamicToolSpec;
|
||||
@@ -21,6 +20,15 @@ use codex_protocol::user_input::UserInput;
|
||||
use core_test_support::apps_test_server::AppsTestServer;
|
||||
use core_test_support::apps_test_server::CALENDAR_CREATE_EVENT_MCP_APP_RESOURCE_URI;
|
||||
use core_test_support::apps_test_server::CALENDAR_CREATE_EVENT_RESOURCE_URI;
|
||||
use core_test_support::apps_test_server::DIRECT_CALENDAR_CREATE_EVENT_TOOL as CALENDAR_CREATE_TOOL;
|
||||
use core_test_support::apps_test_server::DIRECT_CALENDAR_LIST_EVENTS_TOOL as CALENDAR_LIST_TOOL;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_CREATE_TOOL;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_LIST_TOOL;
|
||||
use core_test_support::apps_test_server::SEARCH_CALENDAR_NAMESPACE;
|
||||
use core_test_support::apps_test_server::configure_search_capable_apps;
|
||||
use core_test_support::apps_test_server::configure_search_capable_model;
|
||||
use core_test_support::apps_test_server::recorded_apps_tool_call_by_call_id;
|
||||
use core_test_support::apps_test_server::search_capable_apps_builder as configured_builder;
|
||||
use core_test_support::responses::ResponsesRequest;
|
||||
use core_test_support::responses::ev_assistant_message;
|
||||
use core_test_support::responses::ev_completed;
|
||||
@@ -34,7 +42,6 @@ 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::stdio_server_bin;
|
||||
use core_test_support::test_codex::TestCodexBuilder;
|
||||
use core_test_support::test_codex::test_codex;
|
||||
use core_test_support::wait_for_event;
|
||||
use pretty_assertions::assert_eq;
|
||||
@@ -48,11 +55,6 @@ const SEARCH_TOOL_DESCRIPTION_SNIPPETS: [&str; 2] = [
|
||||
"- Calendar: Plan events and manage your calendar.",
|
||||
];
|
||||
const TOOL_SEARCH_TOOL_NAME: &str = "tool_search";
|
||||
const CALENDAR_CREATE_TOOL: &str = "mcp__codex_apps__calendar_create_event";
|
||||
const CALENDAR_LIST_TOOL: &str = "mcp__codex_apps__calendar_list_events";
|
||||
const SEARCH_CALENDAR_NAMESPACE: &str = "mcp__codex_apps__calendar";
|
||||
const SEARCH_CALENDAR_CREATE_TOOL: &str = "_create_event";
|
||||
const SEARCH_CALENDAR_LIST_TOOL: &str = "_list_events";
|
||||
|
||||
fn tool_names(body: &Value) -> Vec<String> {
|
||||
body.get("tools")
|
||||
@@ -111,28 +113,6 @@ fn tool_search_output_has_namespace_child(
|
||||
namespace_child_tool(&output, namespace, tool_name).is_some()
|
||||
}
|
||||
|
||||
fn configure_search_capable_model(config: &mut Config) {
|
||||
let mut model_catalog = bundled_models_response()
|
||||
.unwrap_or_else(|err| panic!("bundled models.json should parse: {err}"));
|
||||
let model = model_catalog
|
||||
.models
|
||||
.iter_mut()
|
||||
.find(|model| model.slug == "gpt-5.4")
|
||||
.expect("gpt-5.4 exists in bundled models.json");
|
||||
config.model = Some("gpt-5.4".to_string());
|
||||
model.supports_search_tool = true;
|
||||
config.model_catalog = Some(model_catalog);
|
||||
}
|
||||
|
||||
fn configure_search_capable_apps(config: &mut Config, apps_base_url: &str) {
|
||||
config
|
||||
.features
|
||||
.enable(Feature::Apps)
|
||||
.expect("test config should allow feature update");
|
||||
config.chatgpt_base_url = apps_base_url.to_string();
|
||||
configure_search_capable_model(config);
|
||||
}
|
||||
|
||||
fn configure_apps_without_tool_search(config: &mut Config, apps_base_url: &str) {
|
||||
configure_search_capable_apps(config, apps_base_url);
|
||||
config
|
||||
@@ -141,16 +121,6 @@ fn configure_apps_without_tool_search(config: &mut Config, apps_base_url: &str)
|
||||
.expect("test config should allow feature update");
|
||||
}
|
||||
|
||||
fn configure_apps(config: &mut Config, apps_base_url: &str) {
|
||||
configure_search_capable_apps(config, apps_base_url);
|
||||
}
|
||||
|
||||
fn configured_builder(apps_base_url: String) -> TestCodexBuilder {
|
||||
test_codex()
|
||||
.with_auth(CodexAuth::create_dummy_chatgpt_auth_for_testing())
|
||||
.with_config(move |config| configure_apps(config, apps_base_url.as_str()))
|
||||
}
|
||||
|
||||
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
|
||||
async fn search_tool_enabled_by_default_adds_tool_search() -> Result<()> {
|
||||
skip_if_no_network!(Ok(()));
|
||||
@@ -321,7 +291,9 @@ async fn search_tool_is_hidden_for_api_key_auth() -> Result<()> {
|
||||
|
||||
let mut builder = test_codex()
|
||||
.with_auth(CodexAuth::from_api_key("Test API Key"))
|
||||
.with_config(move |config| configure_apps(config, apps_server.chatgpt_base_url.as_str()));
|
||||
.with_config(move |config| {
|
||||
configure_search_capable_apps(config, apps_server.chatgpt_base_url.as_str())
|
||||
});
|
||||
let test = builder.build(&server).await?;
|
||||
|
||||
test.submit_turn_with_approval_and_permission_profile(
|
||||
@@ -585,18 +557,7 @@ async fn tool_search_returns_deferred_tools_without_follow_up_tool_injection() -
|
||||
assert_eq!(requests.len(), 3);
|
||||
let first_request_body = requests[0].body_json();
|
||||
|
||||
let apps_tool_call = server
|
||||
.received_requests()
|
||||
.await
|
||||
.unwrap_or_default()
|
||||
.into_iter()
|
||||
.find_map(|request| {
|
||||
let body: Value = serde_json::from_slice(&request.body).ok()?;
|
||||
(request.url.path() == "/api/codex/apps"
|
||||
&& body.get("method").and_then(Value::as_str) == Some("tools/call"))
|
||||
.then_some(body)
|
||||
})
|
||||
.expect("apps tools/call request should be recorded");
|
||||
let apps_tool_call = recorded_apps_tool_call_by_call_id(&server, "calendar-call-1").await;
|
||||
|
||||
assert_eq!(
|
||||
apps_tool_call.pointer("/params/_meta/_codex_apps"),
|
||||
|
||||
Reference in New Issue
Block a user