Compare commits

..

1 Commits

Author SHA1 Message Date
Ahmed Ibrahim
2481584b88 Release 0.78.0-alpha.3 2025-12-22 12:53:07 -08:00
294 changed files with 5472 additions and 17320 deletions

BIN
.github/codex-cli-login.png vendored Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 2.9 MiB

BIN
.github/codex-cli-permissions.png vendored Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 408 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 818 KiB

After

Width:  |  Height:  |  Size: 3.1 MiB

BIN
.github/demo.gif vendored Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 19 MiB

View File

@@ -12,8 +12,6 @@ permissions:
jobs:
close-stale-contributor-prs:
# Prevent scheduled runs on forks
if: github.repository == 'openai/codex'
runs-on: ubuntu-latest
steps:
- name: Close inactive PRs from contributors

View File

@@ -9,8 +9,7 @@ on:
jobs:
gather-duplicates:
name: Identify potential duplicates
# Prevent runs on forks (requires OpenAI API key, wastes Actions minutes)
if: github.repository == 'openai/codex' && (github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-deduplicate'))
if: ${{ github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-deduplicate') }}
runs-on: ubuntu-latest
permissions:
contents: read

View File

@@ -9,8 +9,7 @@ on:
jobs:
gather-labels:
name: Generate label suggestions
# Prevent runs on forks (requires OpenAI API key, wastes Actions minutes)
if: github.repository == 'openai/codex' && (github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-label'))
if: ${{ github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-label') }}
runs-on: ubuntu-latest
permissions:
contents: read

View File

@@ -14,8 +14,6 @@ permissions:
jobs:
prepare:
# Prevent scheduled runs on forks (no secrets, wastes Actions minutes)
if: github.repository == 'openai/codex'
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6

View File

@@ -323,26 +323,6 @@ jobs:
- name: Checkout repository
uses: actions/checkout@v6
- name: Generate release notes from tag commit message
id: release_notes
shell: bash
run: |
set -euo pipefail
# On tag pushes, GITHUB_SHA may be a tag object for annotated tags;
# peel it to the underlying commit.
commit="$(git rev-parse "${GITHUB_SHA}^{commit}")"
notes_path="${RUNNER_TEMP}/release-notes.md"
# Use the commit message for the commit the tag points at (not the
# annotated tag message).
git log -1 --format=%B "${commit}" > "${notes_path}"
# Ensure trailing newline so GitHub's markdown renderer doesn't
# occasionally run the last line into subsequent content.
echo >> "${notes_path}"
echo "path=${notes_path}" >> "${GITHUB_OUTPUT}"
- uses: actions/download-artifact@v7
with:
path: dist
@@ -415,7 +395,6 @@ jobs:
with:
name: ${{ steps.release_name.outputs.name }}
tag_name: ${{ github.ref_name }}
body_path: ${{ steps.release_notes.outputs.path }}
files: dist/**
# Mark as prerelease only when the version has a suffix after x.y.z
# (e.g. -alpha, -beta). Otherwise publish a normal release.

View File

@@ -77,12 +77,6 @@ If you dont have the tool:
- Prefer deep equals comparisons whenever possible. Perform `assert_eq!()` on entire objects, rather than individual fields.
- Avoid mutating process environment in tests; prefer passing environment-derived flags or dependencies from above.
### Spawning workspace binaries in tests (Cargo vs Buck2)
- Prefer `codex_utils_cargo_bin::cargo_bin("...")` over `assert_cmd::Command::cargo_bin(...)` or `escargot` when tests need to spawn first-party binaries.
- Under Buck2, `CARGO_BIN_EXE_*` may be project-relative (e.g. `buck-out/...`), which breaks if a test changes its working directory. `codex_utils_cargo_bin::cargo_bin` resolves to an absolute path first.
- When locating fixture files under Buck2, avoid `env!("CARGO_MANIFEST_DIR")` (Buck codegen sets it to `"."`). Prefer deriving paths from `codex_utils_cargo_bin::buck_project_root()` when needed.
### Integration tests (core)
- Prefer the utilities in `core_test_support::responses` when writing end-to-end Codex tests.

View File

@@ -1,11 +1,13 @@
<p align="center"><code>npm i -g @openai/codex</code><br />or <code>brew install --cask codex</code></p>
<p align="center"><strong>Codex CLI</strong> is a coding agent from OpenAI that runs locally on your computer.
</br>
</br>If you want Codex in your code editor (VS Code, Cursor, Windsurf), <a href="https://developers.openai.com/codex/ide">install in your IDE</a>
</br>If you are looking for the <em>cloud-based agent</em> from OpenAI, <strong>Codex Web</strong>, go to <a href="https://chatgpt.com/codex">chatgpt.com/codex</a></p>
<p align="center">
<img src="./.github/codex-cli-splash.png" alt="Codex CLI splash" width="80%" />
</p>
</br>
If you want Codex in your code editor (VS Code, Cursor, Windsurf), <a href="https://developers.openai.com/codex/ide">install in your IDE.</a>
</br>If you are looking for the <em>cloud-based agent</em> from OpenAI, <strong>Codex Web</strong>, go to <a href="https://chatgpt.com/codex">chatgpt.com/codex</a>.</p>
</p>
---
@@ -13,19 +15,25 @@ If you want Codex in your code editor (VS Code, Cursor, Windsurf), <a href="http
### Installing and running Codex CLI
Install globally with your preferred package manager:
Install globally with your preferred package manager. If you use npm:
```shell
# Install using npm
npm install -g @openai/codex
```
Alternatively, if you use Homebrew:
```shell
# Install using Homebrew
brew install --cask codex
```
Then simply run `codex` to get started.
Then simply run `codex` to get started:
```shell
codex
```
If you're running into upgrade issues with Homebrew, see the [FAQ entry on brew upgrade codex](./docs/faq.md#brew-upgrade-codex-isnt-upgrading-me).
<details>
<summary>You can also go to the <a href="https://github.com/openai/codex/releases/latest">latest GitHub Release</a> and download the appropriate binary for your platform.</summary>
@@ -45,15 +53,60 @@ Each archive contains a single entry with the platform baked into the name (e.g.
### Using Codex with your ChatGPT plan
<p align="center">
<img src="./.github/codex-cli-login.png" alt="Codex CLI login" width="80%" />
</p>
Run `codex` and select **Sign in with ChatGPT**. We recommend signing into your ChatGPT account to use Codex as part of your Plus, Pro, Team, Edu, or Enterprise plan. [Learn more about what's included in your ChatGPT plan](https://help.openai.com/en/articles/11369540-codex-in-chatgpt).
You can also use Codex with an API key, but this requires [additional setup](https://developers.openai.com/codex/auth#sign-in-with-an-api-key).
You can also use Codex with an API key, but this requires [additional setup](./docs/authentication.md#usage-based-billing-alternative-use-an-openai-api-key). If you previously used an API key for usage-based billing, see the [migration steps](./docs/authentication.md#migrating-from-usage-based-billing-api-key). If you're having trouble with login, please comment on [this issue](https://github.com/openai/codex/issues/1243).
## Docs
### Model Context Protocol (MCP)
- [**Codex Documentation**](https://developers.openai.com/codex)
Codex can access MCP servers. To configure them, refer to the [config docs](./docs/config.md#mcp_servers).
### Configuration
Codex CLI supports a rich set of configuration options, with preferences stored in `~/.codex/config.toml`. For full configuration options, see [Configuration](./docs/config.md).
### Execpolicy
See the [Execpolicy quickstart](./docs/execpolicy.md) to set up rules that govern what commands Codex can execute.
### Docs & FAQ
- [**Getting started**](./docs/getting-started.md)
- [CLI usage](./docs/getting-started.md#cli-usage)
- [Slash Commands](./docs/slash_commands.md)
- [Running with a prompt as input](./docs/getting-started.md#running-with-a-prompt-as-input)
- [Example prompts](./docs/getting-started.md#example-prompts)
- [Custom prompts](./docs/prompts.md)
- [Memory with AGENTS.md](./docs/getting-started.md#memory-with-agentsmd)
- [**Configuration**](./docs/config.md)
- [Example config](./docs/example-config.md)
- [**Sandbox & approvals**](./docs/sandbox.md)
- [**Execpolicy quickstart**](./docs/execpolicy.md)
- [**Authentication**](./docs/authentication.md)
- [Auth methods](./docs/authentication.md#forcing-a-specific-auth-method-advanced)
- [Login on a "Headless" machine](./docs/authentication.md#connecting-on-a-headless-machine)
- **Automating Codex**
- [GitHub Action](https://github.com/openai/codex-action)
- [TypeScript SDK](./sdk/typescript/README.md)
- [Non-interactive mode (`codex exec`)](./docs/exec.md)
- [**Advanced**](./docs/advanced.md)
- [Tracing / verbose logging](./docs/advanced.md#tracing--verbose-logging)
- [Model Context Protocol (MCP)](./docs/advanced.md#model-context-protocol-mcp)
- [**Zero data retention (ZDR)**](./docs/zdr.md)
- [**Contributing**](./docs/contributing.md)
- [**Installing & building**](./docs/install.md)
- [**Install & build**](./docs/install.md)
- [System Requirements](./docs/install.md#system-requirements)
- [DotSlash](./docs/install.md#dotslash)
- [Build from source](./docs/install.md#build-from-source)
- [**FAQ**](./docs/faq.md)
- [**Open source fund**](./docs/open-source-fund.md)
---
## License
This repository is licensed under the [Apache-2.0 License](LICENSE).

View File

@@ -1,16 +0,0 @@
# Example announcement tips for Codex TUI.
# Each [[announcements]] entry is evaluated in order; the last matching one is shown.
# Dates are UTC, formatted as YYYY-MM-DD. The from_date is inclusive and the to_date is exclusive.
# version_regex matches against the CLI version (env!("CARGO_PKG_VERSION")); omit to apply to all versions.
# target_app specify which app should display the announcement (cli, vsce, ...).
[[announcements]]
content = "Welcome to Codex! Check out the new onboarding flow."
from_date = "2024-10-01"
to_date = "2024-10-15"
target_app = "cli"
[[announcements]]
content = "This is a test announcement"
version_regex = "^0\\.0\\.0$"
to_date = "2026-01-10"

View File

@@ -2,7 +2,6 @@
"""Install Codex native binaries (Rust CLI plus ripgrep helpers)."""
import argparse
from contextlib import contextmanager
import json
import os
import shutil
@@ -13,7 +12,6 @@ import zipfile
from dataclasses import dataclass
from concurrent.futures import ThreadPoolExecutor, as_completed
from pathlib import Path
import sys
from typing import Iterable, Sequence
from urllib.parse import urlparse
from urllib.request import urlopen
@@ -79,45 +77,6 @@ RG_TARGET_PLATFORM_PAIRS: list[tuple[str, str]] = [
RG_TARGET_TO_PLATFORM = {target: platform for target, platform in RG_TARGET_PLATFORM_PAIRS}
DEFAULT_RG_TARGETS = [target for target, _ in RG_TARGET_PLATFORM_PAIRS]
# urllib.request.urlopen() defaults to no timeout (can hang indefinitely), which is painful in CI.
DOWNLOAD_TIMEOUT_SECS = 60
def _gha_enabled() -> bool:
# GitHub Actions supports "workflow commands" (e.g. ::group:: / ::error::) that make logs
# much easier to scan: groups collapse noisy sections and error annotations surface the
# failure in the UI without changing the actual exception/traceback output.
return os.environ.get("GITHUB_ACTIONS") == "true"
def _gha_escape(value: str) -> str:
# Workflow commands require percent/newline escaping.
return value.replace("%", "%25").replace("\r", "%0D").replace("\n", "%0A")
def _gha_error(*, title: str, message: str) -> None:
# Emit a GitHub Actions error annotation. This does not replace stdout/stderr logs; it just
# adds a prominent summary line to the job UI so the root cause is easier to spot.
if not _gha_enabled():
return
print(
f"::error title={_gha_escape(title)}::{_gha_escape(message)}",
flush=True,
)
@contextmanager
def _gha_group(title: str):
# Wrap a block in a collapsible log group on GitHub Actions. Outside of GHA this is a no-op
# so local output remains unchanged.
if _gha_enabled():
print(f"::group::{_gha_escape(title)}", flush=True)
try:
yield
finally:
if _gha_enabled():
print("::endgroup::", flush=True)
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description="Install native Codex binaries.")
@@ -172,20 +131,18 @@ def main() -> int:
workflow_id = workflow_url.rstrip("/").split("/")[-1]
print(f"Downloading native artifacts from workflow {workflow_id}...")
with _gha_group(f"Download native artifacts from workflow {workflow_id}"):
with tempfile.TemporaryDirectory(prefix="codex-native-artifacts-") as artifacts_dir_str:
artifacts_dir = Path(artifacts_dir_str)
_download_artifacts(workflow_id, artifacts_dir)
install_binary_components(
artifacts_dir,
vendor_dir,
[BINARY_COMPONENTS[name] for name in components if name in BINARY_COMPONENTS],
)
with tempfile.TemporaryDirectory(prefix="codex-native-artifacts-") as artifacts_dir_str:
artifacts_dir = Path(artifacts_dir_str)
_download_artifacts(workflow_id, artifacts_dir)
install_binary_components(
artifacts_dir,
vendor_dir,
[BINARY_COMPONENTS[name] for name in components if name in BINARY_COMPONENTS],
)
if "rg" in components:
with _gha_group("Fetch ripgrep binaries"):
print("Fetching ripgrep binaries...")
fetch_rg(vendor_dir, DEFAULT_RG_TARGETS, manifest_path=RG_MANIFEST)
print("Fetching ripgrep binaries...")
fetch_rg(vendor_dir, DEFAULT_RG_TARGETS, manifest_path=RG_MANIFEST)
print(f"Installed native dependencies into {vendor_dir}")
return 0
@@ -246,14 +203,7 @@ def fetch_rg(
for future in as_completed(future_map):
target = future_map[future]
try:
results[target] = future.result()
except Exception as exc:
_gha_error(
title="ripgrep install failed",
message=f"target={target} error={exc!r}",
)
raise RuntimeError(f"Failed to install ripgrep for target {target}.") from exc
results[target] = future.result()
print(f" installed ripgrep for {target}")
return [results[target] for target in targets]
@@ -351,8 +301,6 @@ def _fetch_single_rg(
url = providers[0]["url"]
archive_format = platform_info.get("format", "zst")
archive_member = platform_info.get("path")
digest = platform_info.get("digest")
expected_size = platform_info.get("size")
dest_dir = vendor_dir / target / "path"
dest_dir.mkdir(parents=True, exist_ok=True)
@@ -365,32 +313,10 @@ def _fetch_single_rg(
tmp_dir = Path(tmp_dir_str)
archive_filename = os.path.basename(urlparse(url).path)
download_path = tmp_dir / archive_filename
print(
f" downloading ripgrep for {target} ({platform_key}) from {url}",
flush=True,
)
try:
_download_file(url, download_path)
except Exception as exc:
_gha_error(
title="ripgrep download failed",
message=f"target={target} platform={platform_key} url={url} error={exc!r}",
)
raise RuntimeError(
"Failed to download ripgrep "
f"(target={target}, platform={platform_key}, format={archive_format}, "
f"expected_size={expected_size!r}, digest={digest!r}, url={url}, dest={download_path})."
) from exc
_download_file(url, download_path)
dest.unlink(missing_ok=True)
try:
extract_archive(download_path, archive_format, archive_member, dest)
except Exception as exc:
raise RuntimeError(
"Failed to extract ripgrep "
f"(target={target}, platform={platform_key}, format={archive_format}, "
f"member={archive_member!r}, url={url}, archive={download_path})."
) from exc
extract_archive(download_path, archive_format, archive_member, dest)
if not is_windows:
dest.chmod(0o755)
@@ -400,9 +326,7 @@ def _fetch_single_rg(
def _download_file(url: str, dest: Path) -> None:
dest.parent.mkdir(parents=True, exist_ok=True)
dest.unlink(missing_ok=True)
with urlopen(url, timeout=DOWNLOAD_TIMEOUT_SECS) as response, open(dest, "wb") as out:
with urlopen(url) as response, open(dest, "wb") as out:
shutil.copyfileobj(response, out)

341
codex-rs/Cargo.lock generated
View File

@@ -42,7 +42,7 @@ dependencies = [
"bitflags 2.10.0",
"bytes",
"bytestring",
"derive_more 2.1.1",
"derive_more 2.0.1",
"encoding_rs",
"foldhash 0.1.5",
"futures-core",
@@ -137,7 +137,7 @@ dependencies = [
"bytes",
"bytestring",
"cfg-if",
"derive_more 2.1.1",
"derive_more 2.0.1",
"encoding_rs",
"foldhash 0.1.5",
"futures-core",
@@ -329,12 +329,12 @@ name = "app_test_support"
version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"base64",
"chrono",
"codex-app-server-protocol",
"codex-core",
"codex-protocol",
"codex-utils-cargo-bin",
"core_test_support",
"serde",
"serde_json",
@@ -912,9 +912,9 @@ dependencies = [
[[package]]
name = "clap_complete"
version = "4.5.64"
version = "4.5.57"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "4c0da80818b2d95eca9aa614a30783e42f62bf5fdfee24e68cfb960b071ba8d1"
checksum = "4d9501bd3f5f09f7bbee01da9a511073ed30a80cd7a509f1214bb74eadea71ad"
dependencies = [
"clap",
]
@@ -993,6 +993,7 @@ version = "0.0.0"
dependencies = [
"anyhow",
"app_test_support",
"assert_cmd",
"base64",
"chrono",
"codex-app-server-protocol",
@@ -1063,7 +1064,6 @@ dependencies = [
"anyhow",
"assert_cmd",
"assert_matches",
"codex-utils-cargo-bin",
"pretty_assertions",
"similar",
"tempfile",
@@ -1161,7 +1161,6 @@ dependencies = [
"codex-tui",
"codex-tui2",
"codex-utils-absolute-path",
"codex-utils-cargo-bin",
"codex-windows-sandbox",
"ctor 0.5.0",
"libc",
@@ -1286,7 +1285,6 @@ dependencies = [
"codex-protocol",
"codex-rmcp-client",
"codex-utils-absolute-path",
"codex-utils-cargo-bin",
"codex-utils-pty",
"codex-utils-readiness",
"codex-utils-string",
@@ -1336,7 +1334,7 @@ dependencies = [
"tokio",
"tokio-util",
"toml 0.9.5",
"toml_edit 0.24.0+spec-1.1.0",
"toml_edit",
"tracing",
"tracing-subscriber",
"tracing-test",
@@ -1362,7 +1360,6 @@ dependencies = [
"codex-core",
"codex-protocol",
"codex-utils-absolute-path",
"codex-utils-cargo-bin",
"core_test_support",
"libc",
"mcp-types",
@@ -1388,11 +1385,11 @@ name = "codex-exec-server"
version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"async-trait",
"clap",
"codex-core",
"codex-execpolicy",
"codex-utils-cargo-bin",
"exec_server_test_support",
"libc",
"maplit",
@@ -1433,7 +1430,7 @@ dependencies = [
"allocative",
"anyhow",
"clap",
"derive_more 2.1.1",
"derive_more 2.0.1",
"env_logger",
"log",
"multimap",
@@ -1454,7 +1451,6 @@ dependencies = [
"codex-protocol",
"pretty_assertions",
"sentry",
"tracing",
"tracing-subscriber",
]
@@ -1552,6 +1548,7 @@ name = "codex-mcp-server"
version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"codex-arg0",
"codex-common",
"codex-core",
@@ -1610,6 +1607,7 @@ dependencies = [
"serde_json",
"strum_macros 0.27.2",
"tokio",
"tonic",
"tracing",
"tracing-opentelemetry",
"tracing-subscriber",
@@ -1674,8 +1672,8 @@ dependencies = [
"axum",
"codex-keyring-store",
"codex-protocol",
"codex-utils-cargo-bin",
"dirs",
"escargot",
"futures",
"keyring",
"mcp-types",
@@ -1702,7 +1700,6 @@ version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"codex-utils-cargo-bin",
"pretty_assertions",
"tempfile",
"uds_windows",
@@ -1732,7 +1729,7 @@ dependencies = [
"codex-windows-sandbox",
"color-eyre",
"crossterm",
"derive_more 2.1.1",
"derive_more 2.0.1",
"diffy",
"dirs",
"dunce",
@@ -1759,7 +1756,6 @@ dependencies = [
"supports-color 3.0.2",
"tempfile",
"textwrap 0.16.2",
"thiserror 2.0.17",
"tokio",
"tokio-stream",
"tokio-util",
@@ -1774,9 +1770,6 @@ dependencies = [
"url",
"uuid",
"vt100",
"which",
"windows-sys 0.52.0",
"winsplit",
]
[[package]]
@@ -1805,7 +1798,7 @@ dependencies = [
"codex-windows-sandbox",
"color-eyre",
"crossterm",
"derive_more 2.1.1",
"derive_more 2.0.1",
"diffy",
"dirs",
"dunce",
@@ -1820,7 +1813,6 @@ dependencies = [
"pulldown-cmark",
"rand 0.9.2",
"ratatui",
"ratatui-core",
"ratatui-macros",
"regex-lite",
"reqwest",
@@ -1842,7 +1834,6 @@ dependencies = [
"tracing-subscriber",
"tree-sitter-bash",
"tree-sitter-highlight",
"tui-scrollbar",
"unicode-segmentation",
"unicode-width 0.2.1",
"url",
@@ -1871,14 +1862,6 @@ dependencies = [
"tokio",
]
[[package]]
name = "codex-utils-cargo-bin"
version = "0.0.0"
dependencies = [
"assert_cmd",
"thiserror 2.0.17",
]
[[package]]
name = "codex-utils-image"
version = "0.0.0"
@@ -2007,20 +1990,6 @@ dependencies = [
"static_assertions",
]
[[package]]
name = "compact_str"
version = "0.9.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3fdb1325a1cece981e8a296ab8f0f9b63ae357bd0784a9faaf548cc7b480707a"
dependencies = [
"castaway",
"cfg-if",
"itoa",
"rustversion",
"ryu",
"static_assertions",
]
[[package]]
name = "concurrent-queue"
version = "2.5.0"
@@ -2042,18 +2011,6 @@ dependencies = [
"windows-sys 0.59.0",
]
[[package]]
name = "const-hex"
version = "1.17.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3bb320cac8a0750d7f25280aa97b09c26edfe161164238ecbbb31092b079e735"
dependencies = [
"cfg-if",
"cpufeatures",
"proptest",
"serde_core",
]
[[package]]
name = "convert_case"
version = "0.6.0"
@@ -2065,9 +2022,9 @@ dependencies = [
[[package]]
name = "convert_case"
version = "0.10.0"
version = "0.7.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "633458d4ef8c78b72454de2d54fd6ab2e60f9e02be22f3c6104cdc8a4e0fceb9"
checksum = "bb402b8d4c85569410425650ce3eddc7d698ed96d39a73f941b08fb63082f1e7"
dependencies = [
"unicode-segmentation",
]
@@ -2108,7 +2065,6 @@ dependencies = [
"codex-core",
"codex-protocol",
"codex-utils-absolute-path",
"codex-utils-cargo-bin",
"notify",
"pretty_assertions",
"regex-lite",
@@ -2454,11 +2410,11 @@ dependencies = [
[[package]]
name = "derive_more"
version = "2.1.1"
version = "2.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d751e9e49156b02b44f9c1815bcb94b984cdcc4396ecc32521c739452808b134"
checksum = "093242cf7570c207c83073cf82f79706fe7b8317e98620a47d5be7c3d8497678"
dependencies = [
"derive_more-impl 2.1.1",
"derive_more-impl 2.0.1",
]
[[package]]
@@ -2476,14 +2432,13 @@ dependencies = [
[[package]]
name = "derive_more-impl"
version = "2.1.1"
version = "2.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "799a97264921d8623a957f6c3b9011f3b5492f557bbb7a5a19b7fa6d06ba8dcb"
checksum = "bda628edc44c4bb645fbe0f758797143e4e07926f7ebf4e9bdfbd3d2ce621df3"
dependencies = [
"convert_case 0.10.0",
"convert_case 0.7.1",
"proc-macro2",
"quote",
"rustc_version",
"syn 2.0.104",
"unicode-xid",
]
@@ -2599,15 +2554,6 @@ version = "0.3.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "fea41bba32d969b513997752735605054bc0dfa92b4c56bf1189f2e174be7a10"
[[package]]
name = "document-features"
version = "0.2.12"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d4b8a88685455ed29a21542a33abd9cb6510b6b129abadabdcef0f4c55bc8f61"
dependencies = [
"litrs",
]
[[package]]
name = "dotenvy"
version = "0.15.7"
@@ -2838,8 +2784,8 @@ name = "exec_server_test_support"
version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"codex-core",
"codex-utils-cargo-bin",
"rmcp",
"serde_json",
"tokio",
@@ -3774,14 +3720,13 @@ dependencies = [
[[package]]
name = "insta"
version = "1.46.0"
version = "1.44.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1b66886d14d18d420ab5052cbff544fc5d34d0b2cdd35eb5976aaa10a4a472e5"
checksum = "b5c943d4415edd8153251b6f197de5eb1640e56d84e8d9159bea190421c73698"
dependencies = [
"console",
"once_cell",
"similar",
"tempfile",
]
[[package]]
@@ -3806,6 +3751,17 @@ dependencies = [
"rustversion",
]
[[package]]
name = "io-uring"
version = "0.7.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d93587f37623a1a17d94ef2bc9ada592f5465fe7732084ab7beefabe5c77c0c4"
dependencies = [
"bitflags 2.10.0",
"cfg-if",
"libc",
]
[[package]]
name = "ipnet"
version = "2.11.0"
@@ -3934,16 +3890,6 @@ dependencies = [
"wasm-bindgen",
]
[[package]]
name = "kasuari"
version = "0.4.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8fe90c1150662e858c7d5f945089b7517b0a80d8bf7ba4b1b5ffc984e7230a5b"
dependencies = [
"hashbrown 0.16.0",
"thiserror 2.0.17",
]
[[package]]
name = "keyring"
version = "3.6.3"
@@ -4089,12 +4035,6 @@ version = "0.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "241eaef5fd12c88705a01fc1066c48c4b36e0dd4377dcdc7ec3942cea7a69956"
[[package]]
name = "litrs"
version = "1.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "11d3d7f243d5c5a8b9bb5d6dd2b1602c0cb0b9db1621bafc7ed66e35ff9fe092"
[[package]]
name = "local-waker"
version = "0.1.4"
@@ -4213,9 +4153,9 @@ name = "mcp_test_support"
version = "0.0.0"
dependencies = [
"anyhow",
"assert_cmd",
"codex-core",
"codex-mcp-server",
"codex-utils-cargo-bin",
"core_test_support",
"mcp-types",
"os_info",
@@ -4716,9 +4656,9 @@ dependencies = [
[[package]]
name = "opentelemetry"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b84bcd6ae87133e903af7ef497404dda70c60d0ea14895fc8a5e6722754fc2a0"
checksum = "aaf416e4cb72756655126f7dd7bb0af49c674f4c1b9903e80c009e0c37e552e6"
dependencies = [
"futures-core",
"futures-sink",
@@ -4730,9 +4670,9 @@ dependencies = [
[[package]]
name = "opentelemetry-appender-tracing"
version = "0.31.1"
version = "0.30.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ef6a1ac5ca3accf562b8c306fa8483c85f4390f768185ab775f242f7fe8fdcc2"
checksum = "e68f63eca5fad47e570e00e893094fc17be959c80c79a7d6ec1abdd5ae6ffc16"
dependencies = [
"opentelemetry",
"tracing",
@@ -4742,9 +4682,9 @@ dependencies = [
[[package]]
name = "opentelemetry-http"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d7a6d09a73194e6b66df7c8f1b680f156d916a1a942abf2de06823dd02b7855d"
checksum = "50f6639e842a97dbea8886e3439710ae463120091e2e064518ba8e716e6ac36d"
dependencies = [
"async-trait",
"bytes",
@@ -4755,9 +4695,9 @@ dependencies = [
[[package]]
name = "opentelemetry-otlp"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7a2366db2dca4d2ad033cad11e6ee42844fd727007af5ad04a1730f4cb8163bf"
checksum = "dbee664a43e07615731afc539ca60c6d9f1a9425e25ca09c57bc36c87c55852b"
dependencies = [
"http 1.3.1",
"opentelemetry",
@@ -4775,32 +4715,30 @@ dependencies = [
[[package]]
name = "opentelemetry-proto"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a7175df06de5eaee9909d4805a3d07e28bb752c34cab57fa9cff549da596b30f"
checksum = "2e046fd7660710fe5a05e8748e70d9058dc15c94ba914e7c4faa7c728f0e8ddc"
dependencies = [
"base64",
"const-hex",
"hex",
"opentelemetry",
"opentelemetry_sdk",
"prost",
"serde",
"serde_json",
"tonic",
"tonic-prost",
]
[[package]]
name = "opentelemetry-semantic-conventions"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e62e29dfe041afb8ed2a6c9737ab57db4907285d999ef8ad3a59092a36bdc846"
checksum = "83d059a296a47436748557a353c5e6c5705b9470ef6c95cfc52c21a8814ddac2"
[[package]]
name = "opentelemetry_sdk"
version = "0.31.0"
version = "0.30.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e14ae4f5991976fd48df6d843de219ca6d31b01daaab2dad5af2badeded372bd"
checksum = "11f644aa9e5e31d11896e024305d7e3c98a88884d9f8919dbf37a9991bc47a4b"
dependencies = [
"futures-channel",
"futures-executor",
@@ -4808,6 +4746,7 @@ dependencies = [
"opentelemetry",
"percent-encoding",
"rand 0.9.2",
"serde_json",
"thiserror 2.0.17",
"tokio",
"tokio-stream",
@@ -5163,7 +5102,7 @@ version = "3.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "219cb19e96be00ab2e37d6e299658a0cfa83e52429179969b0f0121b4ac46983"
dependencies = [
"toml_edit 0.23.10+spec-1.0.0",
"toml_edit",
]
[[package]]
@@ -5189,26 +5128,11 @@ dependencies = [
"windows 0.61.3",
]
[[package]]
name = "proptest"
version = "1.9.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "bee689443a2bd0a16ab0348b52ee43e3b2d1b1f931c8aa5c9f8de4c86fbe8c40"
dependencies = [
"bitflags 2.10.0",
"num-traits",
"rand 0.9.2",
"rand_chacha 0.9.0",
"rand_xorshift",
"regex-syntax 0.8.5",
"unarray",
]
[[package]]
name = "prost"
version = "0.14.1"
version = "0.13.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7231bd9b3d3d33c86b58adbac74b5ec0ad9f496b19d22801d773636feaa95f3d"
checksum = "2796faa41db3ec313a31f7624d9286acf277b52de526150b7e69f3debf891ee5"
dependencies = [
"bytes",
"prost-derive",
@@ -5216,9 +5140,9 @@ dependencies = [
[[package]]
name = "prost-derive"
version = "0.14.1"
version = "0.13.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9120690fafc389a67ba3803df527d0ec9cbbc9cc45e4cc20b332996dfb672425"
checksum = "8a56d757972c98b346a9b766e3f02746cde6dd1cd1d1d563472929fdd74bec4d"
dependencies = [
"anyhow",
"itertools 0.14.0",
@@ -5418,15 +5342,6 @@ dependencies = [
"getrandom 0.3.3",
]
[[package]]
name = "rand_xorshift"
version = "0.4.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "513962919efc330f829edb2535844d1b912b0fbe2ca165d613e4e8788bb05a5a"
dependencies = [
"rand_core 0.9.3",
]
[[package]]
name = "ratatui"
version = "0.29.0"
@@ -5434,7 +5349,7 @@ source = "git+https://github.com/nornagon/ratatui?branch=nornagon-v0.29.0-patch#
dependencies = [
"bitflags 2.10.0",
"cassowary",
"compact_str 0.8.1",
"compact_str",
"crossterm",
"indoc",
"instability",
@@ -5443,27 +5358,7 @@ dependencies = [
"paste",
"strum 0.26.3",
"unicode-segmentation",
"unicode-truncate 1.1.0",
"unicode-width 0.2.1",
]
[[package]]
name = "ratatui-core"
version = "0.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5ef8dea09a92caaf73bff7adb70b76162e5937524058a7e5bff37869cbbec293"
dependencies = [
"bitflags 2.10.0",
"compact_str 0.9.0",
"hashbrown 0.16.0",
"indoc",
"itertools 0.14.0",
"kasuari",
"lru 0.16.2",
"strum 0.27.2",
"thiserror 2.0.17",
"unicode-segmentation",
"unicode-truncate 2.0.0",
"unicode-truncate",
"unicode-width 0.2.1",
]
@@ -5552,9 +5447,9 @@ dependencies = [
[[package]]
name = "regex-lite"
version = "0.1.8"
version = "0.1.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8d942b98df5e658f56f20d592c7f868833fe38115e65c33003d8cd224b0155da"
checksum = "943f41321c63ef1c92fd763bfe054d2668f7f225a5c29f0105903dc2fc04ba30"
[[package]]
name = "regex-syntax"
@@ -6630,9 +6525,6 @@ name = "strum"
version = "0.27.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "af23d6f6c1a224baef9d3f61e287d2761385a5b88fdab4eb4c6f11aeb54c4bcf"
dependencies = [
"strum_macros 0.27.2",
]
[[package]]
name = "strum_macros"
@@ -7024,26 +6916,29 @@ checksum = "1f3ccbac311fea05f86f61904b462b55fb3df8837a366dfc601a0161d0532f20"
[[package]]
name = "tokio"
version = "1.48.0"
version = "1.47.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ff360e02eab121e0bc37a2d3b4d4dc622e6eda3a8e5253d5435ecf5bd4c68408"
checksum = "89e49afdadebb872d3145a5638b59eb0691ea23e46ca484037cfab3b76b95038"
dependencies = [
"backtrace",
"bytes",
"io-uring",
"libc",
"mio",
"parking_lot",
"pin-project-lite",
"signal-hook-registry",
"slab",
"socket2 0.6.1",
"tokio-macros",
"windows-sys 0.61.1",
"windows-sys 0.59.0",
]
[[package]]
name = "tokio-macros"
version = "2.6.0"
version = "2.5.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "af407857209536a95c8e56f8231ef2c2e2aff839b22e07a1ffcbc617e9db9fa5"
checksum = "6e06d43f1345a3bcd39f6a56dbb7dcab2ba47e68e8ac134855e7e2bdbaf8cab8"
dependencies = [
"proc-macro2",
"quote",
@@ -7072,9 +6967,9 @@ dependencies = [
[[package]]
name = "tokio-stream"
version = "0.1.18"
version = "0.1.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "32da49809aab5c3bc678af03902d4ccddea2a87d028d86392a4b1560c6906c70"
checksum = "eca58d7bba4a75707817a2c44174253f9236b2d5fbd055602e9d5c07c139a047"
dependencies = [
"futures-core",
"pin-project-lite",
@@ -7136,30 +7031,18 @@ dependencies = [
[[package]]
name = "toml_datetime"
version = "0.7.5+spec-1.1.0"
version = "0.7.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "92e1cfed4a3038bc5a127e35a2d360f145e1f4b971b551a2ba5fd7aedf7e1347"
checksum = "f2cdb639ebbc97961c51720f858597f7f24c4fc295327923af55b74c3c724533"
dependencies = [
"serde_core",
]
[[package]]
name = "toml_edit"
version = "0.23.10+spec-1.0.0"
version = "0.23.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "84c8b9f757e028cee9fa244aea147aab2a9ec09d5325a9b01e0a49730c2b5269"
dependencies = [
"indexmap 2.12.0",
"toml_datetime",
"toml_parser",
"winnow",
]
[[package]]
name = "toml_edit"
version = "0.24.0+spec-1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8c740b185920170a6d9191122cafef7010bd6270a3824594bff6784c04d7f09e"
checksum = "6485ef6d0d9b5d0ec17244ff7eb05310113c3f316f2d14200d4de56b3cb98f8d"
dependencies = [
"indexmap 2.12.0",
"toml_datetime",
@@ -7170,28 +7053,30 @@ dependencies = [
[[package]]
name = "toml_parser"
version = "1.0.6+spec-1.1.0"
version = "1.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a3198b4b0a8e11f09dd03e133c0280504d0801269e9afa46362ffde1cbeebf44"
checksum = "c0cbe268d35bdb4bb5a56a2de88d0ad0eb70af5384a99d648cd4b3d04039800e"
dependencies = [
"winnow",
]
[[package]]
name = "toml_writer"
version = "1.0.6+spec-1.1.0"
version = "1.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ab16f14aed21ee8bfd8ec22513f7287cd4a91aa92e44edfe2c17ddd004e92607"
checksum = "df8b2b54733674ad286d16267dcfc7a71ed5c776e4ac7aa3c3e2561f7c637bf2"
[[package]]
name = "tonic"
version = "0.14.2"
version = "0.13.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eb7613188ce9f7df5bfe185db26c5814347d110db17920415cf2fbcad85e7203"
checksum = "7e581ba15a835f4d9ea06c55ab1bd4dce26fc53752c69a04aac00703bfb49ba9"
dependencies = [
"async-trait",
"axum",
"base64",
"bytes",
"h2",
"http 1.3.1",
"http-body",
"http-body-util",
@@ -7200,8 +7085,9 @@ dependencies = [
"hyper-util",
"percent-encoding",
"pin-project",
"prost",
"rustls-native-certs",
"sync_wrapper",
"socket2 0.5.10",
"tokio",
"tokio-rustls",
"tokio-stream",
@@ -7211,17 +7097,6 @@ dependencies = [
"tracing",
]
[[package]]
name = "tonic-prost"
version = "0.14.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "66bd50ad6ce1252d87ef024b3d64fe4c3cf54a86fb9ef4c631fdd0ded7aeaa67"
dependencies = [
"bytes",
"prost",
"tonic",
]
[[package]]
name = "tower"
version = "0.5.2"
@@ -7339,16 +7214,15 @@ dependencies = [
[[package]]
name = "tracing-opentelemetry"
version = "0.32.0"
version = "0.31.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1e6e5658463dd88089aba75c7791e1d3120633b1bfde22478b28f625a9bb1b8e"
checksum = "ddcf5959f39507d0d04d6413119c04f33b623f4f951ebcbdddddfad2d0623a9c"
dependencies = [
"js-sys",
"once_cell",
"opentelemetry",
"opentelemetry_sdk",
"rustversion",
"smallvec",
"thiserror 2.0.17",
"tracing",
"tracing-core",
"tracing-log",
@@ -7358,9 +7232,9 @@ dependencies = [
[[package]]
name = "tracing-subscriber"
version = "0.3.22"
version = "0.3.20"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2f30143827ddab0d256fd843b7a66d164e9f271cfa0dde49142c5ca0ca291f1e"
checksum = "2054a14f5307d601f88daf0553e1cbf472acc4f2c51afab632431cdcd72124d5"
dependencies = [
"matchers",
"nu-ansi-term",
@@ -7479,16 +7353,6 @@ dependencies = [
"termcolor",
]
[[package]]
name = "tui-scrollbar"
version = "0.2.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c42613099915b2e30e9f144670666e858e2538366f77742e1cf1c2f230efcacd"
dependencies = [
"document-features",
"ratatui-core",
]
[[package]]
name = "typenum"
version = "1.18.0"
@@ -7515,12 +7379,6 @@ dependencies = [
"libc",
]
[[package]]
name = "unarray"
version = "0.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eaea85b334db583fe3274d12b4cd1880032beab409c0d774be044d4480ab9a94"
[[package]]
name = "unicase"
version = "2.8.1"
@@ -7556,17 +7414,6 @@ dependencies = [
"unicode-width 0.1.14",
]
[[package]]
name = "unicode-truncate"
version = "2.0.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8fbf03860ff438702f3910ca5f28f8dac63c1c11e7efb5012b8b175493606330"
dependencies = [
"itertools 0.13.0",
"unicode-segmentation",
"unicode-width 0.2.1",
]
[[package]]
name = "unicode-width"
version = "0.1.14"
@@ -8532,12 +8379,6 @@ version = "0.0.19"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d135d17ab770252ad95e9a872d365cf3090e3be864a34ab46f48555993efc904"
[[package]]
name = "winsplit"
version = "0.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3ab703352da6a72f35c39a533526393725640575bb211f61987a2748323ad956"
[[package]]
name = "wiremock"
version = "0.6.5"

View File

@@ -36,7 +36,6 @@ members = [
"tui",
"tui2",
"utils/absolute-path",
"utils/cargo-bin",
"utils/git",
"utils/cache",
"utils/image",
@@ -50,7 +49,7 @@ members = [
resolver = "2"
[workspace.package]
version = "0.79.0-alpha.3"
version = "0.78.0-alpha.3"
# Track the edition for all workspace crates in one place. Individual
# crates can still override this value, but keeping it here means new
# crates created with `cargo new -w ...` automatically inherit the 2024
@@ -94,7 +93,6 @@ codex-tui = { path = "tui" }
codex-tui2 = { path = "tui2" }
codex-utils-absolute-path = { path = "utils/absolute-path" }
codex-utils-cache = { path = "utils/cache" }
codex-utils-cargo-bin = { path = "utils/cargo-bin" }
codex-utils-image = { path = "utils/image" }
codex-utils-json-to-toml = { path = "utils/json-to-toml" }
codex-utils-pty = { path = "utils/pty" }
@@ -145,7 +143,7 @@ ignore = "0.4.23"
image = { version = "^0.25.9", default-features = false }
include_dir = "0.7.4"
indexmap = "2.12.0"
insta = "1.46.0"
insta = "1.44.3"
itertools = "0.14.0"
keyring = { version = "3.6", default-features = false }
landlock = "0.4.4"
@@ -160,12 +158,12 @@ notify = "8.2.0"
nucleo-matcher = "0.3.1"
once_cell = "1.20.2"
openssl-sys = "*"
opentelemetry = "0.31.0"
opentelemetry-appender-tracing = "0.31.0"
opentelemetry-otlp = "0.31.0"
opentelemetry-semantic-conventions = "0.31.0"
opentelemetry_sdk = "0.31.0"
tracing-opentelemetry = "0.32.0"
opentelemetry = "0.30.0"
opentelemetry-appender-tracing = "0.30.0"
opentelemetry-otlp = "0.30.0"
opentelemetry-semantic-conventions = "0.30.0"
opentelemetry_sdk = "0.30.0"
tracing-opentelemetry = "0.31.0"
os_info = "3.12.0"
owo-colors = "4.2.0"
path-absolutize = "3.1.1"
@@ -176,10 +174,9 @@ pretty_assertions = "1.4.1"
pulldown-cmark = "0.10"
rand = "0.9"
ratatui = "0.29.0"
ratatui-core = "0.1.0"
ratatui-macros = "0.6.0"
regex = "1.12.2"
regex-lite = "0.1.8"
regex-lite = "0.1.7"
reqwest = "0.12"
rmcp = { version = "0.12.0", default-features = false }
schemars = "0.8.22"
@@ -207,20 +204,20 @@ thiserror = "2.0.17"
time = "0.3"
tiny_http = "0.12"
tokio = "1"
tokio-stream = "0.1.18"
tokio-stream = "0.1.17"
tokio-test = "0.4"
tokio-util = "0.7.16"
toml = "0.9.5"
toml_edit = "0.24.0"
toml_edit = "0.23.5"
tonic = "0.13.1"
tracing = "0.1.43"
tracing-appender = "0.2.3"
tracing-subscriber = "0.3.22"
tracing-subscriber = "0.3.20"
tracing-test = "0.2.5"
tree-sitter = "0.25.10"
tree-sitter-bash = "0.25"
tree-sitter-highlight = "0.25.10"
ts-rs = "11"
tui-scrollbar = "0.2.1"
uds_windows = "1.1.0"
unicode-segmentation = "1.12.0"
unicode-width = "0.2"

View File

@@ -15,8 +15,8 @@ You can also install via Homebrew (`brew install --cask codex`) or download a pl
## Documentation quickstart
- First run with Codex? Start with [`docs/getting-started.md`](../docs/getting-started.md) (links to the walkthrough for prompts, keyboard shortcuts, and session management).
- Want deeper control? See [`docs/config.md`](../docs/config.md) and [`docs/install.md`](../docs/install.md).
- First run with Codex? Follow the walkthrough in [`docs/getting-started.md`](../docs/getting-started.md) for prompts, keyboard shortcuts, and session management.
- Already shipping with Codex and want deeper control? Jump to [`docs/advanced.md`](../docs/advanced.md) and the configuration reference at [`docs/config.md`](../docs/config.md).
## What's new in the Rust CLI
@@ -30,7 +30,7 @@ Codex supports a rich set of configuration options. Note that the Rust CLI uses
#### MCP client
Codex CLI functions as an MCP client that allows the Codex CLI and IDE extension to connect to MCP servers on startup. See the [`configuration documentation`](../docs/config.md#connecting-to-mcp-servers) for details.
Codex CLI functions as an MCP client that allows the Codex CLI and IDE extension to connect to MCP servers on startup. See the [`configuration documentation`](../docs/config.md#mcp_servers) for details.
#### MCP server (experimental)

View File

@@ -113,10 +113,6 @@ client_request_definitions! {
params: v2::ThreadArchiveParams,
response: v2::ThreadArchiveResponse,
},
ThreadRollback => "thread/rollback" {
params: v2::ThreadRollbackParams,
response: v2::ThreadRollbackResponse,
},
ThreadList => "thread/list" {
params: v2::ThreadListParams,
response: v2::ThreadListResponse,

View File

@@ -6,7 +6,6 @@ use crate::protocol::v2::UserInput;
use codex_protocol::protocol::AgentReasoningEvent;
use codex_protocol::protocol::AgentReasoningRawContentEvent;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::ThreadRolledBackEvent;
use codex_protocol::protocol::TurnAbortedEvent;
use codex_protocol::protocol::UserMessageEvent;
@@ -58,7 +57,6 @@ impl ThreadHistoryBuilder {
EventMsg::TokenCount(_) => {}
EventMsg::EnteredReviewMode(_) => {}
EventMsg::ExitedReviewMode(_) => {}
EventMsg::ThreadRolledBack(payload) => self.handle_thread_rollback(payload),
EventMsg::UndoCompleted(_) => {}
EventMsg::TurnAborted(payload) => self.handle_turn_aborted(payload),
_ => {}
@@ -132,23 +130,6 @@ impl ThreadHistoryBuilder {
turn.status = TurnStatus::Interrupted;
}
fn handle_thread_rollback(&mut self, payload: &ThreadRolledBackEvent) {
self.finish_current_turn();
let n = usize::try_from(payload.num_turns).unwrap_or(usize::MAX);
if n >= self.turns.len() {
self.turns.clear();
} else {
self.turns.truncate(self.turns.len().saturating_sub(n));
}
// Re-number subsequent synthetic ids so the pruned history is consistent.
self.next_turn_index =
i64::try_from(self.turns.len().saturating_add(1)).unwrap_or(i64::MAX);
let item_count: usize = self.turns.iter().map(|t| t.items.len()).sum();
self.next_item_index = i64::try_from(item_count.saturating_add(1)).unwrap_or(i64::MAX);
}
fn finish_current_turn(&mut self) {
if let Some(turn) = self.current_turn.take() {
if turn.items.is_empty() {
@@ -232,7 +213,6 @@ mod tests {
use codex_protocol::protocol::AgentMessageEvent;
use codex_protocol::protocol::AgentReasoningEvent;
use codex_protocol::protocol::AgentReasoningRawContentEvent;
use codex_protocol::protocol::ThreadRolledBackEvent;
use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::TurnAbortedEvent;
use codex_protocol::protocol::UserMessageEvent;
@@ -430,95 +410,4 @@ mod tests {
}
);
}
#[test]
fn drops_last_turns_on_thread_rollback() {
let events = vec![
EventMsg::UserMessage(UserMessageEvent {
message: "First".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A1".into(),
}),
EventMsg::UserMessage(UserMessageEvent {
message: "Second".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A2".into(),
}),
EventMsg::ThreadRolledBack(ThreadRolledBackEvent { num_turns: 1 }),
EventMsg::UserMessage(UserMessageEvent {
message: "Third".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A3".into(),
}),
];
let turns = build_turns_from_event_msgs(&events);
let expected = vec![
Turn {
id: "turn-1".into(),
status: TurnStatus::Completed,
error: None,
items: vec![
ThreadItem::UserMessage {
id: "item-1".into(),
content: vec![UserInput::Text {
text: "First".into(),
}],
},
ThreadItem::AgentMessage {
id: "item-2".into(),
text: "A1".into(),
},
],
},
Turn {
id: "turn-2".into(),
status: TurnStatus::Completed,
error: None,
items: vec![
ThreadItem::UserMessage {
id: "item-3".into(),
content: vec![UserInput::Text {
text: "Third".into(),
}],
},
ThreadItem::AgentMessage {
id: "item-4".into(),
text: "A3".into(),
},
],
},
];
assert_eq!(turns, expected);
}
#[test]
fn thread_rollback_clears_all_turns_when_num_turns_exceeds_history() {
let events = vec![
EventMsg::UserMessage(UserMessageEvent {
message: "One".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A1".into(),
}),
EventMsg::UserMessage(UserMessageEvent {
message: "Two".into(),
images: None,
}),
EventMsg::AgentMessage(AgentMessageEvent {
message: "A2".into(),
}),
EventMsg::ThreadRolledBack(ThreadRolledBackEvent { num_turns: 99 }),
];
let turns = build_turns_from_event_msgs(&events);
assert_eq!(turns, Vec::<Turn>::new());
}
}

View File

@@ -384,8 +384,6 @@ pub struct SendUserTurnParams {
pub model: String,
pub effort: Option<ReasoningEffort>,
pub summary: ReasoningSummary,
/// Optional JSON Schema used to constrain the final assistant message for this turn.
pub output_schema: Option<serde_json::Value>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]

View File

@@ -89,7 +89,6 @@ pub enum CodexErrorInfo {
InternalServerError,
Unauthorized,
BadRequest,
ThreadRollbackFailed,
SandboxError,
/// The response SSE stream disconnected in the middle of a turn before completion.
ResponseStreamDisconnected {
@@ -120,7 +119,6 @@ impl From<CoreCodexErrorInfo> for CodexErrorInfo {
CoreCodexErrorInfo::InternalServerError => CodexErrorInfo::InternalServerError,
CoreCodexErrorInfo::Unauthorized => CodexErrorInfo::Unauthorized,
CoreCodexErrorInfo::BadRequest => CodexErrorInfo::BadRequest,
CoreCodexErrorInfo::ThreadRollbackFailed => CodexErrorInfo::ThreadRollbackFailed,
CoreCodexErrorInfo::SandboxError => CodexErrorInfo::SandboxError,
CoreCodexErrorInfo::ResponseStreamDisconnected { http_status_code } => {
CodexErrorInfo::ResponseStreamDisconnected { http_status_code }
@@ -229,8 +227,6 @@ pub enum ConfigLayerSource {
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
System {
/// This is the path to the system config.toml file, though it is not
/// guaranteed to exist.
file: AbsolutePathBuf,
},
@@ -241,8 +237,6 @@ pub enum ConfigLayerSource {
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
User {
/// This is the path to the user's config.toml file, though it is not
/// guaranteed to exist.
file: AbsolutePathBuf,
},
@@ -332,15 +326,6 @@ pub struct ProfileV2 {
pub additional: HashMap<String, JsonValue>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
pub struct AnalyticsConfig {
pub enabled: Option<bool>,
#[serde(default, flatten)]
pub additional: HashMap<String, JsonValue>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
@@ -365,7 +350,6 @@ pub struct Config {
pub model_reasoning_effort: Option<ReasoningEffort>,
pub model_reasoning_summary: Option<ReasoningSummary>,
pub model_verbosity: Option<Verbosity>,
pub analytics: Option<AnalyticsConfig>,
#[serde(default, flatten)]
pub additional: HashMap<String, JsonValue>,
}
@@ -1057,30 +1041,6 @@ pub struct ThreadArchiveParams {
#[ts(export_to = "v2/")]
pub struct ThreadArchiveResponse {}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ThreadRollbackParams {
pub thread_id: String,
/// The number of turns to drop from the end of the thread. Must be >= 1.
///
/// This only modifies the thread's history and does not revert local file changes
/// that have been made by the agent. Clients are responsible for reverting these changes.
pub num_turns: u32,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ThreadRollbackResponse {
/// The updated thread after applying the rollback, with `turns` populated.
///
/// The ThreadItems stored in each Turn are lossy since we explicitly do not
/// persist all agent interactions, such as command executions. This is the same
/// behavior as `thread/resume`.
pub thread: Thread,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -1219,7 +1179,7 @@ pub struct Thread {
pub source: SessionSource,
/// Optional Git metadata captured when the thread was created.
pub git_info: Option<GitInfo>,
/// Only populated on `thread/resume` and `thread/rollback` responses.
/// Only populated on a `thread/resume` response.
/// For all other responses and notifications returning a Thread,
/// the turns field will be an empty list.
pub turns: Vec<Turn>,
@@ -1310,8 +1270,6 @@ pub struct Turn {
pub struct TurnError {
pub message: String,
pub codex_error_info: Option<CodexErrorInfo>,
#[serde(default)]
pub additional_details: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1355,8 +1313,6 @@ pub struct TurnStartParams {
pub effort: Option<ReasoningEffort>,
/// Override the reasoning summary for this turn and subsequent turns.
pub summary: Option<ReasoningSummary>,
/// Optional JSON Schema used to constrain the final assistant message for this turn.
pub output_schema: Option<JsonValue>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]

View File

@@ -13,7 +13,6 @@ use std::time::Duration;
use anyhow::Context;
use anyhow::Result;
use anyhow::bail;
use clap::ArgAction;
use clap::Parser;
use clap::Subcommand;
use codex_app_server_protocol::AddConversationListenerParams;
@@ -36,8 +35,6 @@ use codex_app_server_protocol::JSONRPCRequest;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::LoginChatGptCompleteNotification;
use codex_app_server_protocol::LoginChatGptResponse;
use codex_app_server_protocol::ModelListParams;
use codex_app_server_protocol::ModelListResponse;
use codex_app_server_protocol::NewConversationParams;
use codex_app_server_protocol::NewConversationResponse;
use codex_app_server_protocol::RequestId;
@@ -68,19 +65,6 @@ struct Cli {
#[arg(long, env = "CODEX_BIN", default_value = "codex")]
codex_bin: String,
/// Forwarded to the `codex` CLI as `--config key=value`. Repeatable.
///
/// Example:
/// `--config 'model_providers.mock.base_url="http://localhost:4010/v2"'`
#[arg(
short = 'c',
long = "config",
value_name = "key=value",
action = ArgAction::Append,
global = true
)]
config_overrides: Vec<String>,
#[command(subcommand)]
command: CliCommand,
}
@@ -129,49 +113,32 @@ enum CliCommand {
TestLogin,
/// Fetch the current account rate limits from the Codex app-server.
GetAccountRateLimits,
/// List the available models from the Codex app-server.
#[command(name = "model-list")]
ModelList,
}
fn main() -> Result<()> {
let Cli {
codex_bin,
config_overrides,
command,
} = Cli::parse();
let Cli { codex_bin, command } = Cli::parse();
match command {
CliCommand::SendMessage { user_message } => {
send_message(&codex_bin, &config_overrides, user_message)
}
CliCommand::SendMessageV2 { user_message } => {
send_message_v2(&codex_bin, &config_overrides, user_message)
}
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, &config_overrides, user_message)
trigger_cmd_approval(codex_bin, user_message)
}
CliCommand::TriggerPatchApproval { user_message } => {
trigger_patch_approval(&codex_bin, &config_overrides, user_message)
trigger_patch_approval(codex_bin, user_message)
}
CliCommand::NoTriggerCmdApproval => no_trigger_cmd_approval(&codex_bin, &config_overrides),
CliCommand::NoTriggerCmdApproval => no_trigger_cmd_approval(codex_bin),
CliCommand::SendFollowUpV2 {
first_message,
follow_up_message,
} => send_follow_up_v2(
&codex_bin,
&config_overrides,
first_message,
follow_up_message,
),
CliCommand::TestLogin => test_login(&codex_bin, &config_overrides),
CliCommand::GetAccountRateLimits => get_account_rate_limits(&codex_bin, &config_overrides),
CliCommand::ModelList => model_list(&codex_bin, &config_overrides),
} => send_follow_up_v2(codex_bin, first_message, follow_up_message),
CliCommand::TestLogin => test_login(codex_bin),
CliCommand::GetAccountRateLimits => get_account_rate_limits(codex_bin),
}
}
fn send_message(codex_bin: &str, config_overrides: &[String], user_message: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
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:?}");
@@ -192,61 +159,46 @@ fn send_message(codex_bin: &str, config_overrides: &[String], user_message: Stri
Ok(())
}
fn send_message_v2(
codex_bin: &str,
config_overrides: &[String],
user_message: String,
) -> Result<()> {
send_message_v2_with_policies(codex_bin, config_overrides, user_message, None, None)
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: &str,
config_overrides: &[String],
user_message: Option<String>,
) -> Result<()> {
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,
config_overrides,
message,
Some(AskForApproval::OnRequest),
Some(SandboxPolicy::ReadOnly),
)
}
fn trigger_patch_approval(
codex_bin: &str,
config_overrides: &[String],
user_message: Option<String>,
) -> Result<()> {
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,
config_overrides,
message,
Some(AskForApproval::OnRequest),
Some(SandboxPolicy::ReadOnly),
)
}
fn no_trigger_cmd_approval(codex_bin: &str, config_overrides: &[String]) -> Result<()> {
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, config_overrides, prompt.to_string(), None, None)
send_message_v2_with_policies(codex_bin, prompt.to_string(), None, None)
}
fn send_message_v2_with_policies(
codex_bin: &str,
config_overrides: &[String],
codex_bin: String,
user_message: String,
approval_policy: Option<AskForApproval>,
sandbox_policy: Option<SandboxPolicy>,
) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
@@ -270,12 +222,11 @@ fn send_message_v2_with_policies(
}
fn send_follow_up_v2(
codex_bin: &str,
config_overrides: &[String],
codex_bin: String,
first_message: String,
follow_up_message: String,
) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
@@ -308,8 +259,8 @@ fn send_follow_up_v2(
Ok(())
}
fn test_login(codex_bin: &str, config_overrides: &[String]) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
fn test_login(codex_bin: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
@@ -338,8 +289,8 @@ fn test_login(codex_bin: &str, config_overrides: &[String]) -> Result<()> {
}
}
fn get_account_rate_limits(codex_bin: &str, config_overrides: &[String]) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
fn get_account_rate_limits(codex_bin: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
@@ -350,18 +301,6 @@ fn get_account_rate_limits(codex_bin: &str, config_overrides: &[String]) -> Resu
Ok(())
}
fn model_list(codex_bin: &str, config_overrides: &[String]) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin, config_overrides)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
let response = client.model_list(ModelListParams::default())?;
println!("< model/list response: {response:?}");
Ok(())
}
struct CodexClient {
child: Child,
stdin: Option<ChildStdin>,
@@ -370,12 +309,8 @@ struct CodexClient {
}
impl CodexClient {
fn spawn(codex_bin: &str, config_overrides: &[String]) -> Result<Self> {
let mut cmd = Command::new(codex_bin);
for override_kv in config_overrides {
cmd.arg("--config").arg(override_kv);
}
let mut codex_app_server = cmd
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())
@@ -517,16 +452,6 @@ impl CodexClient {
self.send_request(request, request_id, "account/rateLimits/read")
}
fn model_list(&mut self, params: ModelListParams) -> Result<ModelListResponse> {
let request_id = self.request_id();
let request = ClientRequest::ModelList {
request_id: request_id.clone(),
params,
};
self.send_request(request, request_id, "model/list")
}
fn stream_conversation(&mut self, conversation_id: &ConversationId) -> Result<()> {
loop {
let notification = self.next_notification()?;

View File

@@ -48,6 +48,7 @@ uuid = { workspace = true, features = ["serde", "v7"] }
[dev-dependencies]
app_test_support = { workspace = true }
assert_cmd = { workspace = true }
base64 = { workspace = true }
core_test_support = { workspace = true }
mcp-types = { workspace = true }

View File

@@ -72,7 +72,6 @@ Example (from OpenAI's official VSCode extension):
- `thread/resume` — reopen an existing thread by id so subsequent `turn/start` calls append to it.
- `thread/list` — page through stored rollouts; supports cursor-based pagination and optional `modelProviders` filtering.
- `thread/archive` — move a threads rollout file into the archived directory; returns `{}` on success.
- `thread/rollback` — drop the last N turns from the agents in-memory context and persist a rollback marker in the rollout so future resumes see the pruned history; returns the updated `thread` (with `turns` populated) on success.
- `turn/start` — add user input to a thread and begin Codex generation; responds with the initial `turn` object and streams `turn/started`, `item/*`, and `turn/completed` notifications.
- `turn/interrupt` — request cancellation of an in-flight turn by `(thread_id, turn_id)`; success is an empty `{}` response and the turn finishes with `status: "interrupted"`.
- `review/start` — kick off Codexs automated reviewer for a thread; responds like `turn/start` and emits `item/started`/`item/completed` notifications with `enteredReviewMode` and `exitedReviewMode` items, plus a final assistant `agentMessage` containing the review.
@@ -163,7 +162,7 @@ Turns attach user input (text or images) to a thread and trigger Codex generatio
- `{"type":"image","url":"https://…png"}`
- `{"type":"localImage","path":"/tmp/screenshot.png"}`
You can optionally specify config overrides on the new turn. If specified, these settings become the default for subsequent turns on the same thread. `outputSchema` applies only to the current turn.
You can optionally specify config overrides on the new turn. If specified, these settings become the default for subsequent turns on the same thread.
```json
{ "method": "turn/start", "id": 30, "params": {
@@ -179,14 +178,7 @@ You can optionally specify config overrides on the new turn. If specified, these
},
"model": "gpt-5.1-codex",
"effort": "medium",
"summary": "concise",
// Optional JSON Schema to constrain the final assistant message for this turn.
"outputSchema": {
"type": "object",
"properties": { "answer": { "type": "string" } },
"required": ["answer"],
"additionalProperties": false
}
"summary": "concise"
} }
{ "id": 30, "result": { "turn": {
"id": "turn_456",
@@ -310,7 +302,7 @@ Event notifications are the server-initiated event stream for thread lifecycles,
The app-server streams JSON-RPC notifications while a turn is running. Each turn starts with `turn/started` (initial `turn`) and ends with `turn/completed` (final `turn` status). Token usage events stream separately via `thread/tokenUsage/updated`. Clients subscribe to the events they care about, rendering each item incrementally as updates arrive. The per-item lifecycle is always: `item/started` → zero or more item-specific deltas → `item/completed`.
- `turn/started``{ turn }` with the turn id, empty `items`, and `status: "inProgress"`.
- `turn/completed``{ turn }` where `turn.status` is `completed`, `interrupted`, or `failed`; failures carry `{ error: { message, codexErrorInfo?, additionalDetails? } }`.
- `turn/completed``{ turn }` where `turn.status` is `completed`, `interrupted`, or `failed`; failures carry `{ error: { message, codexErrorInfo? } }`.
- `turn/diff/updated``{ threadId, turnId, diff }` represents the up-to-date snapshot of the turn-level unified diff, emitted after every FileChange item. `diff` is the latest aggregated unified diff across every file change in the turn. UIs can render this to show the full "what changed" view without stitching individual `fileChange` items.
- `turn/plan/updated``{ turnId, explanation?, plan }` whenever the agent shares or changes its plan; each `plan` entry is `{ step, status }` with `status` in `pending`, `inProgress`, or `completed`.
@@ -360,7 +352,7 @@ There are additional item-specific events:
### Errors
`error` event is emitted whenever the server hits an error mid-turn (for example, upstream model errors or quota limits). Carries the same `{ error: { message, codexErrorInfo?, additionalDetails? } }` payload as `turn.status: "failed"` and may precede that terminal notification.
`error` event is emitted whenever the server hits an error mid-turn (for example, upstream model errors or quota limits). Carries the same `{ error: { message, codexErrorInfo? } }` payload as `turn.status: "failed"` and may precede that terminal notification.
`codexErrorInfo` maps to the `CodexErrorInfo` enum. Common values:

View File

@@ -1,13 +1,7 @@
use crate::codex_message_processor::ApiVersion;
use crate::codex_message_processor::PendingInterrupts;
use crate::codex_message_processor::PendingRollbacks;
use crate::codex_message_processor::TurnSummary;
use crate::codex_message_processor::TurnSummaryStore;
use crate::codex_message_processor::read_event_msgs_from_rollout;
use crate::codex_message_processor::read_summary_from_rollout;
use crate::codex_message_processor::summary_to_thread;
use crate::error_code::INTERNAL_ERROR_CODE;
use crate::error_code::INVALID_REQUEST_ERROR_CODE;
use crate::outgoing_message::OutgoingMessageSender;
use codex_app_server_protocol::AccountRateLimitsUpdatedNotification;
use codex_app_server_protocol::AgentMessageDeltaNotification;
@@ -33,7 +27,6 @@ use codex_app_server_protocol::FileUpdateChange;
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::McpToolCallError;
use codex_app_server_protocol::McpToolCallResult;
use codex_app_server_protocol::McpToolCallStatus;
@@ -47,7 +40,6 @@ use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ServerRequestPayload;
use codex_app_server_protocol::TerminalInteractionNotification;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::ThreadRollbackResponse;
use codex_app_server_protocol::ThreadTokenUsage;
use codex_app_server_protocol::ThreadTokenUsageUpdatedNotification;
use codex_app_server_protocol::Turn;
@@ -58,11 +50,9 @@ use codex_app_server_protocol::TurnInterruptResponse;
use codex_app_server_protocol::TurnPlanStep;
use codex_app_server_protocol::TurnPlanUpdatedNotification;
use codex_app_server_protocol::TurnStatus;
use codex_app_server_protocol::build_turns_from_event_msgs;
use codex_core::CodexConversation;
use codex_core::parse_command::shlex_join;
use codex_core::protocol::ApplyPatchApprovalRequestEvent;
use codex_core::protocol::CodexErrorInfo as CoreCodexErrorInfo;
use codex_core::protocol::Event;
use codex_core::protocol::EventMsg;
use codex_core::protocol::ExecApprovalRequestEvent;
@@ -88,17 +78,14 @@ use tracing::error;
type JsonValue = serde_json::Value;
#[allow(clippy::too_many_arguments)]
pub(crate) async fn apply_bespoke_event_handling(
event: Event,
conversation_id: ConversationId,
conversation: Arc<CodexConversation>,
outgoing: Arc<OutgoingMessageSender>,
pending_interrupts: PendingInterrupts,
pending_rollbacks: PendingRollbacks,
turn_summary_store: TurnSummaryStore,
api_version: ApiVersion,
fallback_model_provider: String,
) {
let Event {
id: event_turn_id,
@@ -350,35 +337,14 @@ pub(crate) async fn apply_bespoke_event_handling(
.await;
}
EventMsg::Error(ev) => {
let message = ev.message.clone();
let codex_error_info = ev.codex_error_info.clone();
// If this error belongs to an in-flight `thread/rollback` request, fail that request
// (and clear pending state) so subsequent rollbacks are unblocked.
//
// Don't send a notification for this error.
if matches!(
codex_error_info,
Some(CoreCodexErrorInfo::ThreadRollbackFailed)
) {
return handle_thread_rollback_failed(
conversation_id,
message,
&pending_rollbacks,
&outgoing,
)
.await;
};
let turn_error = TurnError {
message: ev.message,
codex_error_info: ev.codex_error_info.map(V2CodexErrorInfo::from),
additional_details: None,
};
handle_error(conversation_id, turn_error.clone(), &turn_summary_store).await;
outgoing
.send_server_notification(ServerNotification::Error(ErrorNotification {
error: turn_error.clone(),
error: turn_error,
will_retry: false,
thread_id: conversation_id.to_string(),
turn_id: event_turn_id.clone(),
@@ -391,7 +357,6 @@ pub(crate) async fn apply_bespoke_event_handling(
let turn_error = TurnError {
message: ev.message,
codex_error_info: ev.codex_error_info.map(V2CodexErrorInfo::from),
additional_details: ev.additional_details,
};
outgoing
.send_server_notification(ServerNotification::Error(ErrorNotification {
@@ -723,58 +688,6 @@ pub(crate) async fn apply_bespoke_event_handling(
)
.await;
}
EventMsg::ThreadRolledBack(_rollback_event) => {
let pending = {
let mut map = pending_rollbacks.lock().await;
map.remove(&conversation_id)
};
if let Some(request_id) = pending {
let rollout_path = conversation.rollout_path();
let response = match read_summary_from_rollout(
rollout_path.as_path(),
fallback_model_provider.as_str(),
)
.await
{
Ok(summary) => {
let mut thread = summary_to_thread(summary);
match read_event_msgs_from_rollout(rollout_path.as_path()).await {
Ok(events) => {
thread.turns = build_turns_from_event_msgs(&events);
ThreadRollbackResponse { thread }
}
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!(
"failed to load rollout `{}`: {err}",
rollout_path.display()
),
data: None,
};
outgoing.send_error(request_id, error).await;
return;
}
}
}
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!(
"failed to load rollout `{}`: {err}",
rollout_path.display()
),
data: None,
};
outgoing.send_error(request_id, error).await;
return;
}
};
outgoing.send_response(request_id, response).await;
}
}
EventMsg::TurnDiff(turn_diff_event) => {
handle_turn_diff(
conversation_id,
@@ -991,31 +904,6 @@ async fn handle_turn_interrupted(
.await;
}
async fn handle_thread_rollback_failed(
conversation_id: ConversationId,
message: String,
pending_rollbacks: &PendingRollbacks,
outgoing: &OutgoingMessageSender,
) {
let pending_rollback = {
let mut map = pending_rollbacks.lock().await;
map.remove(&conversation_id)
};
if let Some(request_id) = pending_rollback {
outgoing
.send_error(
request_id,
JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: message.clone(),
data: None,
},
)
.await;
}
}
async fn handle_token_count_event(
conversation_id: ConversationId,
turn_id: String,
@@ -1452,7 +1340,6 @@ mod tests {
TurnError {
message: "boom".to_string(),
codex_error_info: Some(V2CodexErrorInfo::InternalServerError),
additional_details: None,
},
&turn_summary_store,
)
@@ -1464,7 +1351,6 @@ mod tests {
Some(TurnError {
message: "boom".to_string(),
codex_error_info: Some(V2CodexErrorInfo::InternalServerError),
additional_details: None,
})
);
Ok(())
@@ -1512,7 +1398,6 @@ mod tests {
TurnError {
message: "oops".to_string(),
codex_error_info: None,
additional_details: None,
},
&turn_summary_store,
)
@@ -1554,7 +1439,6 @@ mod tests {
TurnError {
message: "bad".to_string(),
codex_error_info: Some(V2CodexErrorInfo::Other),
additional_details: None,
},
&turn_summary_store,
)
@@ -1583,7 +1467,6 @@ mod tests {
Some(TurnError {
message: "bad".to_string(),
codex_error_info: Some(V2CodexErrorInfo::Other),
additional_details: None,
})
);
}
@@ -1808,7 +1691,6 @@ mod tests {
TurnError {
message: "a1".to_string(),
codex_error_info: Some(V2CodexErrorInfo::BadRequest),
additional_details: None,
},
&turn_summary_store,
)
@@ -1828,7 +1710,6 @@ mod tests {
TurnError {
message: "b1".to_string(),
codex_error_info: None,
additional_details: None,
},
&turn_summary_store,
)
@@ -1865,7 +1746,6 @@ mod tests {
Some(TurnError {
message: "a1".to_string(),
codex_error_info: Some(V2CodexErrorInfo::BadRequest),
additional_details: None,
})
);
}
@@ -1886,7 +1766,6 @@ mod tests {
Some(TurnError {
message: "b1".to_string(),
codex_error_info: None,
additional_details: None,
})
);
}

View File

@@ -91,7 +91,6 @@ use codex_app_server_protocol::ThreadListParams;
use codex_app_server_protocol::ThreadListResponse;
use codex_app_server_protocol::ThreadResumeParams;
use codex_app_server_protocol::ThreadResumeResponse;
use codex_app_server_protocol::ThreadRollbackParams;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::ThreadStartedNotification;
@@ -179,8 +178,6 @@ use uuid::Uuid;
type PendingInterruptQueue = Vec<(RequestId, ApiVersion)>;
pub(crate) type PendingInterrupts = Arc<Mutex<HashMap<ConversationId, PendingInterruptQueue>>>;
pub(crate) type PendingRollbacks = Arc<Mutex<HashMap<ConversationId, RequestId>>>;
/// Per-conversation accumulation of the latest states e.g. error message while a turn runs.
#[derive(Default, Clone)]
pub(crate) struct TurnSummary {
@@ -223,8 +220,6 @@ pub(crate) struct CodexMessageProcessor {
active_login: Arc<Mutex<Option<ActiveLogin>>>,
// Queue of pending interrupt requests per conversation. We reply when TurnAborted arrives.
pending_interrupts: PendingInterrupts,
// Queue of pending rollback requests per conversation. We reply when ThreadRollback arrives.
pending_rollbacks: PendingRollbacks,
turn_summary_store: TurnSummaryStore,
pending_fuzzy_searches: Arc<Mutex<HashMap<String, Arc<AtomicBool>>>>,
feedback: CodexFeedback,
@@ -280,7 +275,6 @@ impl CodexMessageProcessor {
conversation_listeners: HashMap::new(),
active_login: Arc::new(Mutex::new(None)),
pending_interrupts: Arc::new(Mutex::new(HashMap::new())),
pending_rollbacks: Arc::new(Mutex::new(HashMap::new())),
turn_summary_store: Arc::new(Mutex::new(HashMap::new())),
pending_fuzzy_searches: Arc::new(Mutex::new(HashMap::new())),
feedback,
@@ -371,9 +365,6 @@ impl CodexMessageProcessor {
ClientRequest::ThreadArchive { request_id, params } => {
self.thread_archive(request_id, params).await;
}
ClientRequest::ThreadRollback { request_id, params } => {
self.thread_rollback(request_id, params).await;
}
ClientRequest::ThreadList { request_id, params } => {
self.thread_list(request_id, params).await;
}
@@ -1112,7 +1103,7 @@ impl CodexMessageProcessor {
}
async fn get_user_saved_config(&self, request_id: RequestId) {
let service = ConfigService::new_with_defaults(self.config.codex_home.clone());
let service = ConfigService::new(self.config.codex_home.clone(), Vec::new());
let user_saved_config: UserSavedConfig = match service.load_user_saved_config().await {
Ok(config) => config,
Err(err) => {
@@ -1255,14 +1246,14 @@ impl CodexMessageProcessor {
cwd,
approval_policy,
sandbox: sandbox_mode,
config: request_overrides,
config: cli_overrides,
base_instructions,
developer_instructions,
compact_prompt,
include_apply_patch_tool,
} = params;
let typesafe_overrides = ConfigOverrides {
let overrides = ConfigOverrides {
model,
config_profile: profile,
cwd: cwd.clone().map(PathBuf::from),
@@ -1279,21 +1270,15 @@ impl CodexMessageProcessor {
// Persist windows sandbox feature.
// TODO: persist default config in general.
let mut request_overrides = request_overrides.unwrap_or_default();
let mut cli_overrides = cli_overrides.unwrap_or_default();
if cfg!(windows) && self.config.features.enabled(Feature::WindowsSandbox) {
request_overrides.insert(
cli_overrides.insert(
"features.experimental_windows_sandbox".to_string(),
serde_json::json!(true),
);
}
let config = match derive_config_from_params(
&self.cli_overrides,
Some(request_overrides),
typesafe_overrides,
)
.await
{
let config = match derive_config_from_params(overrides, Some(cli_overrides)).await {
Ok(config) => config,
Err(err) => {
let error = JSONRPCErrorError {
@@ -1333,7 +1318,7 @@ impl CodexMessageProcessor {
}
async fn thread_start(&mut self, request_id: RequestId, params: ThreadStartParams) {
let typesafe_overrides = self.build_thread_config_overrides(
let overrides = self.build_thread_config_overrides(
params.model,
params.model_provider,
params.cwd,
@@ -1343,21 +1328,18 @@ impl CodexMessageProcessor {
params.developer_instructions,
);
let config =
match derive_config_from_params(&self.cli_overrides, params.config, typesafe_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;
}
};
let config = match derive_config_from_params(overrides, params.config).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;
}
};
match self.conversation_manager.new_conversation(config).await {
Ok(new_conv) => {
@@ -1524,52 +1506,6 @@ impl CodexMessageProcessor {
}
}
async fn thread_rollback(&mut self, request_id: RequestId, params: ThreadRollbackParams) {
let ThreadRollbackParams {
thread_id,
num_turns,
} = params;
if num_turns == 0 {
self.send_invalid_request_error(request_id, "numTurns must be >= 1".to_string())
.await;
return;
}
let (conversation_id, conversation) =
match self.conversation_from_thread_id(&thread_id).await {
Ok(v) => v,
Err(error) => {
self.outgoing.send_error(request_id, error).await;
return;
}
};
{
let mut map = self.pending_rollbacks.lock().await;
if map.contains_key(&conversation_id) {
self.send_invalid_request_error(
request_id,
"rollback already in progress for this thread".to_string(),
)
.await;
return;
}
map.insert(conversation_id, request_id.clone());
}
if let Err(err) = conversation.submit(Op::ThreadRollback { num_turns }).await {
// No ThreadRollback event will arrive if an error occurs.
// Clean up and reply immediately.
let mut map = self.pending_rollbacks.lock().await;
map.remove(&conversation_id);
self.send_internal_error(request_id, format!("failed to start rollback: {err}"))
.await;
}
}
async fn thread_list(&self, request_id: RequestId, params: ThreadListParams) {
let ThreadListParams {
cursor,
@@ -1607,7 +1543,7 @@ impl CodexMessageProcessor {
cwd,
approval_policy,
sandbox,
config: request_overrides,
config: cli_overrides,
base_instructions,
developer_instructions,
} = params;
@@ -1617,12 +1553,12 @@ impl CodexMessageProcessor {
|| cwd.is_some()
|| approval_policy.is_some()
|| sandbox.is_some()
|| request_overrides.is_some()
|| cli_overrides.is_some()
|| base_instructions.is_some()
|| developer_instructions.is_some();
let config = if overrides_requested {
let typesafe_overrides = self.build_thread_config_overrides(
let overrides = self.build_thread_config_overrides(
model,
model_provider,
cwd,
@@ -1631,13 +1567,7 @@ impl CodexMessageProcessor {
base_instructions,
developer_instructions,
);
match derive_config_from_params(
&self.cli_overrides,
request_overrides,
typesafe_overrides,
)
.await
{
match derive_config_from_params(overrides, cli_overrides).await {
Ok(config) => config,
Err(err) => {
let error = JSONRPCErrorError {
@@ -2267,7 +2197,7 @@ impl CodexMessageProcessor {
cwd,
approval_policy,
sandbox: sandbox_mode,
config: request_overrides,
config: cli_overrides,
base_instructions,
developer_instructions,
compact_prompt,
@@ -2275,15 +2205,15 @@ impl CodexMessageProcessor {
} = overrides;
// Persist windows sandbox feature.
let mut request_overrides = request_overrides.unwrap_or_default();
let mut cli_overrides = cli_overrides.unwrap_or_default();
if cfg!(windows) && self.config.features.enabled(Feature::WindowsSandbox) {
request_overrides.insert(
cli_overrides.insert(
"features.experimental_windows_sandbox".to_string(),
serde_json::json!(true),
);
}
let typesafe_overrides = ConfigOverrides {
let overrides = ConfigOverrides {
model,
config_profile: profile,
cwd: cwd.map(PathBuf::from),
@@ -2298,12 +2228,7 @@ impl CodexMessageProcessor {
..Default::default()
};
derive_config_from_params(
&self.cli_overrides,
Some(request_overrides),
typesafe_overrides,
)
.await
derive_config_from_params(overrides, Some(cli_overrides)).await
}
None => Ok(self.config.as_ref().clone()),
};
@@ -2654,7 +2579,6 @@ impl CodexMessageProcessor {
let _ = conversation
.submit(Op::UserInput {
items: mapped_items,
final_output_json_schema: None,
})
.await;
@@ -2674,7 +2598,6 @@ impl CodexMessageProcessor {
model,
effort,
summary,
output_schema,
} = params;
let Ok(conversation) = self
@@ -2709,7 +2632,7 @@ impl CodexMessageProcessor {
model,
effort,
summary,
final_output_json_schema: output_schema,
final_output_json_schema: None,
})
.await;
@@ -2727,17 +2650,19 @@ impl CodexMessageProcessor {
};
let skills_manager = self.conversation_manager.skills_manager();
let mut data = Vec::new();
for cwd in cwds {
let outcome = skills_manager.skills_for_cwd(&cwd, force_reload).await;
let errors = errors_to_info(&outcome.errors);
let skills = skills_to_info(&outcome.skills);
data.push(codex_app_server_protocol::SkillsListEntry {
cwd,
skills,
errors,
});
}
let data = cwds
.into_iter()
.map(|cwd| {
let outcome = skills_manager.skills_for_cwd_with_options(&cwd, force_reload);
let errors = errors_to_info(&outcome.errors);
let skills = skills_to_info(&outcome.skills);
codex_app_server_protocol::SkillsListEntry {
cwd,
skills,
errors,
}
})
.collect();
self.outgoing
.send_response(request_id, SkillsListResponse { data })
.await;
@@ -2816,7 +2741,6 @@ impl CodexMessageProcessor {
let turn_id = conversation
.submit(Op::UserInput {
items: mapped_items,
final_output_json_schema: params.output_schema,
})
.await;
@@ -3170,10 +3094,8 @@ impl CodexMessageProcessor {
let outgoing_for_task = self.outgoing.clone();
let pending_interrupts = self.pending_interrupts.clone();
let pending_rollbacks = self.pending_rollbacks.clone();
let turn_summary_store = self.turn_summary_store.clone();
let api_version_for_task = api_version;
let fallback_model_provider = self.config.model_provider_id.clone();
tokio::spawn(async move {
loop {
tokio::select! {
@@ -3229,10 +3151,8 @@ impl CodexMessageProcessor {
conversation.clone(),
outgoing_for_task.clone(),
pending_interrupts.clone(),
pending_rollbacks.clone(),
turn_summary_store.clone(),
api_version_for_task,
fallback_model_provider.clone(),
)
.await;
}
@@ -3420,37 +3340,20 @@ fn errors_to_info(
.collect()
}
/// Derive the effective [`Config`] by layering three override sources.
///
/// Precedence (lowest to highest):
/// - `cli_overrides`: process-wide startup `--config` flags.
/// - `request_overrides`: per-request dotted-path overrides (`params.config`), converted JSON->TOML.
/// - `typesafe_overrides`: Request objects such as `NewConversationParams` and
/// `ThreadStartParams` support a limited set of _explicit_ config overrides, so
/// `typesafe_overrides` is a `ConfigOverrides` derived from the respective request object.
/// Because the overrides are defined explicitly in the `*Params`, this takes priority over
/// the more general "bag of config options" provided by `cli_overrides` and `request_overrides`.
async fn derive_config_from_params(
cli_overrides: &[(String, TomlValue)],
request_overrides: Option<HashMap<String, serde_json::Value>>,
typesafe_overrides: ConfigOverrides,
overrides: ConfigOverrides,
cli_overrides: Option<HashMap<String, serde_json::Value>>,
) -> std::io::Result<Config> {
let merged_cli_overrides = cli_overrides
.iter()
.cloned()
.chain(
request_overrides
.unwrap_or_default()
.into_iter()
.map(|(k, v)| (k, json_to_toml(v))),
)
.collect::<Vec<_>>();
let cli_overrides = cli_overrides
.unwrap_or_default()
.into_iter()
.map(|(k, v)| (k, json_to_toml(v)))
.collect();
Config::load_with_cli_overrides_and_harness_overrides(merged_cli_overrides, typesafe_overrides)
.await
Config::load_with_cli_overrides_and_harness_overrides(cli_overrides, overrides).await
}
pub(crate) async fn read_summary_from_rollout(
async fn read_summary_from_rollout(
path: &Path,
fallback_provider: &str,
) -> std::io::Result<ConversationSummary> {
@@ -3509,24 +3412,6 @@ pub(crate) async fn read_summary_from_rollout(
})
}
pub(crate) async fn read_event_msgs_from_rollout(
path: &Path,
) -> std::io::Result<Vec<codex_protocol::protocol::EventMsg>> {
let items = match RolloutRecorder::get_rollout_history(path).await? {
InitialHistory::New => Vec::new(),
InitialHistory::Forked(items) => items,
InitialHistory::Resumed(resumed) => resumed.history,
};
Ok(items
.into_iter()
.filter_map(|item| match item {
RolloutItem::EventMsg(event) => Some(event),
_ => None,
})
.collect())
}
fn extract_conversation_summary(
path: PathBuf,
head: &[serde_json::Value],
@@ -3588,7 +3473,7 @@ fn parse_datetime(timestamp: Option<&str>) -> Option<DateTime<Utc>> {
})
}
pub(crate) fn summary_to_thread(summary: ConversationSummary) -> Thread {
fn summary_to_thread(summary: ConversationSummary) -> Thread {
let ConversationSummary {
conversation_id,
path,

View File

@@ -9,7 +9,6 @@ use codex_app_server_protocol::ConfigWriteResponse;
use codex_app_server_protocol::JSONRPCErrorError;
use codex_core::config::ConfigService;
use codex_core::config::ConfigServiceError;
use codex_core::config_loader::LoaderOverrides;
use serde_json::json;
use std::path::PathBuf;
use toml::Value as TomlValue;
@@ -20,13 +19,9 @@ pub(crate) struct ConfigApi {
}
impl ConfigApi {
pub(crate) fn new(
codex_home: PathBuf,
cli_overrides: Vec<(String, TomlValue)>,
loader_overrides: LoaderOverrides,
) -> Self {
pub(crate) fn new(codex_home: PathBuf, cli_overrides: Vec<(String, TomlValue)>) -> Self {
Self {
service: ConfigService::new(codex_home, cli_overrides, loader_overrides),
service: ConfigService::new(codex_home, cli_overrides),
}
}

View File

@@ -1,8 +1,7 @@
#![deny(clippy::print_stdout, clippy::print_stderr)]
use codex_common::CliConfigOverrides;
use codex_core::config::ConfigBuilder;
use codex_core::config_loader::LoaderOverrides;
use codex_core::config::Config;
use std::io::ErrorKind;
use std::io::Result as IoResult;
use std::path::PathBuf;
@@ -18,11 +17,13 @@ use tokio::io::BufReader;
use tokio::io::{self};
use tokio::sync::mpsc;
use toml::Value as TomlValue;
use tracing::Level;
use tracing::debug;
use tracing::error;
use tracing::info;
use tracing_subscriber::EnvFilter;
use tracing_subscriber::Layer;
use tracing_subscriber::filter::Targets;
use tracing_subscriber::layer::SubscriberExt;
use tracing_subscriber::util::SubscriberInitExt;
@@ -43,7 +44,6 @@ const CHANNEL_CAPACITY: usize = 128;
pub async fn run_main(
codex_linux_sandbox_exe: Option<PathBuf>,
cli_config_overrides: CliConfigOverrides,
loader_overrides: LoaderOverrides,
) -> IoResult<()> {
// Set up channels.
let (incoming_tx, mut incoming_rx) = mpsc::channel::<JSONRPCMessage>(CHANNEL_CAPACITY);
@@ -80,11 +80,7 @@ pub async fn run_main(
format!("error parsing -c overrides: {e}"),
)
})?;
let loader_overrides_for_config_api = loader_overrides.clone();
let config = ConfigBuilder::default()
.cli_overrides(cli_kv_overrides.clone())
.loader_overrides(loader_overrides)
.build()
let config = Config::load_with_cli_overrides(cli_kv_overrides.clone())
.await
.map_err(|e| {
std::io::Error::new(ErrorKind::InvalidData, format!("error loading config: {e}"))
@@ -107,8 +103,11 @@ pub async fn run_main(
.with_span_events(tracing_subscriber::fmt::format::FmtSpan::FULL)
.with_filter(EnvFilter::from_default_env());
let feedback_layer = feedback.logger_layer();
let feedback_metadata_layer = feedback.metadata_layer();
let feedback_layer = tracing_subscriber::fmt::layer()
.with_writer(feedback.make_writer())
.with_ansi(false)
.with_target(false)
.with_filter(Targets::new().with_default(Level::TRACE));
let otel_logger_layer = otel.as_ref().and_then(|o| o.logger_layer());
@@ -117,7 +116,6 @@ pub async fn run_main(
let _ = tracing_subscriber::registry()
.with(stderr_fmt)
.with(feedback_layer)
.with(feedback_metadata_layer)
.with(otel_logger_layer)
.with(otel_tracing_layer)
.try_init();
@@ -126,13 +124,11 @@ pub async fn run_main(
let processor_handle = tokio::spawn({
let outgoing_message_sender = OutgoingMessageSender::new(outgoing_tx);
let cli_overrides: Vec<(String, TomlValue)> = cli_kv_overrides.clone();
let loader_overrides = loader_overrides_for_config_api;
let mut processor = MessageProcessor::new(
outgoing_message_sender,
codex_linux_sandbox_exe,
std::sync::Arc::new(config),
cli_overrides,
loader_overrides,
feedback.clone(),
);
async move {

View File

@@ -1,42 +1,10 @@
use codex_app_server::run_main;
use codex_arg0::arg0_dispatch_or_else;
use codex_common::CliConfigOverrides;
use codex_core::config_loader::LoaderOverrides;
use std::path::PathBuf;
// Debug-only test hook: lets integration tests point the server at a temporary
// managed config file without writing to /etc.
const MANAGED_CONFIG_PATH_ENV_VAR: &str = "CODEX_APP_SERVER_MANAGED_CONFIG_PATH";
fn main() -> anyhow::Result<()> {
arg0_dispatch_or_else(|codex_linux_sandbox_exe| async move {
let managed_config_path = managed_config_path_from_debug_env();
let loader_overrides = LoaderOverrides {
managed_config_path,
..Default::default()
};
run_main(
codex_linux_sandbox_exe,
CliConfigOverrides::default(),
loader_overrides,
)
.await?;
run_main(codex_linux_sandbox_exe, CliConfigOverrides::default()).await?;
Ok(())
})
}
fn managed_config_path_from_debug_env() -> Option<PathBuf> {
#[cfg(debug_assertions)]
{
if let Ok(value) = std::env::var(MANAGED_CONFIG_PATH_ENV_VAR) {
return if value.is_empty() {
None
} else {
Some(PathBuf::from(value))
};
}
}
None
}

View File

@@ -20,7 +20,6 @@ use codex_app_server_protocol::RequestId;
use codex_core::AuthManager;
use codex_core::ConversationManager;
use codex_core::config::Config;
use codex_core::config_loader::LoaderOverrides;
use codex_core::default_client::USER_AGENT_SUFFIX;
use codex_core::default_client::get_codex_user_agent;
use codex_feedback::CodexFeedback;
@@ -42,7 +41,6 @@ impl MessageProcessor {
codex_linux_sandbox_exe: Option<PathBuf>,
config: Arc<Config>,
cli_overrides: Vec<(String, TomlValue)>,
loader_overrides: LoaderOverrides,
feedback: CodexFeedback,
) -> Self {
let outgoing = Arc::new(outgoing);
@@ -64,7 +62,7 @@ impl MessageProcessor {
cli_overrides.clone(),
feedback,
);
let config_api = ConfigApi::new(config.codex_home.clone(), cli_overrides, loader_overrides);
let config_api = ConfigApi::new(config.codex_home.clone(), cli_overrides);
Self {
outgoing,

View File

@@ -9,12 +9,12 @@ path = "lib.rs"
[dependencies]
anyhow = { workspace = true }
assert_cmd = { workspace = true }
base64 = { workspace = true }
chrono = { workspace = true }
codex-app-server-protocol = { workspace = true }
codex-core = { workspace = true, features = ["test-support"] }
codex-protocol = { workspace = true }
codex-utils-cargo-bin = { workspace = true }
serde = { workspace = true }
serde_json = { workspace = true }
tokio = { workspace = true, features = [

View File

@@ -11,6 +11,7 @@ use tokio::process::ChildStdin;
use tokio::process::ChildStdout;
use anyhow::Context;
use assert_cmd::prelude::*;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::ArchiveConversationParams;
use codex_app_server_protocol::CancelLoginAccountParams;
@@ -45,10 +46,10 @@ use codex_app_server_protocol::SetDefaultModelParams;
use codex_app_server_protocol::ThreadArchiveParams;
use codex_app_server_protocol::ThreadListParams;
use codex_app_server_protocol::ThreadResumeParams;
use codex_app_server_protocol::ThreadRollbackParams;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::TurnInterruptParams;
use codex_app_server_protocol::TurnStartParams;
use std::process::Command as StdCommand;
use tokio::process::Command;
pub struct McpProcess {
@@ -77,8 +78,12 @@ impl McpProcess {
codex_home: &Path,
env_overrides: &[(&str, Option<&str>)],
) -> anyhow::Result<Self> {
let program = codex_utils_cargo_bin::cargo_bin("codex-app-server")
.context("should find binary for codex-app-server")?;
// Use assert_cmd to locate the binary path and then switch to tokio::process::Command
let std_cmd = StdCommand::cargo_bin("codex-app-server")
.context("should find binary for codex-mcp-server")?;
let program = std_cmd.get_program().to_owned();
let mut cmd = Command::new(program);
cmd.stdin(Stdio::piped());
@@ -317,15 +322,6 @@ impl McpProcess {
self.send_request("thread/archive", params).await
}
/// Send a `thread/rollback` JSON-RPC request.
pub async fn send_thread_rollback_request(
&mut self,
params: ThreadRollbackParams,
) -> anyhow::Result<i64> {
let params = Some(serde_json::to_value(params)?);
self.send_request("thread/rollback", params).await
}
/// Send a `thread/list` JSON-RPC request.
pub async fn send_thread_list_request(
&mut self,

View File

@@ -5,6 +5,7 @@ use codex_protocol::openai_models::ConfigShellToolType;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ModelPreset;
use codex_protocol::openai_models::ModelVisibility;
use codex_protocol::openai_models::ReasoningSummaryFormat;
use codex_protocol::openai_models::TruncationPolicyConfig;
use serde_json::json;
use std::path::Path;
@@ -34,6 +35,7 @@ fn preset_to_info(preset: &ModelPreset, priority: i32) -> ModelInfo {
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
context_window: None,
reasoning_summary_format: ReasoningSummaryFormat::None,
experimental_supported_tools: Vec::new(),
}
}

View File

@@ -305,7 +305,6 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
model: "mock-model".to_string(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: None,
})
.await?;
// Acknowledge sendUserTurn
@@ -419,7 +418,6 @@ async fn test_send_user_turn_updates_sandbox_and_cwd_between_turns() -> Result<(
model: model.clone(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: None,
})
.await?;
timeout(
@@ -445,7 +443,6 @@ async fn test_send_user_turn_updates_sandbox_and_cwd_between_turns() -> Result<(
model: model.clone(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: None,
})
.await?;
timeout(

View File

@@ -7,7 +7,6 @@ mod fuzzy_file_search;
mod interrupt;
mod list_resume;
mod login;
mod output_schema;
mod send_message;
mod set_default_model;
mod user_agent;

View File

@@ -1,282 +0,0 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::to_response;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::InputItem;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::NewConversationParams;
use codex_app_server_protocol::NewConversationResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::SendUserTurnParams;
use codex_app_server_protocol::SendUserTurnResponse;
use codex_core::protocol::AskForApproval;
use codex_core::protocol::SandboxPolicy;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::openai_models::ReasoningEffort;
use core_test_support::responses;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
use std::path::Path;
use tempfile::TempDir;
use tokio::time::timeout;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
#[tokio::test]
async fn send_user_turn_accepts_output_schema_v1() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let response_mock = responses::mount_sse_once(&server, body).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 new_conv_id = mcp
.send_new_conversation_request(NewConversationParams {
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(new_conv_id)),
)
.await??;
let NewConversationResponse {
conversation_id, ..
} = to_response::<NewConversationResponse>(new_conv_resp)?;
let listener_id = mcp
.send_add_conversation_listener_request(AddConversationListenerParams {
conversation_id,
experimental_raw_events: false,
})
.await?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(listener_id)),
)
.await??;
let output_schema = serde_json::json!({
"type": "object",
"properties": {
"answer": { "type": "string" }
},
"required": ["answer"],
"additionalProperties": false
});
let send_turn_id = mcp
.send_send_user_turn_request(SendUserTurnParams {
conversation_id,
items: vec![InputItem::Text {
text: "Hello".to_string(),
}],
cwd: codex_home.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::new_read_only_policy(),
model: "mock-model".to_string(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: Some(output_schema.clone()),
})
.await?;
let _send_turn_resp: SendUserTurnResponse = to_response::<SendUserTurnResponse>(
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(send_turn_id)),
)
.await??,
)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/task_complete"),
)
.await??;
let request = response_mock.single_request();
let payload = request.body_json();
let text = payload.get("text").expect("request missing text field");
let format = text
.get("format")
.expect("request missing text.format field");
assert_eq!(
format,
&serde_json::json!({
"name": "codex_output_schema",
"type": "json_schema",
"strict": true,
"schema": output_schema,
})
);
Ok(())
}
#[tokio::test]
async fn send_user_turn_output_schema_is_per_turn_v1() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = responses::start_mock_server().await;
let body1 = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let response_mock1 = responses::mount_sse_once(&server, body1).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 new_conv_id = mcp
.send_new_conversation_request(NewConversationParams {
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(new_conv_id)),
)
.await??;
let NewConversationResponse {
conversation_id, ..
} = to_response::<NewConversationResponse>(new_conv_resp)?;
let listener_id = mcp
.send_add_conversation_listener_request(AddConversationListenerParams {
conversation_id,
experimental_raw_events: false,
})
.await?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(listener_id)),
)
.await??;
let output_schema = serde_json::json!({
"type": "object",
"properties": {
"answer": { "type": "string" }
},
"required": ["answer"],
"additionalProperties": false
});
let send_turn_id = mcp
.send_send_user_turn_request(SendUserTurnParams {
conversation_id,
items: vec![InputItem::Text {
text: "Hello".to_string(),
}],
cwd: codex_home.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::new_read_only_policy(),
model: "mock-model".to_string(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: Some(output_schema.clone()),
})
.await?;
let _send_turn_resp: SendUserTurnResponse = to_response::<SendUserTurnResponse>(
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(send_turn_id)),
)
.await??,
)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/task_complete"),
)
.await??;
let payload1 = response_mock1.single_request().body_json();
assert_eq!(
payload1.pointer("/text/format"),
Some(&serde_json::json!({
"name": "codex_output_schema",
"type": "json_schema",
"strict": true,
"schema": output_schema,
}))
);
let body2 = responses::sse(vec![
responses::ev_response_created("resp-2"),
responses::ev_assistant_message("msg-2", "Done"),
responses::ev_completed("resp-2"),
]);
let response_mock2 = responses::mount_sse_once(&server, body2).await;
let send_turn_id_2 = mcp
.send_send_user_turn_request(SendUserTurnParams {
conversation_id,
items: vec![InputItem::Text {
text: "Hello again".to_string(),
}],
cwd: codex_home.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::new_read_only_policy(),
model: "mock-model".to_string(),
effort: Some(ReasoningEffort::Medium),
summary: ReasoningSummary::Auto,
output_schema: None,
})
.await?;
let _send_turn_resp_2: SendUserTurnResponse = to_response::<SendUserTurnResponse>(
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(send_turn_id_2)),
)
.await??,
)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/task_complete"),
)
.await??;
let payload2 = response_mock2.single_request().body_json();
assert_eq!(payload2.pointer("/text/format"), None);
Ok(())
}
fn create_config_toml(codex_home: &Path, server_uri: &str) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
std::fs::write(
config_toml,
format!(
r#"
model = "mock-model"
approval_policy = "never"
sandbox_mode = "read-only"
model_provider = "mock_provider"
[model_providers.mock_provider]
name = "Mock provider for test"
base_url = "{server_uri}/v1"
wire_api = "responses"
request_max_retries = 0
stream_max_retries = 0
"#
),
)
}

View File

@@ -18,7 +18,6 @@ use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::SandboxMode;
use codex_app_server_protocol::ToolsV2;
use codex_app_server_protocol::WriteStatus;
use codex_core::config_loader::SYSTEM_CONFIG_TOML_FILE_UNIX;
use codex_utils_absolute_path::AbsolutePathBuf;
use pretty_assertions::assert_eq;
use serde_json::json;
@@ -74,7 +73,8 @@ sandbox_mode = "workspace-write"
}
);
let layers = layers.expect("layers present");
assert_layers_user_then_optional_system(&layers, user_file)?;
assert_eq!(layers.len(), 1);
assert_eq!(layers[0].name, ConfigLayerSource::User { file: user_file });
Ok(())
}
@@ -136,7 +136,8 @@ view_image = false
);
let layers = layers.expect("layers present");
assert_layers_user_then_optional_system(&layers, user_file)?;
assert_eq!(layers.len(), 1);
assert_eq!(layers[0].name, ConfigLayerSource::User { file: user_file });
Ok(())
}
@@ -184,10 +185,7 @@ writable_roots = [{}]
let mut mcp = McpProcess::new_with_env(
codex_home.path(),
&[(
"CODEX_APP_SERVER_MANAGED_CONFIG_PATH",
Some(&managed_path_str),
)],
&[("CODEX_MANAGED_CONFIG_PATH", Some(&managed_path_str))],
)
.await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
@@ -259,7 +257,12 @@ writable_roots = [{}]
);
let layers = layers.expect("layers present");
assert_layers_managed_user_then_optional_system(&layers, managed_file, user_file)?;
assert_eq!(layers.len(), 2);
assert_eq!(
layers[0].name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile { file: managed_file }
);
assert_eq!(layers[1].name, ConfigLayerSource::User { file: user_file });
Ok(())
}
@@ -430,50 +433,3 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
Ok(())
}
fn assert_layers_user_then_optional_system(
layers: &[codex_app_server_protocol::ConfigLayer],
user_file: AbsolutePathBuf,
) -> Result<()> {
if cfg!(unix) {
let system_file = AbsolutePathBuf::from_absolute_path(SYSTEM_CONFIG_TOML_FILE_UNIX)?;
assert_eq!(layers.len(), 2);
assert_eq!(layers[0].name, ConfigLayerSource::User { file: user_file });
assert_eq!(
layers[1].name,
ConfigLayerSource::System { file: system_file }
);
} else {
assert_eq!(layers.len(), 1);
assert_eq!(layers[0].name, ConfigLayerSource::User { file: user_file });
}
Ok(())
}
fn assert_layers_managed_user_then_optional_system(
layers: &[codex_app_server_protocol::ConfigLayer],
managed_file: AbsolutePathBuf,
user_file: AbsolutePathBuf,
) -> Result<()> {
if cfg!(unix) {
let system_file = AbsolutePathBuf::from_absolute_path(SYSTEM_CONFIG_TOML_FILE_UNIX)?;
assert_eq!(layers.len(), 3);
assert_eq!(
layers[0].name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile { file: managed_file }
);
assert_eq!(layers[1].name, ConfigLayerSource::User { file: user_file });
assert_eq!(
layers[2].name,
ConfigLayerSource::System { file: system_file }
);
} else {
assert_eq!(layers.len(), 2);
assert_eq!(
layers[0].name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile { file: managed_file }
);
assert_eq!(layers[1].name, ConfigLayerSource::User { file: user_file });
}
Ok(())
}

View File

@@ -1,13 +1,11 @@
mod account;
mod config_rpc;
mod model_list;
mod output_schema;
mod rate_limits;
mod review;
mod thread_archive;
mod thread_list;
mod thread_resume;
mod thread_rollback;
mod thread_start;
mod turn_interrupt;
mod turn_start;

View File

@@ -1,231 +0,0 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::to_response;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
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 core_test_support::responses;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
use std::path::Path;
use tempfile::TempDir;
use tokio::time::timeout;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
#[tokio::test]
async fn turn_start_accepts_output_schema_v2() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let response_mock = responses::mount_sse_once(&server, body).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 thread_req = mcp
.send_thread_start_request(ThreadStartParams {
..Default::default()
})
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
let output_schema = serde_json::json!({
"type": "object",
"properties": {
"answer": { "type": "string" }
},
"required": ["answer"],
"additionalProperties": false
});
let turn_req = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "Hello".to_string(),
}],
output_schema: Some(output_schema.clone()),
..Default::default()
})
.await?;
let turn_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
)
.await??;
let _turn: TurnStartResponse = to_response::<TurnStartResponse>(turn_resp)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
let request = response_mock.single_request();
let payload = request.body_json();
let text = payload.get("text").expect("request missing text field");
let format = text
.get("format")
.expect("request missing text.format field");
assert_eq!(
format,
&serde_json::json!({
"name": "codex_output_schema",
"type": "json_schema",
"strict": true,
"schema": output_schema,
})
);
Ok(())
}
#[tokio::test]
async fn turn_start_output_schema_is_per_turn_v2() -> Result<()> {
skip_if_no_network!(Ok(()));
let server = responses::start_mock_server().await;
let body1 = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let response_mock1 = responses::mount_sse_once(&server, body1).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 thread_req = mcp
.send_thread_start_request(ThreadStartParams {
..Default::default()
})
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
let output_schema = serde_json::json!({
"type": "object",
"properties": {
"answer": { "type": "string" }
},
"required": ["answer"],
"additionalProperties": false
});
let turn_req_1 = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "Hello".to_string(),
}],
output_schema: Some(output_schema.clone()),
..Default::default()
})
.await?;
let turn_resp_1: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req_1)),
)
.await??;
let _turn: TurnStartResponse = to_response::<TurnStartResponse>(turn_resp_1)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
let payload1 = response_mock1.single_request().body_json();
assert_eq!(
payload1.pointer("/text/format"),
Some(&serde_json::json!({
"name": "codex_output_schema",
"type": "json_schema",
"strict": true,
"schema": output_schema,
}))
);
let body2 = responses::sse(vec![
responses::ev_response_created("resp-2"),
responses::ev_assistant_message("msg-2", "Done"),
responses::ev_completed("resp-2"),
]);
let response_mock2 = responses::mount_sse_once(&server, body2).await;
let turn_req_2 = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "Hello again".to_string(),
}],
output_schema: None,
..Default::default()
})
.await?;
let turn_resp_2: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req_2)),
)
.await??;
let _turn: TurnStartResponse = to_response::<TurnStartResponse>(turn_resp_2)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
let payload2 = response_mock2.single_request().body_json();
assert_eq!(payload2.pointer("/text/format"), None);
Ok(())
}
fn create_config_toml(codex_home: &Path, server_uri: &str) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
std::fs::write(
config_toml,
format!(
r#"
model = "mock-model"
approval_policy = "never"
sandbox_mode = "read-only"
model_provider = "mock_provider"
[model_providers.mock_provider]
name = "Mock provider for test"
base_url = "{server_uri}/v1"
wire_api = "responses"
request_max_retries = 0
stream_max_retries = 0
"#
),
)
}

View File

@@ -1,177 +0,0 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::create_final_assistant_message_sse_response;
use app_test_support::create_mock_chat_completions_server_unchecked;
use app_test_support::to_response;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::ThreadResumeParams;
use codex_app_server_protocol::ThreadResumeResponse;
use codex_app_server_protocol::ThreadRollbackParams;
use codex_app_server_protocol::ThreadRollbackResponse;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::TurnStartParams;
use codex_app_server_protocol::UserInput as V2UserInput;
use pretty_assertions::assert_eq;
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_rollback_drops_last_turns_and_persists_to_rollout() -> Result<()> {
// Three Codex turns hit the mock model (session start + two turn/start calls).
let responses = vec![
create_final_assistant_message_sse_response("Done")?,
create_final_assistant_message_sse_response("Done")?,
create_final_assistant_message_sse_response("Done")?,
];
let server = create_mock_chat_completions_server_unchecked(responses).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("mock-model".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)?;
// Two turns.
let first_text = "First";
let turn1_id = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: first_text.to_string(),
}],
..Default::default()
})
.await?;
let _turn1_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn1_id)),
)
.await??;
let _completed1 = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
let turn2_id = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "Second".to_string(),
}],
..Default::default()
})
.await?;
let _turn2_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn2_id)),
)
.await??;
let _completed2 = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
// Roll back the last turn.
let rollback_id = mcp
.send_thread_rollback_request(ThreadRollbackParams {
thread_id: thread.id.clone(),
num_turns: 1,
})
.await?;
let rollback_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(rollback_id)),
)
.await??;
let ThreadRollbackResponse {
thread: rolled_back_thread,
} = to_response::<ThreadRollbackResponse>(rollback_resp)?;
assert_eq!(rolled_back_thread.turns.len(), 1);
assert_eq!(rolled_back_thread.turns[0].items.len(), 2);
match &rolled_back_thread.turns[0].items[0] {
ThreadItem::UserMessage { content, .. } => {
assert_eq!(
content,
&vec![V2UserInput::Text {
text: first_text.to_string()
}]
);
}
other => panic!("expected user message item, got {other:?}"),
}
// Resume and confirm the history is pruned.
let resume_id = mcp
.send_thread_resume_request(ThreadResumeParams {
thread_id: thread.id,
..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, .. } = to_response::<ThreadResumeResponse>(resume_resp)?;
assert_eq!(thread.turns.len(), 1);
assert_eq!(thread.turns[0].items.len(), 2);
match &thread.turns[0].items[0] {
ThreadItem::UserMessage { content, .. } => {
assert_eq!(
content,
&vec![V2UserInput::Text {
text: first_text.to_string()
}]
);
}
other => panic!("expected user message item, got {other:?}"),
}
Ok(())
}
fn create_config_toml(codex_home: &std::path::Path, server_uri: &str) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
std::fs::write(
config_toml,
format!(
r#"
model = "mock-model"
approval_policy = "never"
sandbox_mode = "read-only"
model_provider = "mock_provider"
[model_providers.mock_provider]
name = "Mock provider for test"
base_url = "{server_uri}/v1"
wire_api = "chat"
request_max_retries = 0
stream_max_retries = 0
"#
),
)
}

View File

@@ -540,7 +540,6 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
model: Some("mock-model".to_string()),
effort: Some(ReasoningEffort::Medium),
summary: Some(ReasoningSummary::Auto),
output_schema: None,
})
.await?;
timeout(
@@ -567,7 +566,6 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
model: Some("mock-model".to_string()),
effort: Some(ReasoningEffort::Medium),
summary: Some(ReasoningSummary::Auto),
output_schema: None,
})
.await?;
timeout(

View File

@@ -25,6 +25,5 @@ tree-sitter-bash = { workspace = true }
[dev-dependencies]
assert_cmd = { workspace = true }
assert_matches = { workspace = true }
codex-utils-cargo-bin = { workspace = true }
pretty_assertions = { workspace = true }
tempfile = { workspace = true }

View File

@@ -227,14 +227,11 @@ fn check_start_and_end_lines_strict(
first_line: Option<&&str>,
last_line: Option<&&str>,
) -> Result<(), ParseError> {
let first_line = first_line.map(|line| line.trim());
let last_line = last_line.map(|line| line.trim());
match (first_line, last_line) {
(Some(first), Some(last)) if first == BEGIN_PATCH_MARKER && last == END_PATCH_MARKER => {
(Some(&first), Some(&last)) if first == BEGIN_PATCH_MARKER && last == END_PATCH_MARKER => {
Ok(())
}
(Some(first), _) if first != BEGIN_PATCH_MARKER => Err(InvalidPatchError(String::from(
(Some(&first), _) if first != BEGIN_PATCH_MARKER => Err(InvalidPatchError(String::from(
"The first line of the patch must be '*** Begin Patch'",
))),
_ => Err(InvalidPatchError(String::from(
@@ -447,25 +444,6 @@ fn test_parse_patch() {
"The last line of the patch must be '*** End Patch'".to_string()
))
);
assert_eq!(
parse_patch_text(
concat!(
"*** Begin Patch",
" ",
"\n*** Add File: foo\n+hi\n",
" ",
"*** End Patch"
),
ParseMode::Strict
)
.unwrap()
.hunks,
vec![AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
assert_eq!(
parse_patch_text(
"*** Begin Patch\n\

View File

@@ -1,3 +0,0 @@
*** Begin Patch
*** Delete File: obsolete.txt
*** End Patch

View File

@@ -1,6 +0,0 @@
*** Begin Patch
*** Update File: file.txt
@@
-one
+two
*** End Patch

View File

@@ -1,7 +0,0 @@
*** Begin Patch
*** Update File: lines.txt
@@
line1
-line2
line3
*** End Patch

View File

@@ -1,8 +0,0 @@
*** Begin Patch
*** Update File: tail.txt
@@
first
-second
+second updated
*** End of File
*** End Patch

View File

@@ -1,13 +1,8 @@
use assert_cmd::Command;
use assert_cmd::prelude::*;
use std::fs;
use std::process::Command;
use tempfile::tempdir;
fn apply_patch_command() -> anyhow::Result<Command> {
Ok(Command::new(codex_utils_cargo_bin::cargo_bin(
"apply_patch",
)?))
}
#[test]
fn test_apply_patch_cli_add_and_update() -> anyhow::Result<()> {
let tmp = tempdir()?;
@@ -21,7 +16,8 @@ fn test_apply_patch_cli_add_and_update() -> anyhow::Result<()> {
+hello
*** End Patch"#
);
apply_patch_command()?
Command::cargo_bin("apply_patch")
.expect("should find apply_patch binary")
.arg(add_patch)
.current_dir(tmp.path())
.assert()
@@ -38,7 +34,8 @@ fn test_apply_patch_cli_add_and_update() -> anyhow::Result<()> {
+world
*** End Patch"#
);
apply_patch_command()?
Command::cargo_bin("apply_patch")
.expect("should find apply_patch binary")
.arg(update_patch)
.current_dir(tmp.path())
.assert()
@@ -62,9 +59,10 @@ fn test_apply_patch_cli_stdin_add_and_update() -> anyhow::Result<()> {
+hello
*** End Patch"#
);
apply_patch_command()?
.current_dir(tmp.path())
.write_stdin(add_patch)
let mut cmd =
assert_cmd::Command::cargo_bin("apply_patch").expect("should find apply_patch binary");
cmd.current_dir(tmp.path());
cmd.write_stdin(add_patch)
.assert()
.success()
.stdout(format!("Success. Updated the following files:\nA {file}\n"));
@@ -79,9 +77,10 @@ fn test_apply_patch_cli_stdin_add_and_update() -> anyhow::Result<()> {
+world
*** End Patch"#
);
apply_patch_command()?
.current_dir(tmp.path())
.write_stdin(update_patch)
let mut cmd =
assert_cmd::Command::cargo_bin("apply_patch").expect("should find apply_patch binary");
cmd.current_dir(tmp.path());
cmd.write_stdin(update_patch)
.assert()
.success()
.stdout(format!("Success. Updated the following files:\nM {file}\n"));

View File

@@ -1,3 +1,4 @@
use assert_cmd::prelude::*;
use pretty_assertions::assert_eq;
use std::collections::BTreeMap;
use std::fs;
@@ -8,8 +9,7 @@ use tempfile::tempdir;
#[test]
fn test_apply_patch_scenarios() -> anyhow::Result<()> {
let scenarios_dir = Path::new(env!("CARGO_MANIFEST_DIR")).join("tests/fixtures/scenarios");
for scenario in fs::read_dir(scenarios_dir)? {
for scenario in fs::read_dir("tests/fixtures/scenarios")? {
let scenario = scenario?;
let path = scenario.path();
if path.is_dir() {
@@ -36,7 +36,7 @@ fn run_apply_patch_scenario(dir: &Path) -> anyhow::Result<()> {
// Run apply_patch in the temporary directory. We intentionally do not assert
// on the exit status here; the scenarios are specified purely in terms of
// final filesystem state, which we compare below.
Command::new(codex_utils_cargo_bin::cargo_bin("apply_patch")?)
Command::cargo_bin("apply_patch")?
.arg(patch)
.current_dir(tmp.path())
.output()?;
@@ -82,15 +82,11 @@ fn snapshot_dir_recursive(
continue;
};
let rel = stripped.to_path_buf();
// Under Buck2, files in `__srcs` are often materialized as symlinks.
// Use `metadata()` (follows symlinks) so our fixture snapshots work
// under both Cargo and Buck2.
let metadata = fs::metadata(&path)?;
if metadata.is_dir() {
let file_type = entry.file_type()?;
if file_type.is_dir() {
entries.insert(rel.clone(), Entry::Dir);
snapshot_dir_recursive(base, &path, entries)?;
} else if metadata.is_file() {
} else if file_type.is_file() {
let contents = fs::read(&path)?;
entries.insert(rel, Entry::File(contents));
}
@@ -102,14 +98,12 @@ fn copy_dir_recursive(src: &Path, dst: &Path) -> anyhow::Result<()> {
for entry in fs::read_dir(src)? {
let entry = entry?;
let path = entry.path();
let file_type = entry.file_type()?;
let dest_path = dst.join(entry.file_name());
// See note in `snapshot_dir_recursive` about Buck2 symlink trees.
let metadata = fs::metadata(&path)?;
if metadata.is_dir() {
if file_type.is_dir() {
fs::create_dir_all(&dest_path)?;
copy_dir_recursive(&path, &dest_path)?;
} else if metadata.is_file() {
} else if file_type.is_file() {
if let Some(parent) = dest_path.parent() {
fs::create_dir_all(parent)?;
}

View File

@@ -5,13 +5,13 @@ use std::path::Path;
use tempfile::tempdir;
fn run_apply_patch_in_dir(dir: &Path, patch: &str) -> anyhow::Result<assert_cmd::assert::Assert> {
let mut cmd = Command::new(codex_utils_cargo_bin::cargo_bin("apply_patch")?);
let mut cmd = Command::cargo_bin("apply_patch")?;
cmd.current_dir(dir);
Ok(cmd.arg(patch).assert())
}
fn apply_patch_command(dir: &Path) -> anyhow::Result<Command> {
let mut cmd = Command::new(codex_utils_cargo_bin::cargo_bin("apply_patch")?);
let mut cmd = Command::cargo_bin("apply_patch")?;
cmd.current_dir(dir);
Ok(cmd)
}

View File

@@ -60,7 +60,6 @@ codex_windows_sandbox = { package = "codex-windows-sandbox", path = "../windows-
[dev-dependencies]
assert_cmd = { workspace = true }
assert_matches = { workspace = true }
codex-utils-cargo-bin = { workspace = true }
predicates = { workspace = true }
pretty_assertions = { workspace = true }
tempfile = { workspace = true }

View File

@@ -480,12 +480,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
}
Some(Subcommand::AppServer(app_server_cli)) => match app_server_cli.subcommand {
None => {
codex_app_server::run_main(
codex_linux_sandbox_exe,
root_config_overrides,
codex_core::config_loader::LoaderOverrides::default(),
)
.await?;
codex_app_server::run_main(codex_linux_sandbox_exe, root_config_overrides).await?;
}
Some(AppServerSubcommand::GenerateTs(gen_cli)) => {
codex_app_server_protocol::generate_ts(

View File

@@ -24,7 +24,7 @@ prefix_rule(
"#,
)?;
let output = Command::new(codex_utils_cargo_bin::cargo_bin("codex")?)
let output = Command::cargo_bin("codex")?
.env("CODEX_HOME", codex_home.path())
.args([
"execpolicy",
@@ -59,61 +59,3 @@ prefix_rule(
Ok(())
}
#[test]
fn execpolicy_check_includes_justification_when_present() -> Result<(), Box<dyn std::error::Error>>
{
let codex_home = TempDir::new()?;
let policy_path = codex_home.path().join("rules").join("policy.rules");
fs::create_dir_all(
policy_path
.parent()
.expect("policy path should have a parent"),
)?;
fs::write(
&policy_path,
r#"
prefix_rule(
pattern = ["git", "push"],
decision = "forbidden",
justification = "pushing is blocked in this repo",
)
"#,
)?;
let output = Command::new(codex_utils_cargo_bin::cargo_bin("codex")?)
.env("CODEX_HOME", codex_home.path())
.args([
"execpolicy",
"check",
"--rules",
policy_path
.to_str()
.expect("policy path should be valid UTF-8"),
"git",
"push",
"origin",
"main",
])
.output()?;
assert!(output.status.success());
let result: serde_json::Value = serde_json::from_slice(&output.stdout)?;
assert_eq!(
result,
json!({
"decision": "forbidden",
"matchedRules": [
{
"prefixRuleMatch": {
"matchedPrefix": ["git", "push"],
"decision": "forbidden",
"justification": "pushing is blocked in this repo"
}
}
]
})
);
Ok(())
}

View File

@@ -8,7 +8,7 @@ use pretty_assertions::assert_eq;
use tempfile::TempDir;
fn codex_command(codex_home: &Path) -> Result<assert_cmd::Command> {
let mut cmd = assert_cmd::Command::new(codex_utils_cargo_bin::cargo_bin("codex")?);
let mut cmd = assert_cmd::Command::cargo_bin("codex")?;
cmd.env("CODEX_HOME", codex_home);
Ok(cmd)
}

View File

@@ -12,7 +12,7 @@ use serde_json::json;
use tempfile::TempDir;
fn codex_command(codex_home: &Path) -> Result<assert_cmd::Command> {
let mut cmd = assert_cmd::Command::new(codex_utils_cargo_bin::cargo_bin("codex")?);
let mut cmd = assert_cmd::Command::cargo_bin("codex")?;
cmd.env("CODEX_HOME", codex_home);
Ok(cmd)
}

View File

@@ -59,7 +59,6 @@ pub enum ResponseEvent {
summary_index: i64,
},
RateLimits(RateLimitSnapshot),
ModelsEtag(String),
}
#[derive(Debug, Serialize, Clone)]

View File

@@ -152,9 +152,6 @@ impl Stream for AggregatedStream {
Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot)))) => {
return Poll::Ready(Some(Ok(ResponseEvent::RateLimits(snapshot))));
}
Poll::Ready(Some(Ok(ResponseEvent::ModelsEtag(etag)))) => {
return Poll::Ready(Some(Ok(ResponseEvent::ModelsEtag(etag))));
}
Poll::Ready(Some(Ok(ResponseEvent::Completed {
response_id,
token_usage,

View File

@@ -5,7 +5,6 @@ use crate::provider::Provider;
use crate::telemetry::run_with_request_telemetry;
use codex_client::HttpTransport;
use codex_client::RequestTelemetry;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ModelsResponse;
use http::HeaderMap;
use http::Method;
@@ -42,7 +41,7 @@ impl<T: HttpTransport, A: AuthProvider> ModelsClient<T, A> {
&self,
client_version: &str,
extra_headers: HeaderMap,
) -> Result<(Vec<ModelInfo>, Option<String>), ApiError> {
) -> Result<ModelsResponse, ApiError> {
let builder = || {
let mut req = self.provider.build_request(Method::GET, self.path());
req.headers.extend(extra_headers.clone());
@@ -67,7 +66,7 @@ impl<T: HttpTransport, A: AuthProvider> ModelsClient<T, A> {
.and_then(|value| value.to_str().ok())
.map(ToString::to_string);
let ModelsResponse { models } = serde_json::from_slice::<ModelsResponse>(&resp.body)
let ModelsResponse { models, etag } = serde_json::from_slice::<ModelsResponse>(&resp.body)
.map_err(|e| {
ApiError::Stream(format!(
"failed to decode models response: {e}; body: {}",
@@ -75,7 +74,9 @@ impl<T: HttpTransport, A: AuthProvider> ModelsClient<T, A> {
))
})?;
Ok((models, header_etag))
let etag = header_etag.unwrap_or(etag);
Ok(ModelsResponse { models, etag })
}
}
@@ -101,15 +102,16 @@ mod tests {
struct CapturingTransport {
last_request: Arc<Mutex<Option<Request>>>,
body: Arc<ModelsResponse>,
etag: Option<String>,
}
impl Default for CapturingTransport {
fn default() -> Self {
Self {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(ModelsResponse { models: Vec::new() }),
etag: None,
body: Arc::new(ModelsResponse {
models: Vec::new(),
etag: String::new(),
}),
}
}
}
@@ -120,8 +122,8 @@ mod tests {
*self.last_request.lock().unwrap() = Some(req);
let body = serde_json::to_vec(&*self.body).unwrap();
let mut headers = HeaderMap::new();
if let Some(etag) = &self.etag {
headers.insert(ETAG, etag.parse().unwrap());
if !self.body.etag.is_empty() {
headers.insert(ETAG, self.body.etag.parse().unwrap());
}
Ok(Response {
status: StatusCode::OK,
@@ -164,12 +166,14 @@ mod tests {
#[tokio::test]
async fn appends_client_version_query() {
let response = ModelsResponse { models: Vec::new() };
let response = ModelsResponse {
models: Vec::new(),
etag: String::new(),
};
let transport = CapturingTransport {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(response),
etag: None,
};
let client = ModelsClient::new(
@@ -178,12 +182,12 @@ mod tests {
DummyAuth,
);
let (models, _) = client
let result = client
.list_models("0.99.0", HeaderMap::new())
.await
.expect("request should succeed");
assert_eq!(models.len(), 0);
assert_eq!(result.models.len(), 0);
let url = transport
.last_request
@@ -223,16 +227,17 @@ mod tests {
"truncation_policy": {"mode": "bytes", "limit": 10_000},
"supports_parallel_tool_calls": false,
"context_window": null,
"reasoning_summary_format": "none",
"experimental_supported_tools": [],
}))
.unwrap(),
],
etag: String::new(),
};
let transport = CapturingTransport {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(response),
etag: None,
};
let client = ModelsClient::new(
@@ -241,25 +246,27 @@ mod tests {
DummyAuth,
);
let (models, _) = client
let result = client
.list_models("0.99.0", HeaderMap::new())
.await
.expect("request should succeed");
assert_eq!(models.len(), 1);
assert_eq!(models[0].slug, "gpt-test");
assert_eq!(models[0].supported_in_api, true);
assert_eq!(models[0].priority, 1);
assert_eq!(result.models.len(), 1);
assert_eq!(result.models[0].slug, "gpt-test");
assert_eq!(result.models[0].supported_in_api, true);
assert_eq!(result.models[0].priority, 1);
}
#[tokio::test]
async fn list_models_includes_etag() {
let response = ModelsResponse { models: Vec::new() };
let response = ModelsResponse {
models: Vec::new(),
etag: "\"abc\"".to_string(),
};
let transport = CapturingTransport {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(response),
etag: Some("\"abc\"".to_string()),
};
let client = ModelsClient::new(
@@ -268,12 +275,12 @@ mod tests {
DummyAuth,
);
let (models, etag) = client
let result = client
.list_models("0.1.0", HeaderMap::new())
.await
.expect("request should succeed");
assert_eq!(models.len(), 0);
assert_eq!(etag, Some("\"abc\"".to_string()));
assert_eq!(result.models.len(), 0);
assert_eq!(result.etag, "\"abc\"");
}
}

View File

@@ -204,16 +204,24 @@ impl<'a> ChatRequestBuilder<'a> {
call_id,
..
} => {
let reasoning = reasoning_by_anchor_index.get(&idx).map(String::as_str);
let tool_call = json!({
"id": call_id,
"type": "function",
"function": {
"name": name,
"arguments": arguments,
}
let mut msg = json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": call_id,
"type": "function",
"function": {
"name": name,
"arguments": arguments,
}
}]
});
push_tool_call_message(&mut messages, tool_call, reasoning);
if let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
ResponseItem::LocalShellCall {
id,
@@ -221,14 +229,22 @@ impl<'a> ChatRequestBuilder<'a> {
status,
action,
} => {
let reasoning = reasoning_by_anchor_index.get(&idx).map(String::as_str);
let tool_call = json!({
"id": id.clone().unwrap_or_default(),
"type": "local_shell_call",
"status": status,
"action": action,
let mut msg = json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": id.clone().unwrap_or_default(),
"type": "local_shell_call",
"status": status,
"action": action,
}]
});
push_tool_call_message(&mut messages, tool_call, reasoning);
if let Some(reasoning) = reasoning_by_anchor_index.get(&idx)
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
ResponseItem::FunctionCallOutput { call_id, output } => {
let content_value = if let Some(items) = &output.content_items {
@@ -261,16 +277,18 @@ impl<'a> ChatRequestBuilder<'a> {
input,
status: _,
} => {
let tool_call = json!({
"id": id,
"type": "custom",
"custom": {
"name": name,
"input": input,
}
});
let reasoning = reasoning_by_anchor_index.get(&idx).map(String::as_str);
push_tool_call_message(&mut messages, tool_call, reasoning);
messages.push(json!({
"role": "assistant",
"content": null,
"tool_calls": [{
"id": id,
"type": "custom",
"custom": {
"name": name,
"input": input,
}
}]
}));
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
messages.push(json!({
@@ -310,50 +328,11 @@ impl<'a> ChatRequestBuilder<'a> {
}
}
fn push_tool_call_message(messages: &mut Vec<Value>, tool_call: Value, reasoning: Option<&str>) {
// Chat Completions requires that tool calls are grouped into a single assistant message
// (with `tool_calls: [...]`) followed by tool role responses.
if let Some(Value::Object(obj)) = messages.last_mut()
&& obj.get("role").and_then(Value::as_str) == Some("assistant")
&& obj.get("content").is_some_and(Value::is_null)
&& let Some(tool_calls) = obj.get_mut("tool_calls").and_then(Value::as_array_mut)
{
tool_calls.push(tool_call);
if let Some(reasoning) = reasoning {
if let Some(Value::String(existing)) = obj.get_mut("reasoning") {
if !existing.is_empty() {
existing.push('\n');
}
existing.push_str(reasoning);
} else {
obj.insert(
"reasoning".to_string(),
Value::String(reasoning.to_string()),
);
}
}
return;
}
let mut msg = json!({
"role": "assistant",
"content": null,
"tool_calls": [tool_call],
});
if let Some(reasoning) = reasoning
&& let Some(obj) = msg.as_object_mut()
{
obj.insert("reasoning".to_string(), json!(reasoning));
}
messages.push(msg);
}
#[cfg(test)]
mod tests {
use super::*;
use crate::provider::RetryConfig;
use crate::provider::WireApi;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use http::HeaderValue;
@@ -406,89 +385,4 @@ mod tests {
Some(&HeaderValue::from_static("review"))
);
}
#[test]
fn groups_consecutive_tool_calls_into_a_single_assistant_message() {
let prompt_input = vec![
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "read these".to_string(),
}],
},
ResponseItem::FunctionCall {
id: None,
name: "read_file".to_string(),
arguments: r#"{"path":"a.txt"}"#.to_string(),
call_id: "call-a".to_string(),
},
ResponseItem::FunctionCall {
id: None,
name: "read_file".to_string(),
arguments: r#"{"path":"b.txt"}"#.to_string(),
call_id: "call-b".to_string(),
},
ResponseItem::FunctionCall {
id: None,
name: "read_file".to_string(),
arguments: r#"{"path":"c.txt"}"#.to_string(),
call_id: "call-c".to_string(),
},
ResponseItem::FunctionCallOutput {
call_id: "call-a".to_string(),
output: FunctionCallOutputPayload {
content: "A".to_string(),
..Default::default()
},
},
ResponseItem::FunctionCallOutput {
call_id: "call-b".to_string(),
output: FunctionCallOutputPayload {
content: "B".to_string(),
..Default::default()
},
},
ResponseItem::FunctionCallOutput {
call_id: "call-c".to_string(),
output: FunctionCallOutputPayload {
content: "C".to_string(),
..Default::default()
},
},
];
let req = ChatRequestBuilder::new("gpt-test", "inst", &prompt_input, &[])
.build(&provider())
.expect("request");
let messages = req
.body
.get("messages")
.and_then(|v| v.as_array())
.expect("messages array");
// system + user + assistant(tool_calls=[...]) + 3 tool outputs
assert_eq!(messages.len(), 6);
assert_eq!(messages[0]["role"], "system");
assert_eq!(messages[1]["role"], "user");
let tool_calls_msg = &messages[2];
assert_eq!(tool_calls_msg["role"], "assistant");
assert_eq!(tool_calls_msg["content"], serde_json::Value::Null);
let tool_calls = tool_calls_msg["tool_calls"]
.as_array()
.expect("tool_calls array");
assert_eq!(tool_calls.len(), 3);
assert_eq!(tool_calls[0]["id"], "call-a");
assert_eq!(tool_calls[1]["id"], "call-b");
assert_eq!(tool_calls[2]["id"], "call-c");
assert_eq!(messages[3]["role"], "tool");
assert_eq!(messages[3]["tool_call_id"], "call-a");
assert_eq!(messages[4]["role"], "tool");
assert_eq!(messages[4]["tool_call_id"], "call-b");
assert_eq!(messages[5]["role"], "tool");
assert_eq!(messages[5]["tool_call_id"], "call-c");
}
}

View File

@@ -30,21 +30,6 @@ pub(crate) fn spawn_chat_stream(
ResponseStream { rx_event }
}
/// Processes Server-Sent Events from the legacy Chat Completions streaming API.
///
/// The upstream protocol terminates a streaming response with a final sentinel event
/// (`data: [DONE]`). Historically, some of our test stubs have emitted `data: DONE`
/// (without brackets) instead.
///
/// `eventsource_stream` delivers these sentinels as regular events rather than signaling
/// end-of-stream. If we try to parse them as JSON, we log and skip them, then keep
/// polling for more events.
///
/// On servers that keep the HTTP connection open after emitting the sentinel (notably
/// wiremock on Windows), skipping the sentinel means we never emit `ResponseEvent::Completed`.
/// Higher-level workflows/tests that wait for completion before issuing subsequent model
/// calls will then stall, which shows up as "expected N requests, got 1" verification
/// failures in the mock server.
pub async fn process_chat_sse<S>(
stream: S,
tx_event: mpsc::Sender<Result<ResponseEvent, ApiError>>,
@@ -72,31 +57,6 @@ pub async fn process_chat_sse<S>(
let mut reasoning_item: Option<ResponseItem> = None;
let mut completed_sent = false;
async fn flush_and_complete(
tx_event: &mpsc::Sender<Result<ResponseEvent, ApiError>>,
reasoning_item: &mut Option<ResponseItem>,
assistant_item: &mut Option<ResponseItem>,
) {
if let Some(reasoning) = reasoning_item.take() {
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemDone(reasoning)))
.await;
}
if let Some(assistant) = assistant_item.take() {
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemDone(assistant)))
.await;
}
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
}
loop {
let start = Instant::now();
let response = timeout(idle_timeout, stream.next()).await;
@@ -110,8 +70,24 @@ pub async fn process_chat_sse<S>(
return;
}
Ok(None) => {
if let Some(reasoning) = reasoning_item {
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemDone(reasoning)))
.await;
}
if let Some(assistant) = assistant_item {
let _ = tx_event
.send(Ok(ResponseEvent::OutputItemDone(assistant)))
.await;
}
if !completed_sent {
flush_and_complete(&tx_event, &mut reasoning_item, &mut assistant_item).await;
let _ = tx_event
.send(Ok(ResponseEvent::Completed {
response_id: String::new(),
token_usage: None,
}))
.await;
}
return;
}
@@ -125,25 +101,16 @@ pub async fn process_chat_sse<S>(
trace!("SSE event: {}", sse.data);
let data = sse.data.trim();
if data.is_empty() {
if sse.data.trim().is_empty() {
continue;
}
if data == "[DONE]" || data == "DONE" {
if !completed_sent {
flush_and_complete(&tx_event, &mut reasoning_item, &mut assistant_item).await;
}
return;
}
let value: serde_json::Value = match serde_json::from_str(data) {
let value: serde_json::Value = match serde_json::from_str(&sse.data) {
Ok(val) => val,
Err(err) => {
debug!(
"Failed to parse ChatCompletions SSE event: {err}, data: {}",
data
&sse.data
);
continue;
}
@@ -395,16 +362,6 @@ mod tests {
body
}
/// Regression test: the stream should complete when we see a `[DONE]` sentinel.
///
/// This is important for tests/mocks that don't immediately close the underlying
/// connection after emitting the sentinel.
#[tokio::test]
async fn completes_on_done_sentinel_without_json() {
let events = collect_events("event: message\ndata: [DONE]\n\n").await;
assert_matches!(&events[..], [ResponseEvent::Completed { .. }]);
}
async fn collect_events(body: &str) -> Vec<ResponseEvent> {
let reader = ReaderStream::new(std::io::Cursor::new(body.to_string()))
.map_err(|err| codex_client::TransportError::Network(err.to_string()));

View File

@@ -51,19 +51,11 @@ pub fn spawn_response_stream(
telemetry: Option<Arc<dyn SseTelemetry>>,
) -> ResponseStream {
let rate_limits = parse_rate_limit(&stream_response.headers);
let models_etag = stream_response
.headers
.get("X-Models-Etag")
.and_then(|v| v.to_str().ok())
.map(ToString::to_string);
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent, ApiError>>(1600);
tokio::spawn(async move {
if let Some(snapshot) = rate_limits {
let _ = tx_event.send(Ok(ResponseEvent::RateLimits(snapshot))).await;
}
if let Some(etag) = models_etag {
let _ = tx_event.send(Ok(ResponseEvent::ModelsEtag(etag))).await;
}
process_sse(stream_response.bytes, tx_event, idle_timeout, telemetry).await;
});

View File

@@ -10,6 +10,7 @@ use codex_protocol::openai_models::ModelVisibility;
use codex_protocol::openai_models::ModelsResponse;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::openai_models::ReasoningEffortPreset;
use codex_protocol::openai_models::ReasoningSummaryFormat;
use codex_protocol::openai_models::TruncationPolicyConfig;
use http::HeaderMap;
use http::Method;
@@ -84,8 +85,10 @@ async fn models_client_hits_models_endpoint() {
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
context_window: None,
reasoning_summary_format: ReasoningSummaryFormat::None,
experimental_supported_tools: Vec::new(),
}],
etag: String::new(),
};
Mock::given(method("GET"))
@@ -101,13 +104,13 @@ async fn models_client_hits_models_endpoint() {
let transport = ReqwestTransport::new(reqwest::Client::new());
let client = ModelsClient::new(transport, provider(&base_url), DummyAuth);
let (models, _) = client
let result = client
.list_models("0.1.0", HeaderMap::new())
.await
.expect("models request should succeed");
assert_eq!(models.len(), 1);
assert_eq!(models[0].slug, "gpt-test");
assert_eq!(result.models.len(), 1);
assert_eq!(result.models[0].slug, "gpt-test");
let received = server
.received_requests()

View File

@@ -69,15 +69,6 @@ impl ReqwestTransport {
#[async_trait]
impl HttpTransport for ReqwestTransport {
async fn execute(&self, req: Request) -> Result<Response, TransportError> {
if enabled!(Level::TRACE) {
trace!(
"{} to {}: {}",
req.method,
req.url,
req.body.as_ref().unwrap_or_default()
);
}
let builder = self.build(req)?;
let resp = builder.send().await.map_err(Self::map_error)?;
let status = resp.status();

View File

@@ -123,7 +123,6 @@ assert_cmd = { workspace = true }
assert_matches = { workspace = true }
codex-arg0 = { workspace = true }
codex-core = { path = ".", features = ["deterministic_process_ids"] }
codex-utils-cargo-bin = { workspace = true }
core_test_support = { workspace = true }
ctor = { workspace = true }
escargot = { workspace = true }

View File

@@ -1,189 +0,0 @@
use crate::CodexConversation;
use crate::agent::AgentStatus;
use crate::conversation_manager::ConversationManagerState;
use crate::error::CodexErr;
use crate::error::Result as CodexResult;
use codex_protocol::ConversationId;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::Op;
use codex_protocol::user_input::UserInput;
use std::sync::Arc;
use std::sync::Weak;
/// Control-plane handle for multi-agent operations.
/// `AgentControl` is held by each session (via `SessionServices`). It provides capability to
/// spawn new agents and the inter-agent communication layer.
#[derive(Clone, Default)]
pub(crate) struct AgentControl {
/// Weak handle back to the global conversation registry/state.
/// This is `Weak` to avoid reference cycles and shadow persistence of the form
/// `ConversationManagerState -> CodexConversation -> Session -> SessionServices -> ConversationManagerState`.
manager: Weak<ConversationManagerState>,
}
impl AgentControl {
/// Construct a new `AgentControl` that can spawn/message agents via the given manager state.
pub(crate) fn new(manager: Weak<ConversationManagerState>) -> Self {
Self { manager }
}
#[allow(dead_code)] // Used by upcoming multi-agent tooling.
/// Spawn a new agent conversation and submit the initial prompt.
///
/// If `headless` is true, a background drain task is spawned to prevent unbounded event growth
/// of the channel queue when there is no client actively reading the conversation events.
pub(crate) async fn spawn_agent(
&self,
config: crate::config::Config,
prompt: String,
headless: bool,
) -> CodexResult<ConversationId> {
let state = self.upgrade()?;
let new_conversation = state.spawn_new_conversation(config, self.clone()).await?;
if headless {
spawn_headless_drain(Arc::clone(&new_conversation.conversation));
}
self.send_prompt(new_conversation.conversation_id, prompt)
.await?;
Ok(new_conversation.conversation_id)
}
#[allow(dead_code)] // Used by upcoming multi-agent tooling.
/// Send a `user` prompt to an existing agent conversation.
pub(crate) async fn send_prompt(
&self,
agent_id: ConversationId,
prompt: String,
) -> CodexResult<String> {
let state = self.upgrade()?;
state
.send_op(
agent_id,
Op::UserInput {
items: vec![UserInput::Text { text: prompt }],
final_output_json_schema: None,
},
)
.await
}
#[allow(dead_code)] // Used by upcoming multi-agent tooling.
/// Fetch the last known status for `agent_id`, returning `NotFound` when unavailable.
pub(crate) async fn get_status(&self, agent_id: ConversationId) -> AgentStatus {
let Ok(state) = self.upgrade() else {
// No agent available if upgrade fails.
return AgentStatus::NotFound;
};
let Ok(conversation) = state.get_conversation(agent_id).await else {
return AgentStatus::NotFound;
};
conversation.agent_status().await
}
fn upgrade(&self) -> CodexResult<Arc<ConversationManagerState>> {
self.manager.upgrade().ok_or_else(|| {
CodexErr::UnsupportedOperation("conversation manager dropped".to_string())
})
}
}
/// When an agent is spawned "headless" (no UI/view attached), there may be no consumer polling
/// `CodexConversation::next_event()`. The underlying event channel is unbounded, so the producer can
/// accumulate events indefinitely. This drain task prevents that memory growth by polling and
/// discarding events until shutdown.
fn spawn_headless_drain(conversation: Arc<CodexConversation>) {
tokio::spawn(async move {
loop {
match conversation.next_event().await {
Ok(event) => {
if matches!(event.msg, EventMsg::ShutdownComplete) {
break;
}
}
Err(err) => {
tracing::warn!("failed to receive event from agent: {err:?}");
break;
}
}
}
});
}
#[cfg(test)]
mod tests {
use super::*;
use crate::agent::agent_status_from_event;
use codex_protocol::protocol::ErrorEvent;
use codex_protocol::protocol::TaskCompleteEvent;
use codex_protocol::protocol::TaskStartedEvent;
use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::TurnAbortedEvent;
use pretty_assertions::assert_eq;
#[tokio::test]
async fn send_prompt_errors_when_manager_dropped() {
let control = AgentControl::default();
let err = control
.send_prompt(ConversationId::new(), "hello".to_string())
.await
.expect_err("send_prompt should fail without a manager");
assert_eq!(
err.to_string(),
"unsupported operation: conversation manager dropped"
);
}
#[tokio::test]
async fn get_status_returns_not_found_without_manager() {
let control = AgentControl::default();
let got = control.get_status(ConversationId::new()).await;
assert_eq!(got, AgentStatus::NotFound);
}
#[tokio::test]
async fn on_event_updates_status_from_task_started() {
let status = agent_status_from_event(&EventMsg::TaskStarted(TaskStartedEvent {
model_context_window: None,
}));
assert_eq!(status, Some(AgentStatus::Running));
}
#[tokio::test]
async fn on_event_updates_status_from_task_complete() {
let status = agent_status_from_event(&EventMsg::TaskComplete(TaskCompleteEvent {
last_agent_message: Some("done".to_string()),
}));
let expected = AgentStatus::Completed(Some("done".to_string()));
assert_eq!(status, Some(expected));
}
#[tokio::test]
async fn on_event_updates_status_from_error() {
let status = agent_status_from_event(&EventMsg::Error(ErrorEvent {
message: "boom".to_string(),
codex_error_info: None,
}));
let expected = AgentStatus::Errored("boom".to_string());
assert_eq!(status, Some(expected));
}
#[tokio::test]
async fn on_event_updates_status_from_turn_aborted() {
let status = agent_status_from_event(&EventMsg::TurnAborted(TurnAbortedEvent {
reason: TurnAbortReason::Interrupted,
}));
let expected = AgentStatus::Errored("Interrupted".to_string());
assert_eq!(status, Some(expected));
}
#[tokio::test]
async fn on_event_updates_status_from_shutdown_complete() {
let status = agent_status_from_event(&EventMsg::ShutdownComplete);
assert_eq!(status, Some(AgentStatus::Shutdown));
}
}

View File

@@ -1,6 +0,0 @@
pub(crate) mod control;
pub(crate) mod status;
pub(crate) use codex_protocol::protocol::AgentStatus;
pub(crate) use control::AgentControl;
pub(crate) use status::agent_status_from_event;

View File

@@ -1,15 +0,0 @@
use codex_protocol::protocol::AgentStatus;
use codex_protocol::protocol::EventMsg;
/// Derive the next agent status from a single emitted event.
/// Returns `None` when the event does not affect status tracking.
pub(crate) fn agent_status_from_event(msg: &EventMsg) -> Option<AgentStatus> {
match msg {
EventMsg::TaskStarted(_) => Some(AgentStatus::Running),
EventMsg::TaskComplete(ev) => Some(AgentStatus::Completed(ev.last_agent_message.clone())),
EventMsg::TurnAborted(ev) => Some(AgentStatus::Errored(format!("{:?}", ev.reason))),
EventMsg::Error(ev) => Some(AgentStatus::Errored(ev.message.clone())),
EventMsg::ShutdownComplete => Some(AgentStatus::Shutdown),
_ => None,
}
}

View File

@@ -195,13 +195,8 @@ pub(crate) mod tools {
LocalShell {},
// TODO: Understand why we get an error on web_search although the API docs say it's supported.
// https://platform.openai.com/docs/guides/tools-web-search?api-mode=responses#:~:text=%7B%20type%3A%20%22web_search%22%20%7D%2C
// The `external_web_access` field determines whether the web search is over cached or live content.
// https://platform.openai.com/docs/guides/tools-web-search#live-internet-access
#[serde(rename = "web_search")]
WebSearch {
#[serde(skip_serializing_if = "Option::is_none")]
external_web_access: Option<bool>,
},
WebSearch {},
#[serde(rename = "custom")]
Freeform(FreeformTool),
}
@@ -211,7 +206,7 @@ pub(crate) mod tools {
match self {
ToolSpec::Function(tool) => tool.name.as_str(),
ToolSpec::LocalShell {} => "local_shell",
ToolSpec::WebSearch { .. } => "web_search",
ToolSpec::WebSearch {} => "web_search",
ToolSpec::Freeform(tool) => tool.name.as_str(),
}
}

View File

@@ -8,9 +8,6 @@ use std::sync::atomic::Ordering;
use crate::AuthManager;
use crate::SandboxState;
use crate::agent::AgentControl;
use crate::agent::AgentStatus;
use crate::agent::agent_status_from_event;
use crate::client_common::REVIEW_PROMPT;
use crate::compact;
use crate::compact::run_inline_auto_compact_task;
@@ -91,7 +88,6 @@ use crate::error::Result as CodexResult;
#[cfg(test)]
use crate::exec::StreamOutput;
use crate::exec_policy::ExecPolicyUpdateError;
use crate::feedback_tags;
use crate::mcp::auth::compute_auth_statuses;
use crate::mcp_connection_manager::McpConnectionManager;
use crate::model_provider_info::CHAT_WIRE_API_DEPRECATION_SUMMARY;
@@ -170,8 +166,6 @@ pub struct Codex {
pub(crate) next_id: AtomicU64,
pub(crate) tx_sub: Sender<Submission>,
pub(crate) rx_event: Receiver<Event>,
// Last known status of the agent.
pub(crate) agent_status: Arc<RwLock<AgentStatus>>,
}
/// Wrapper returned by [`Codex::spawn`] containing the spawned [`Codex`],
@@ -212,23 +206,21 @@ fn maybe_push_chat_wire_api_deprecation(
impl Codex {
/// Spawn a new [`Codex`] and initialize the session.
pub(crate) async fn spawn(
pub async fn spawn(
config: Config,
auth_manager: Arc<AuthManager>,
models_manager: Arc<ModelsManager>,
skills_manager: Arc<SkillsManager>,
conversation_history: InitialHistory,
session_source: SessionSource,
agent_control: AgentControl,
) -> CodexResult<CodexSpawnOk> {
let (tx_sub, rx_sub) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
let (tx_event, rx_event) = async_channel::unbounded();
let loaded_skills = if config.features.enabled(Feature::Skills) {
Some(skills_manager.skills_for_config(&config))
} else {
None
};
let loaded_skills = config
.features
.enabled(Feature::Skills)
.then(|| skills_manager.skills_for_cwd(&config.cwd));
if let Some(outcome) = &loaded_skills {
for err in &outcome.errors {
@@ -248,15 +240,13 @@ impl Codex {
)
.await;
let exec_policy = ExecPolicyManager::load(&config.features, &config.config_layer_stack)
let exec_policy = ExecPolicyManager::load(&config.features, &config.codex_home)
.await
.map_err(|err| CodexErr::Fatal(format!("failed to load execpolicy: {err}")))?;
let config = Arc::new(config);
if config.features.enabled(Feature::RemoteModels)
&& let Err(err) = models_manager
.refresh_available_models_with_cache(&config)
.await
&& let Err(err) = models_manager.refresh_available_models(&config).await
{
error!("failed to refresh available models: {err:?}");
}
@@ -279,7 +269,6 @@ impl Codex {
// Generate a unique ID for the lifetime of this Codex session.
let session_source_clone = session_configuration.session_source.clone();
let agent_status = Arc::new(RwLock::new(AgentStatus::PendingInit));
let session = Session::new(
session_configuration,
@@ -288,11 +277,9 @@ impl Codex {
models_manager.clone(),
exec_policy,
tx_event.clone(),
Arc::clone(&agent_status),
conversation_history,
session_source_clone,
skills_manager,
agent_control,
)
.await
.map_err(|e| {
@@ -307,7 +294,6 @@ impl Codex {
next_id: AtomicU64::new(0),
tx_sub,
rx_event,
agent_status,
};
Ok(CodexSpawnOk {
@@ -345,11 +331,6 @@ impl Codex {
.map_err(|_| CodexErr::InternalAgentDied)?;
Ok(event)
}
pub(crate) async fn agent_status(&self) -> AgentStatus {
let status = self.agent_status.read().await;
status.clone()
}
}
/// Context for an initialized model agent
@@ -358,7 +339,6 @@ impl Codex {
pub(crate) struct Session {
conversation_id: ConversationId,
tx_event: Sender<Event>,
agent_status: Arc<RwLock<AgentStatus>>,
state: Mutex<SessionState>,
/// The set of enabled features should be invariant for the lifetime of the
/// session.
@@ -562,11 +542,9 @@ impl Session {
models_manager: Arc<ModelsManager>,
exec_policy: ExecPolicyManager,
tx_event: Sender<Event>,
agent_status: Arc<RwLock<AgentStatus>>,
initial_history: InitialHistory,
session_source: SessionSource,
skills_manager: Arc<SkillsManager>,
agent_control: AgentControl,
) -> anyhow::Result<Arc<Self>> {
debug!(
"Configuring session: model={}; provider={:?}",
@@ -688,13 +666,11 @@ impl Session {
models_manager: Arc::clone(&models_manager),
tool_approvals: Mutex::new(ApprovalStore::default()),
skills_manager,
agent_control,
};
let sess = Arc::new(Session {
conversation_id,
tx_event: tx_event.clone(),
agent_status: Arc::clone(&agent_status),
state: Mutex::new(state),
features: config.features.clone(),
active_turn: Mutex::new(None),
@@ -832,13 +808,6 @@ impl Session {
.await;
}
// Seed usage info from the recorded rollout so UIs can show token counts
// immediately on resume/fork.
if let Some(info) = Self::last_token_info_from_rollout(&rollout_items) {
let mut state = self.state.lock().await;
state.set_token_info(Some(info));
}
// If persisting, persist all rollout items as-is (recorder filters)
if persist && !rollout_items.is_empty() {
self.persist_rollout_items(&rollout_items).await;
@@ -849,13 +818,6 @@ impl Session {
}
}
fn last_token_info_from_rollout(rollout_items: &[RolloutItem]) -> Option<TokenUsageInfo> {
rollout_items.iter().rev().find_map(|item| match item {
RolloutItem::EventMsg(EventMsg::TokenCount(ev)) => ev.info.clone(),
_ => None,
})
}
pub(crate) async fn update_settings(
&self,
updates: SessionSettingsUpdate,
@@ -1015,11 +977,6 @@ impl Session {
}
pub(crate) async fn send_event_raw(&self, event: Event) {
// Record the last known agent status.
if let Some(status) = agent_status_from_event(&event.msg) {
let mut guard = self.agent_status.write().await;
*guard = status;
}
// Persist the event into rollout (recorder filters as needed)
let rollout_items = vec![RolloutItem::EventMsg(event.msg.clone())];
self.persist_rollout_items(&rollout_items).await;
@@ -1028,25 +985,6 @@ impl Session {
}
}
/// Persist the event to the rollout file, flush it, and only then deliver it to clients.
///
/// Most events can be delivered immediately after queueing the rollout write, but some
/// clients (e.g. app-server thread/rollback) re-read the rollout file synchronously on
/// receipt of the event and depend on the marker already being visible on disk.
pub(crate) async fn send_event_raw_flushed(&self, event: Event) {
// Record the last known agent status.
if let Some(status) = agent_status_from_event(&event.msg) {
let mut guard = self.agent_status.write().await;
*guard = status;
}
self.persist_rollout_items(&[RolloutItem::EventMsg(event.msg.clone())])
.await;
self.flush_rollout().await;
if let Err(e) = self.tx_event.send(event).await {
error!("failed to send tool call event: {e}");
}
}
pub(crate) async fn emit_turn_item_started(&self, turn_context: &TurnContext, item: &TurnItem) {
self.send_event(
turn_context,
@@ -1264,9 +1202,6 @@ impl Session {
history.replace(rebuilt);
}
}
RolloutItem::EventMsg(EventMsg::ThreadRolledBack(rollback)) => {
history.drop_last_n_user_turns(rollback.num_turns);
}
_ => {}
}
}
@@ -1487,14 +1422,12 @@ impl Session {
message: impl Into<String>,
codex_error: CodexErr,
) {
let additional_details = codex_error.to_string();
let codex_error_info = CodexErrorInfo::ResponseStreamDisconnected {
http_status_code: codex_error.http_status_code_value(),
};
let event = EventMsg::StreamError(StreamErrorEvent {
message: message.into(),
codex_error_info: Some(codex_error_info),
additional_details: Some(additional_details),
});
self.send_event(turn_context, event).await;
}
@@ -1706,9 +1639,6 @@ async fn submission_loop(sess: Arc<Session>, config: Arc<Config>, rx_sub: Receiv
Op::Compact => {
handlers::compact(&sess, sub.id.clone()).await;
}
Op::ThreadRollback { num_turns } => {
handlers::thread_rollback(&sess, sub.id.clone(), num_turns).await;
}
Op::RunUserShellCommand { command } => {
handlers::run_user_shell_command(
&sess,
@@ -1766,7 +1696,6 @@ mod handlers {
use codex_protocol::protocol::ReviewDecision;
use codex_protocol::protocol::ReviewRequest;
use codex_protocol::protocol::SkillsListEntry;
use codex_protocol::protocol::ThreadRolledBackEvent;
use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::WarningEvent;
@@ -1828,16 +1757,7 @@ mod handlers {
final_output_json_schema: Some(final_output_json_schema),
},
),
Op::UserInput {
items,
final_output_json_schema,
} => (
items,
SessionSettingsUpdate {
final_output_json_schema: Some(final_output_json_schema),
..Default::default()
},
),
Op::UserInput { items } => (items, SessionSettingsUpdate::default()),
_ => unreachable!(),
};
@@ -2039,18 +1959,18 @@ mod handlers {
};
let skills = if sess.enabled(Feature::Skills) {
let skills_manager = &sess.services.skills_manager;
let mut entries = Vec::new();
for cwd in cwds {
let outcome = skills_manager.skills_for_cwd(&cwd, force_reload).await;
let errors = super::errors_to_info(&outcome.errors);
let skills = super::skills_to_info(&outcome.skills);
entries.push(SkillsListEntry {
cwd,
skills,
errors,
});
}
entries
cwds.into_iter()
.map(|cwd| {
let outcome = skills_manager.skills_for_cwd_with_options(&cwd, force_reload);
let errors = super::errors_to_info(&outcome.errors);
let skills = super::skills_to_info(&outcome.skills);
SkillsListEntry {
cwd,
skills,
errors,
}
})
.collect()
} else {
cwds.into_iter()
.map(|cwd| SkillsListEntry {
@@ -2086,46 +2006,6 @@ mod handlers {
.await;
}
pub async fn thread_rollback(sess: &Arc<Session>, sub_id: String, num_turns: u32) {
if num_turns == 0 {
sess.send_event_raw(Event {
id: sub_id,
msg: EventMsg::Error(ErrorEvent {
message: "num_turns must be >= 1".to_string(),
codex_error_info: Some(CodexErrorInfo::ThreadRollbackFailed),
}),
})
.await;
return;
}
let has_active_turn = { sess.active_turn.lock().await.is_some() };
if has_active_turn {
sess.send_event_raw(Event {
id: sub_id,
msg: EventMsg::Error(ErrorEvent {
message: "Cannot rollback while a turn is in progress.".to_string(),
codex_error_info: Some(CodexErrorInfo::ThreadRollbackFailed),
}),
})
.await;
return;
}
let turn_context = sess.new_default_turn_with_sub_id(sub_id).await;
let mut history = sess.clone_history().await;
history.drop_last_n_user_turns(num_turns);
sess.replace_history(history.get_history()).await;
sess.recompute_token_usage(turn_context.as_ref()).await;
sess.send_event_raw_flushed(Event {
id: turn_context.sub_id.clone(),
msg: EventMsg::ThreadRolledBack(ThreadRolledBackEvent { num_turns }),
})
.await;
}
pub async fn shutdown(sess: &Arc<Session>, sub_id: String) -> bool {
sess.abort_all_tasks(TurnAbortReason::Interrupted).await;
sess.services
@@ -2169,7 +2049,7 @@ mod handlers {
review_request: ReviewRequest,
) {
let turn_context = sess.new_default_turn_with_sub_id(sub_id.clone()).await;
match resolve_review_request(review_request, turn_context.cwd.as_path()) {
match resolve_review_request(review_request, config.cwd.as_path()) {
Ok(resolved) => {
spawn_review_thread(
Arc::clone(sess),
@@ -2212,7 +2092,6 @@ async fn spawn_review_thread(
let mut review_features = sess.features.clone();
review_features
.disable(crate::features::Feature::WebSearchRequest)
.disable(crate::features::Feature::WebSearchCached)
.disable(crate::features::Feature::ViewImageTool);
let tools_config = ToolsConfig::new(&ToolsConfigParams {
model_family: &review_model_family,
@@ -2345,16 +2224,11 @@ pub(crate) async fn run_task(
});
sess.send_event(&turn_context, event).await;
let skills_outcome = if sess.enabled(Feature::Skills) {
Some(
sess.services
.skills_manager
.skills_for_cwd(&turn_context.cwd, false)
.await,
)
} else {
None
};
let skills_outcome = sess.enabled(Feature::Skills).then(|| {
sess.services
.skills_manager
.skills_for_cwd(&turn_context.cwd)
});
let SkillInjections {
items: skill_items,
@@ -2640,22 +2514,8 @@ async fn try_run_turn(
model: turn_context.client.get_model(),
effort: turn_context.client.get_reasoning_effort(),
summary: turn_context.client.get_reasoning_summary(),
base_instructions: turn_context.base_instructions.clone(),
user_instructions: turn_context.user_instructions.clone(),
developer_instructions: turn_context.developer_instructions.clone(),
final_output_json_schema: turn_context.final_output_json_schema.clone(),
truncation_policy: Some(turn_context.truncation_policy.into()),
});
feedback_tags!(
model = turn_context.client.get_model(),
approval_policy = turn_context.approval_policy,
sandbox_policy = turn_context.sandbox_policy,
effort = turn_context.client.get_reasoning_effort(),
auth_mode = sess.services.auth_manager.get_auth_mode(),
features = sess.features.enabled_features(),
);
sess.persist_rollout_items(&[rollout_item]).await;
let mut stream = turn_context
.client
@@ -2746,13 +2606,6 @@ async fn try_run_turn(
// token usage is available to avoid duplicate TokenCount events.
sess.update_rate_limits(&turn_context, snapshot).await;
}
ResponseEvent::ModelsEtag(etag) => {
// Update internal state with latest models etag
sess.services
.models_manager
.refresh_if_new_etag(etag, sess.features.enabled(Feature::RemoteModels))
.await;
}
ResponseEvent::Completed {
response_id: _,
token_usage,
@@ -2893,9 +2746,6 @@ mod tests {
use crate::protocol::RateLimitSnapshot;
use crate::protocol::RateLimitWindow;
use crate::protocol::ResumedHistory;
use crate::protocol::TokenCountEvent;
use crate::protocol::TokenUsage;
use crate::protocol::TokenUsageInfo;
use crate::state::TaskKind;
use crate::tasks::SessionTask;
use crate::tasks::SessionTaskContext;
@@ -2950,83 +2800,6 @@ mod tests {
assert_eq!(expected, actual);
}
#[tokio::test]
async fn record_initial_history_seeds_token_info_from_rollout() {
let (session, turn_context) = make_session_and_context().await;
let (mut rollout_items, _expected) = sample_rollout(&session, &turn_context);
let info1 = TokenUsageInfo {
total_token_usage: TokenUsage {
input_tokens: 10,
cached_input_tokens: 0,
output_tokens: 20,
reasoning_output_tokens: 0,
total_tokens: 30,
},
last_token_usage: TokenUsage {
input_tokens: 3,
cached_input_tokens: 0,
output_tokens: 4,
reasoning_output_tokens: 0,
total_tokens: 7,
},
model_context_window: Some(1_000),
};
let info2 = TokenUsageInfo {
total_token_usage: TokenUsage {
input_tokens: 100,
cached_input_tokens: 50,
output_tokens: 200,
reasoning_output_tokens: 25,
total_tokens: 375,
},
last_token_usage: TokenUsage {
input_tokens: 10,
cached_input_tokens: 0,
output_tokens: 20,
reasoning_output_tokens: 5,
total_tokens: 35,
},
model_context_window: Some(2_000),
};
rollout_items.push(RolloutItem::EventMsg(EventMsg::TokenCount(
TokenCountEvent {
info: Some(info1),
rate_limits: None,
},
)));
rollout_items.push(RolloutItem::EventMsg(EventMsg::TokenCount(
TokenCountEvent {
info: None,
rate_limits: None,
},
)));
rollout_items.push(RolloutItem::EventMsg(EventMsg::TokenCount(
TokenCountEvent {
info: Some(info2.clone()),
rate_limits: None,
},
)));
rollout_items.push(RolloutItem::EventMsg(EventMsg::TokenCount(
TokenCountEvent {
info: None,
rate_limits: None,
},
)));
session
.record_initial_history(InitialHistory::Resumed(ResumedHistory {
conversation_id: ConversationId::default(),
history: rollout_items,
rollout_path: PathBuf::from("/tmp/resume.jsonl"),
}))
.await;
let actual = session.state.lock().await.token_info();
assert_eq!(actual, Some(info2));
}
#[tokio::test]
async fn record_initial_history_reconstructs_forked_transcript() {
let (session, turn_context) = make_session_and_context().await;
@@ -3040,131 +2813,6 @@ mod tests {
assert_eq!(expected, actual);
}
#[tokio::test]
async fn thread_rollback_drops_last_turn_from_history() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref());
sess.record_into_history(&initial_context, tc.as_ref())
.await;
let turn_1 = vec![
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "turn 1 user".to_string(),
}],
},
ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: "turn 1 assistant".to_string(),
}],
},
];
sess.record_into_history(&turn_1, tc.as_ref()).await;
let turn_2 = vec![
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "turn 2 user".to_string(),
}],
},
ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: "turn 2 assistant".to_string(),
}],
},
];
sess.record_into_history(&turn_2, tc.as_ref()).await;
handlers::thread_rollback(&sess, "sub-1".to_string(), 1).await;
let rollback_event = wait_for_thread_rolled_back(&rx).await;
assert_eq!(rollback_event.num_turns, 1);
let mut expected = Vec::new();
expected.extend(initial_context);
expected.extend(turn_1);
let actual = sess.clone_history().await.get_history();
assert_eq!(expected, actual);
}
#[tokio::test]
async fn thread_rollback_clears_history_when_num_turns_exceeds_existing_turns() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref());
sess.record_into_history(&initial_context, tc.as_ref())
.await;
let turn_1 = vec![ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "turn 1 user".to_string(),
}],
}];
sess.record_into_history(&turn_1, tc.as_ref()).await;
handlers::thread_rollback(&sess, "sub-1".to_string(), 99).await;
let rollback_event = wait_for_thread_rolled_back(&rx).await;
assert_eq!(rollback_event.num_turns, 99);
let actual = sess.clone_history().await.get_history();
assert_eq!(initial_context, actual);
}
#[tokio::test]
async fn thread_rollback_fails_when_turn_in_progress() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref());
sess.record_into_history(&initial_context, tc.as_ref())
.await;
*sess.active_turn.lock().await = Some(crate::state::ActiveTurn::default());
handlers::thread_rollback(&sess, "sub-1".to_string(), 1).await;
let error_event = wait_for_thread_rollback_failed(&rx).await;
assert_eq!(
error_event.codex_error_info,
Some(CodexErrorInfo::ThreadRollbackFailed)
);
let actual = sess.clone_history().await.get_history();
assert_eq!(initial_context, actual);
}
#[tokio::test]
async fn thread_rollback_fails_when_num_turns_is_zero() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref());
sess.record_into_history(&initial_context, tc.as_ref())
.await;
handlers::thread_rollback(&sess, "sub-1".to_string(), 0).await;
let error_event = wait_for_thread_rollback_failed(&rx).await;
assert_eq!(error_event.message, "num_turns must be >= 1");
assert_eq!(
error_event.codex_error_info,
Some(CodexErrorInfo::ThreadRollbackFailed)
);
let actual = sess.clone_history().await.get_history();
assert_eq!(initial_context, actual);
}
#[tokio::test]
async fn set_rate_limits_retains_previous_credits() {
let codex_home = tempfile::tempdir().expect("create temp dir");
@@ -3398,44 +3046,6 @@ mod tests {
assert_eq!(expected, got);
}
async fn wait_for_thread_rolled_back(
rx: &async_channel::Receiver<Event>,
) -> crate::protocol::ThreadRolledBackEvent {
let deadline = StdDuration::from_secs(2);
let start = std::time::Instant::now();
loop {
let remaining = deadline.saturating_sub(start.elapsed());
let evt = tokio::time::timeout(remaining, rx.recv())
.await
.expect("timeout waiting for event")
.expect("event");
match evt.msg {
EventMsg::ThreadRolledBack(payload) => return payload,
_ => continue,
}
}
}
async fn wait_for_thread_rollback_failed(rx: &async_channel::Receiver<Event>) -> ErrorEvent {
let deadline = StdDuration::from_secs(2);
let start = std::time::Instant::now();
loop {
let remaining = deadline.saturating_sub(start.elapsed());
let evt = tokio::time::timeout(remaining, rx.recv())
.await
.expect("timeout waiting for event")
.expect("event");
match evt.msg {
EventMsg::Error(payload)
if payload.codex_error_info == Some(CodexErrorInfo::ThreadRollbackFailed) =>
{
return payload;
}
_ => continue,
}
}
}
fn text_block(s: &str) -> ContentBlock {
ContentBlock::TextContent(TextContent {
annotations: None,
@@ -3480,9 +3090,7 @@ mod tests {
let auth_manager =
AuthManager::from_auth_for_testing(CodexAuth::from_api_key("Test API Key"));
let models_manager = Arc::new(ModelsManager::new(auth_manager.clone()));
let agent_control = AgentControl::default();
let exec_policy = ExecPolicyManager::default();
let agent_status = Arc::new(RwLock::new(AgentStatus::PendingInit));
let model = ModelsManager::get_model_offline(config.model.as_deref());
let session_configuration = SessionConfiguration {
provider: config.model_provider.clone(),
@@ -3525,10 +3133,9 @@ mod tests {
exec_policy,
auth_manager: auth_manager.clone(),
otel_manager: otel_manager.clone(),
models_manager: Arc::clone(&models_manager),
models_manager,
tool_approvals: Mutex::new(ApprovalStore::default()),
skills_manager,
agent_control,
};
let turn_context = Session::make_turn_context(
@@ -3545,7 +3152,6 @@ mod tests {
let session = Session {
conversation_id,
tx_event,
agent_status: Arc::clone(&agent_status),
state: Mutex::new(state),
features: config.features.clone(),
active_turn: Mutex::new(None),
@@ -3571,9 +3177,7 @@ mod tests {
let auth_manager =
AuthManager::from_auth_for_testing(CodexAuth::from_api_key("Test API Key"));
let models_manager = Arc::new(ModelsManager::new(auth_manager.clone()));
let agent_control = AgentControl::default();
let exec_policy = ExecPolicyManager::default();
let agent_status = Arc::new(RwLock::new(AgentStatus::PendingInit));
let model = ModelsManager::get_model_offline(config.model.as_deref());
let session_configuration = SessionConfiguration {
provider: config.model_provider.clone(),
@@ -3616,10 +3220,9 @@ mod tests {
exec_policy,
auth_manager: Arc::clone(&auth_manager),
otel_manager: otel_manager.clone(),
models_manager: Arc::clone(&models_manager),
models_manager,
tool_approvals: Mutex::new(ApprovalStore::default()),
skills_manager,
agent_control,
};
let turn_context = Arc::new(Session::make_turn_context(
@@ -3636,7 +3239,6 @@ mod tests {
let session = Arc::new(Session {
conversation_id,
tx_event,
agent_status: Arc::clone(&agent_status),
state: Mutex::new(state),
features: config.features.clone(),
active_turn: Mutex::new(None),

View File

@@ -1,4 +1,3 @@
use crate::agent::AgentStatus;
use crate::codex::Codex;
use crate::error::Result as CodexResult;
use crate::protocol::Event;
@@ -34,10 +33,6 @@ impl CodexConversation {
self.codex.next_event().await
}
pub async fn agent_status(&self) -> AgentStatus {
self.codex.agent_status().await
}
pub fn rollout_path(&self) -> PathBuf {
self.rollout_path.clone()
}

View File

@@ -52,7 +52,6 @@ pub(crate) async fn run_codex_conversation_interactive(
Arc::clone(&parent_session.services.skills_manager),
initial_history.unwrap_or(InitialHistory::New),
SessionSource::SubAgent(SubAgentSource::Review),
parent_session.services.agent_control.clone(),
)
.await?;
let codex = Arc::new(codex);
@@ -87,7 +86,6 @@ pub(crate) async fn run_codex_conversation_interactive(
next_id: AtomicU64::new(0),
tx_sub: tx_ops,
rx_event: rx_sub,
agent_status: Arc::clone(&codex.agent_status),
})
}
@@ -120,16 +118,11 @@ pub(crate) async fn run_codex_conversation_one_shot(
.await?;
// Send the initial input to kick off the one-shot turn.
io.submit(Op::UserInput {
items: input,
final_output_json_schema: None,
})
.await?;
io.submit(Op::UserInput { items: input }).await?;
// Bridge events so we can observe completion and shut down automatically.
let (tx_bridge, rx_bridge) = async_channel::bounded(SUBMISSION_CHANNEL_CAPACITY);
let ops_tx = io.tx_sub.clone();
let agent_status = Arc::clone(&io.agent_status);
let io_for_bridge = io;
tokio::spawn(async move {
while let Ok(event) = io_for_bridge.next_event().await {
@@ -161,7 +154,6 @@ pub(crate) async fn run_codex_conversation_one_shot(
next_id: AtomicU64::new(0),
rx_event: rx_bridge,
tx_sub: tx_closed,
agent_status,
})
}
@@ -192,10 +184,6 @@ async fn forward_events(
id: _,
msg: EventMsg::AgentMessageDelta(_) | EventMsg::AgentReasoningDelta(_),
} => {}
Event {
id: _,
msg: EventMsg::TokenCount(_),
} => {}
Event {
id: _,
msg: EventMsg::SessionConfigured(_),
@@ -376,7 +364,6 @@ mod tests {
next_id: AtomicU64::new(0),
tx_sub,
rx_event: rx_events,
agent_status: Default::default(),
});
let (session, ctx, _rx_evt) = crate::codex::make_session_and_context_with_rx().await;

View File

@@ -86,11 +86,6 @@ async fn run_compact_task_inner(
model: turn_context.client.get_model(),
effort: turn_context.client.get_reasoning_effort(),
summary: turn_context.client.get_reasoning_summary(),
base_instructions: turn_context.base_instructions.clone(),
user_instructions: turn_context.user_instructions.clone(),
developer_instructions: turn_context.developer_instructions.clone(),
final_output_json_schema: turn_context.final_output_json_schema.clone(),
truncation_policy: Some(turn_context.truncation_policy.into()),
});
sess.persist_rollout_items(&[rollout_item]).await;

View File

@@ -13,7 +13,6 @@ use crate::config::types::ShellEnvironmentPolicy;
use crate::config::types::ShellEnvironmentPolicyToml;
use crate::config::types::Tui;
use crate::config::types::UriBasedFileOpener;
use crate::config_loader::ConfigLayerStack;
use crate::config_loader::ConfigRequirements;
use crate::config_loader::LoaderOverrides;
use crate::config_loader::load_config_layers_state;
@@ -38,6 +37,7 @@ use codex_protocol::config_types::SandboxMode;
use codex_protocol::config_types::TrustLevel;
use codex_protocol::config_types::Verbosity;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::openai_models::ReasoningSummaryFormat;
use codex_rmcp_client::OAuthCredentialsStoreMode;
use codex_utils_absolute_path::AbsolutePathBuf;
use codex_utils_absolute_path::AbsolutePathBufGuard;
@@ -94,10 +94,6 @@ pub(crate) fn test_config() -> Config {
/// Application configuration loaded from disk and merged with overrides.
#[derive(Debug, Clone, PartialEq)]
pub struct Config {
/// Provenance for how this [`Config`] was derived (merged layers + enforced
/// requirements).
pub config_layer_stack: ConfigLayerStack,
/// Optional override of model selection.
pub model: Option<String>,
@@ -302,6 +298,9 @@ pub struct Config {
/// Optional override to force-enable reasoning summaries for the configured model.
pub model_supports_reasoning_summaries: Option<bool>,
/// Optional override to force reasoning summary format for the configured model.
pub model_reasoning_summary_format: Option<ReasoningSummaryFormat>,
/// Optional verbosity control for GPT-5 models (Responses API `text.verbosity`).
pub model_verbosity: Option<Verbosity>,
@@ -353,10 +352,6 @@ pub struct Config {
/// or placeholder replacement will occur for fast keypress bursts.
pub disable_paste_burst: bool,
/// When `false`, disables analytics across Codex product surfaces in this machine.
/// Defaults to `true`.
pub analytics: bool,
/// OTEL configuration (exporter type, endpoint, headers, etc.).
pub otel: crate::config::types::OtelConfig,
}
@@ -417,11 +412,11 @@ impl ConfigBuilder {
let config_toml: ConfigToml = merged_toml
.try_into()
.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e))?;
Config::load_config_with_layer_stack(
Config::load_config_with_requirements(
config_toml,
harness_overrides,
codex_home,
config_layer_stack,
config_layer_stack.requirements().clone(),
)
}
}
@@ -786,6 +781,9 @@ pub struct ConfigToml {
/// Override to force-enable reasoning summaries for the configured model.
pub model_supports_reasoning_summaries: Option<bool>,
/// Override to force reasoning summary format for the configured model.
pub model_reasoning_summary_format: Option<ReasoningSummaryFormat>,
/// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
pub chatgpt_base_url: Option<String>,
@@ -817,10 +815,6 @@ pub struct ConfigToml {
/// or placeholder replacement will occur for fast keypress bursts.
pub disable_paste_burst: Option<bool>,
/// When `false`, disables analytics across Codex product surfaces in this machine.
/// Defaults to `true`.
pub analytics: Option<crate::config::types::AnalyticsConfigToml>,
/// OTEL configuration.
pub otel: Option<crate::config::types::OtelConfigToml>,
@@ -1074,17 +1068,16 @@ impl Config {
codex_home: PathBuf,
) -> std::io::Result<Self> {
// Note this ignores requirements.toml enforcement for tests.
let config_layer_stack = ConfigLayerStack::default();
Self::load_config_with_layer_stack(cfg, overrides, codex_home, config_layer_stack)
let requirements = ConfigRequirements::default();
Self::load_config_with_requirements(cfg, overrides, codex_home, requirements)
}
fn load_config_with_layer_stack(
fn load_config_with_requirements(
cfg: ConfigToml,
overrides: ConfigOverrides,
codex_home: PathBuf,
config_layer_stack: ConfigLayerStack,
requirements: ConfigRequirements,
) -> std::io::Result<Self> {
let requirements = config_layer_stack.requirements().clone();
let user_instructions = Self::load_instructions(Some(&codex_home));
// Destructure ConfigOverrides fully to ensure all overrides are applied.
@@ -1362,7 +1355,6 @@ impl Config {
.collect(),
tool_output_token_limit: cfg.tool_output_token_limit,
codex_home,
config_layer_stack,
history,
file_opener: cfg.file_opener.unwrap_or(UriBasedFileOpener::VsCode),
codex_linux_sandbox_exe,
@@ -1380,6 +1372,7 @@ impl Config {
.or(cfg.model_reasoning_summary)
.unwrap_or_default(),
model_supports_reasoning_summaries: cfg.model_supports_reasoning_summaries,
model_reasoning_summary_format: cfg.model_reasoning_summary_format.clone(),
model_verbosity: config_profile.model_verbosity.or(cfg.model_verbosity),
chatgpt_base_url: config_profile
.chatgpt_base_url
@@ -1398,12 +1391,6 @@ impl Config {
notices: cfg.notice.unwrap_or_default(),
check_for_update_on_startup,
disable_paste_burst: cfg.disable_paste_burst.unwrap_or(false),
analytics: config_profile
.analytics
.as_ref()
.and_then(|a| a.enabled)
.or(cfg.analytics.as_ref().and_then(|a| a.enabled))
.unwrap_or(true),
tui_notifications: cfg
.tui
.as_ref()
@@ -2064,7 +2051,6 @@ trust_level = "trusted"
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
};
let cwd = AbsolutePathBuf::try_from(codex_home.path())?;
@@ -2185,7 +2171,6 @@ trust_level = "trusted"
managed_config_path: Some(managed_path),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
};
let cwd = AbsolutePathBuf::try_from(codex_home.path())?;
@@ -3053,9 +3038,6 @@ approval_policy = "untrusted"
# `ConfigOverrides`.
profile = "gpt3"
[analytics]
enabled = true
[model_providers.openai-chat-completions]
name = "OpenAI using Chat Completions"
base_url = "https://api.openai.com/v1"
@@ -3081,9 +3063,6 @@ model = "o3"
model_provider = "openai"
approval_policy = "on-failure"
[profiles.zdr.analytics]
enabled = false
[profiles.gpt5]
model = "gpt-5.1"
model_provider = "openai"
@@ -3194,7 +3173,6 @@ model_verbosity = "high"
project_doc_fallback_filenames: Vec::new(),
tool_output_token_limit: None,
codex_home: fixture.codex_home(),
config_layer_stack: Default::default(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
codex_linux_sandbox_exe: None,
@@ -3203,6 +3181,7 @@ model_verbosity = "high"
model_reasoning_effort: Some(ReasoningEffort::High),
model_reasoning_summary: ReasoningSummary::Detailed,
model_supports_reasoning_summaries: None,
model_reasoning_summary_format: None,
model_verbosity: None,
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
base_instructions: None,
@@ -3224,7 +3203,6 @@ model_verbosity = "high"
tui_notifications: Default::default(),
animations: true,
show_tooltips: true,
analytics: true,
tui_scroll_events_per_tick: None,
tui_scroll_wheel_lines: None,
tui_scroll_trackpad_lines: None,
@@ -3278,7 +3256,6 @@ model_verbosity = "high"
project_doc_fallback_filenames: Vec::new(),
tool_output_token_limit: None,
codex_home: fixture.codex_home(),
config_layer_stack: Default::default(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
codex_linux_sandbox_exe: None,
@@ -3287,6 +3264,7 @@ model_verbosity = "high"
model_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::default(),
model_supports_reasoning_summaries: None,
model_reasoning_summary_format: None,
model_verbosity: None,
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
base_instructions: None,
@@ -3308,7 +3286,6 @@ model_verbosity = "high"
tui_notifications: Default::default(),
animations: true,
show_tooltips: true,
analytics: true,
tui_scroll_events_per_tick: None,
tui_scroll_wheel_lines: None,
tui_scroll_trackpad_lines: None,
@@ -3377,7 +3354,6 @@ model_verbosity = "high"
project_doc_fallback_filenames: Vec::new(),
tool_output_token_limit: None,
codex_home: fixture.codex_home(),
config_layer_stack: Default::default(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
codex_linux_sandbox_exe: None,
@@ -3386,6 +3362,7 @@ model_verbosity = "high"
model_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::default(),
model_supports_reasoning_summaries: None,
model_reasoning_summary_format: None,
model_verbosity: None,
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
base_instructions: None,
@@ -3407,7 +3384,6 @@ model_verbosity = "high"
tui_notifications: Default::default(),
animations: true,
show_tooltips: true,
analytics: false,
tui_scroll_events_per_tick: None,
tui_scroll_wheel_lines: None,
tui_scroll_trackpad_lines: None,
@@ -3462,7 +3438,6 @@ model_verbosity = "high"
project_doc_fallback_filenames: Vec::new(),
tool_output_token_limit: None,
codex_home: fixture.codex_home(),
config_layer_stack: Default::default(),
history: History::default(),
file_opener: UriBasedFileOpener::VsCode,
codex_linux_sandbox_exe: None,
@@ -3471,6 +3446,7 @@ model_verbosity = "high"
model_reasoning_effort: Some(ReasoningEffort::High),
model_reasoning_summary: ReasoningSummary::Detailed,
model_supports_reasoning_summaries: None,
model_reasoning_summary_format: None,
model_verbosity: Some(Verbosity::High),
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
base_instructions: None,
@@ -3492,7 +3468,6 @@ model_verbosity = "high"
tui_notifications: Default::default(),
animations: true,
show_tooltips: true,
analytics: true,
tui_scroll_events_per_tick: None,
tui_scroll_wheel_lines: None,
tui_scroll_trackpad_lines: None,

View File

@@ -29,7 +29,6 @@ pub struct ConfigProfile {
pub experimental_use_freeform_apply_patch: Option<bool>,
pub tools_web_search: Option<bool>,
pub tools_view_image: Option<bool>,
pub analytics: Option<crate::config::types::AnalyticsConfigToml>,
/// Optional feature toggles scoped to this profile.
#[serde(default)]
pub features: Option<crate::features::FeaturesToml>,

View File

@@ -106,7 +106,16 @@ pub struct ConfigService {
}
impl ConfigService {
pub fn new(
pub fn new(codex_home: PathBuf, cli_overrides: Vec<(String, TomlValue)>) -> Self {
Self {
codex_home,
cli_overrides,
loader_overrides: LoaderOverrides::default(),
}
}
#[cfg(test)]
fn with_overrides(
codex_home: PathBuf,
cli_overrides: Vec<(String, TomlValue)>,
loader_overrides: LoaderOverrides,
@@ -118,14 +127,6 @@ impl ConfigService {
}
}
pub fn new_with_defaults(codex_home: PathBuf) -> Self {
Self {
codex_home,
cli_overrides: Vec::new(),
loader_overrides: LoaderOverrides::default(),
}
}
pub async fn read(
&self,
params: ConfigReadParams,
@@ -706,7 +707,7 @@ unified_exec = true
"#;
std::fs::write(tmp.path().join(CONFIG_TOML_FILE), original)?;
let service = ConfigService::new_with_defaults(tmp.path().to_path_buf());
let service = ConfigService::new(tmp.path().to_path_buf(), vec![]);
service
.write_value(ConfigValueWriteParams {
file_path: Some(tmp.path().join(CONFIG_TOML_FILE).display().to_string()),
@@ -747,14 +748,13 @@ remote_compaction = true
std::fs::write(&managed_path, "approval_policy = \"never\"").unwrap();
let managed_file = AbsolutePathBuf::try_from(managed_path.clone()).expect("managed file");
let service = ConfigService::new(
let service = ConfigService::with_overrides(
tmp.path().to_path_buf(),
vec![],
LoaderOverrides {
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
},
);
@@ -778,41 +778,15 @@ remote_compaction = true
},
);
let layers = response.layers.expect("layers present");
if cfg!(unix) {
let system_file = AbsolutePathBuf::from_absolute_path(
crate::config_loader::SYSTEM_CONFIG_TOML_FILE_UNIX,
)
.expect("system file");
assert_eq!(layers.len(), 3, "expected three layers on unix");
assert_eq!(
layers.first().unwrap().name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile {
file: managed_file.clone()
}
);
assert_eq!(
layers.get(1).unwrap().name,
ConfigLayerSource::User {
file: user_file.clone()
}
);
assert_eq!(
layers.get(2).unwrap().name,
ConfigLayerSource::System { file: system_file }
);
} else {
assert_eq!(layers.len(), 2, "expected two layers");
assert_eq!(
layers.first().unwrap().name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile {
file: managed_file.clone()
}
);
assert_eq!(
layers.get(1).unwrap().name,
ConfigLayerSource::User { file: user_file }
);
}
assert_eq!(layers.len(), 2, "expected two layers");
assert_eq!(
layers.first().unwrap().name,
ConfigLayerSource::LegacyManagedConfigTomlFromFile { file: managed_file }
);
assert_eq!(
layers.get(1).unwrap().name,
ConfigLayerSource::User { file: user_file }
);
}
#[tokio::test]
@@ -828,14 +802,13 @@ remote_compaction = true
std::fs::write(&managed_path, "approval_policy = \"never\"").unwrap();
let managed_file = AbsolutePathBuf::try_from(managed_path.clone()).expect("managed file");
let service = ConfigService::new(
let service = ConfigService::with_overrides(
tmp.path().to_path_buf(),
vec![],
LoaderOverrides {
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
},
);
@@ -880,7 +853,7 @@ remote_compaction = true
let user_path = tmp.path().join(CONFIG_TOML_FILE);
std::fs::write(&user_path, "model = \"user\"").unwrap();
let service = ConfigService::new_with_defaults(tmp.path().to_path_buf());
let service = ConfigService::new(tmp.path().to_path_buf(), vec![]);
let error = service
.write_value(ConfigValueWriteParams {
file_path: Some(tmp.path().join(CONFIG_TOML_FILE).display().to_string()),
@@ -903,7 +876,7 @@ remote_compaction = true
let tmp = tempdir().expect("tempdir");
std::fs::write(tmp.path().join(CONFIG_TOML_FILE), "").unwrap();
let service = ConfigService::new_with_defaults(tmp.path().to_path_buf());
let service = ConfigService::new(tmp.path().to_path_buf(), vec![]);
service
.write_value(ConfigValueWriteParams {
file_path: None,
@@ -931,14 +904,13 @@ remote_compaction = true
let managed_path = tmp.path().join("managed_config.toml");
std::fs::write(&managed_path, "approval_policy = \"never\"").unwrap();
let service = ConfigService::new(
let service = ConfigService::with_overrides(
tmp.path().to_path_buf(),
vec![],
LoaderOverrides {
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
},
);
@@ -979,14 +951,13 @@ remote_compaction = true
TomlValue::String("session".to_string()),
)];
let service = ConfigService::new(
let service = ConfigService::with_overrides(
tmp.path().to_path_buf(),
cli_overrides,
LoaderOverrides {
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
},
);
@@ -1025,14 +996,13 @@ remote_compaction = true
std::fs::write(&managed_path, "approval_policy = \"never\"").unwrap();
let managed_file = AbsolutePathBuf::try_from(managed_path.clone()).expect("managed file");
let service = ConfigService::new(
let service = ConfigService::with_overrides(
tmp.path().to_path_buf(),
vec![],
LoaderOverrides {
managed_config_path: Some(managed_path.clone()),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
},
);
@@ -1084,7 +1054,7 @@ alpha = "a"
std::fs::write(&path, base)?;
let service = ConfigService::new_with_defaults(tmp.path().to_path_buf());
let service = ConfigService::new(tmp.path().to_path_buf(), vec![]);
service
.write_value(ConfigValueWriteParams {
file_path: Some(path.display().to_string()),

View File

@@ -273,15 +273,6 @@ pub enum HistoryPersistence {
None,
}
// ===== Analytics configuration =====
/// Analytics settings loaded from config.toml. Fields are optional so we can apply defaults.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default)]
pub struct AnalyticsConfigToml {
/// When `false`, disables analytics across Codex product surfaces in this profile.
pub enabled: Option<bool>,
}
// ===== OTEL configuration =====
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]

View File

@@ -33,13 +33,11 @@ pub(super) async fn load_config_layers_internal(
let LoaderOverrides {
managed_config_path,
managed_preferences_base64,
..
} = overrides;
#[cfg(not(target_os = "macos"))]
let LoaderOverrides {
managed_config_path,
..
} = overrides;
let managed_config_path = AbsolutePathBuf::from_absolute_path(
@@ -93,8 +91,12 @@ pub(super) async fn read_config_from_path(
}
}
/// Return the default managed config path.
/// Return the default managed config path (honoring `CODEX_MANAGED_CONFIG_PATH`).
pub(super) fn managed_config_default_path(codex_home: &Path) -> PathBuf {
if let Ok(path) = std::env::var("CODEX_MANAGED_CONFIG_PATH") {
return PathBuf::from(path);
}
#[cfg(unix)]
{
let _ = codex_home;

View File

@@ -1,4 +1,3 @@
use super::config_requirements::ConfigRequirementsToml;
use base64::Engine;
use base64::prelude::BASE64_STANDARD;
use core_foundation::base::TCFType;
@@ -11,7 +10,6 @@ use toml::Value as TomlValue;
const MANAGED_PREFERENCES_APPLICATION_ID: &str = "com.openai.codex";
const MANAGED_PREFERENCES_CONFIG_KEY: &str = "config_toml_base64";
const MANAGED_PREFERENCES_REQUIREMENTS_KEY: &str = "requirements_toml_base64";
pub(crate) async fn load_managed_admin_config_layer(
override_base64: Option<&str>,
@@ -21,126 +19,82 @@ pub(crate) async fn load_managed_admin_config_layer(
return if trimmed.is_empty() {
Ok(None)
} else {
parse_managed_config_base64(trimmed).map(Some)
parse_managed_preferences_base64(trimmed).map(Some)
};
}
const LOAD_ERROR: &str = "Failed to load managed preferences configuration";
match task::spawn_blocking(load_managed_admin_config).await {
Ok(result) => result,
Err(join_err) => {
if join_err.is_cancelled() {
tracing::error!("Managed config load task was cancelled");
tracing::error!("Managed preferences load task was cancelled");
} else {
tracing::error!("Managed config load task failed: {join_err}");
tracing::error!("Managed preferences load task failed: {join_err}");
}
Err(io::Error::other("Failed to load managed config"))
Err(io::Error::other(LOAD_ERROR))
}
}
}
fn load_managed_admin_config() -> io::Result<Option<TomlValue>> {
load_managed_preference(MANAGED_PREFERENCES_CONFIG_KEY)?
.as_deref()
.map(str::trim)
.map(parse_managed_config_base64)
.transpose()
}
pub(crate) async fn load_managed_admin_requirements_toml(
target: &mut ConfigRequirementsToml,
override_base64: Option<&str>,
) -> io::Result<()> {
if let Some(encoded) = override_base64 {
let trimmed = encoded.trim();
if !trimmed.is_empty() {
target.merge_unset_fields(parse_managed_requirements_base64(trimmed)?);
}
return Ok(());
}
match task::spawn_blocking(load_managed_admin_requirements).await {
Ok(result) => {
if let Some(requirements) = result? {
target.merge_unset_fields(requirements);
}
Ok(())
}
Err(join_err) => {
if join_err.is_cancelled() {
tracing::error!("Managed requirements load task was cancelled");
} else {
tracing::error!("Managed requirements load task failed: {join_err}");
}
Err(io::Error::other("Failed to load managed requirements"))
}
}
}
fn load_managed_admin_requirements() -> io::Result<Option<ConfigRequirementsToml>> {
load_managed_preference(MANAGED_PREFERENCES_REQUIREMENTS_KEY)?
.as_deref()
.map(str::trim)
.map(parse_managed_requirements_base64)
.transpose()
}
fn load_managed_preference(key_name: &str) -> io::Result<Option<String>> {
#[link(name = "CoreFoundation", kind = "framework")]
unsafe extern "C" {
fn CFPreferencesCopyAppValue(key: CFStringRef, application_id: CFStringRef) -> *mut c_void;
}
let application_id = CFString::new(MANAGED_PREFERENCES_APPLICATION_ID);
let key = CFString::new(MANAGED_PREFERENCES_CONFIG_KEY);
let value_ref = unsafe {
CFPreferencesCopyAppValue(
CFString::new(key_name).as_concrete_TypeRef(),
CFString::new(MANAGED_PREFERENCES_APPLICATION_ID).as_concrete_TypeRef(),
key.as_concrete_TypeRef(),
application_id.as_concrete_TypeRef(),
)
};
if value_ref.is_null() {
tracing::debug!(
"Managed preferences for {MANAGED_PREFERENCES_APPLICATION_ID} key {key_name} not found",
"Managed preferences for {} key {} not found",
MANAGED_PREFERENCES_APPLICATION_ID,
MANAGED_PREFERENCES_CONFIG_KEY
);
return Ok(None);
}
let value = unsafe { CFString::wrap_under_create_rule(value_ref as _) }.to_string();
Ok(Some(value))
let value = unsafe { CFString::wrap_under_create_rule(value_ref as _) };
let contents = value.to_string();
let trimmed = contents.trim();
parse_managed_preferences_base64(trimmed).map(Some)
}
fn parse_managed_config_base64(encoded: &str) -> io::Result<TomlValue> {
match toml::from_str::<TomlValue>(&decode_managed_preferences_base64(encoded)?) {
fn parse_managed_preferences_base64(encoded: &str) -> io::Result<TomlValue> {
let decoded = BASE64_STANDARD.decode(encoded.as_bytes()).map_err(|err| {
tracing::error!("Failed to decode managed preferences as base64: {err}");
io::Error::new(io::ErrorKind::InvalidData, err)
})?;
let decoded_str = String::from_utf8(decoded).map_err(|err| {
tracing::error!("Managed preferences base64 contents were not valid UTF-8: {err}");
io::Error::new(io::ErrorKind::InvalidData, err)
})?;
match toml::from_str::<TomlValue>(&decoded_str) {
Ok(TomlValue::Table(parsed)) => Ok(TomlValue::Table(parsed)),
Ok(other) => {
tracing::error!("Managed config TOML must have a table at the root, found {other:?}",);
tracing::error!(
"Managed preferences TOML must have a table at the root, found {other:?}",
);
Err(io::Error::new(
io::ErrorKind::InvalidData,
"managed config root must be a table",
"managed preferences root must be a table",
))
}
Err(err) => {
tracing::error!("Failed to parse managed config TOML: {err}");
tracing::error!("Failed to parse managed preferences TOML: {err}");
Err(io::Error::new(io::ErrorKind::InvalidData, err))
}
}
}
fn parse_managed_requirements_base64(encoded: &str) -> io::Result<ConfigRequirementsToml> {
toml::from_str::<ConfigRequirementsToml>(&decode_managed_preferences_base64(encoded)?).map_err(
|err| {
tracing::error!("Failed to parse managed requirements TOML: {err}");
io::Error::new(io::ErrorKind::InvalidData, err)
},
)
}
fn decode_managed_preferences_base64(encoded: &str) -> io::Result<String> {
String::from_utf8(BASE64_STANDARD.decode(encoded.as_bytes()).map_err(|err| {
tracing::error!("Failed to decode managed value as base64: {err}",);
io::Error::new(io::ErrorKind::InvalidData, err)
})?)
.map_err(|err| {
tracing::error!("Managed value base64 contents were not valid UTF-8: {err}",);
io::Error::new(io::ErrorKind::InvalidData, err)
})
}

View File

@@ -28,17 +28,10 @@ pub use config_requirements::ConfigRequirements;
pub use merge::merge_toml_values;
pub use state::ConfigLayerEntry;
pub use state::ConfigLayerStack;
pub use state::ConfigLayerStackOrdering;
pub use state::LoaderOverrides;
/// On Unix systems, load requirements from this file path, if present.
const DEFAULT_REQUIREMENTS_TOML_FILE_UNIX: &str = "/etc/codex/requirements.toml";
/// On Unix systems, load default settings from this file path, if present.
/// Note that /etc/codex/ is treated as a "config folder," so subfolders such
/// as skills/ and rules/ will also be honored.
pub const SYSTEM_CONFIG_TOML_FILE_UNIX: &str = "/etc/codex/config.toml";
const DEFAULT_PROJECT_ROOT_MARKERS: &[&str] = &[".git"];
/// To build up the set of admin-enforced constraints, we build up from multiple
@@ -78,14 +71,8 @@ pub async fn load_config_layers_state(
) -> io::Result<ConfigLayerStack> {
let mut config_requirements_toml = ConfigRequirementsToml::default();
#[cfg(target_os = "macos")]
macos::load_managed_admin_requirements_toml(
&mut config_requirements_toml,
overrides
.macos_managed_config_requirements_base64
.as_deref(),
)
.await?;
// TODO(mbolin): Support an entry in MDM for config requirements and use it
// with `config_requirements_toml.merge_unset_fields(...)`, if present.
// Honor /etc/codex/requirements.toml.
if cfg!(unix) {
@@ -107,44 +94,52 @@ pub async fn load_config_layers_state(
let mut layers = Vec::<ConfigLayerEntry>::new();
// Include an entry for the "system" config folder, loading its config.toml,
// if it exists.
let system_config_toml_file = if cfg!(unix) {
Some(AbsolutePathBuf::from_absolute_path(
SYSTEM_CONFIG_TOML_FILE_UNIX,
)?)
} else {
// TODO(gt): Determine the path to load on Windows.
None
};
if let Some(system_config_toml_file) = system_config_toml_file {
let system_layer =
load_config_toml_for_required_layer(&system_config_toml_file, |config_toml| {
ConfigLayerEntry::new(
ConfigLayerSource::System {
file: system_config_toml_file.clone(),
},
config_toml,
)
})
.await?;
layers.push(system_layer);
}
// TODO(mbolin): Honor managed preferences (macOS only).
// TODO(mbolin): Honor /etc/codex/config.toml.
// Add a layer for $CODEX_HOME/config.toml if it exists. Note if the file
// exists, but is malformed, then this error should be propagated to the
// user.
let user_file = AbsolutePathBuf::resolve_path_against_base(CONFIG_TOML_FILE, codex_home)?;
let user_layer = load_config_toml_for_required_layer(&user_file, |config_toml| {
ConfigLayerEntry::new(
ConfigLayerSource::User {
file: user_file.clone(),
},
config_toml,
)
})
.await?;
layers.push(user_layer);
match tokio::fs::read_to_string(&user_file).await {
Ok(contents) => {
let user_config: TomlValue = toml::from_str(&contents).map_err(|e| {
io::Error::new(
io::ErrorKind::InvalidData,
format!(
"Error parsing user config file {}: {e}",
user_file.as_path().display(),
),
)
})?;
let user_config_parent = user_file.as_path().parent().ok_or_else(|| {
io::Error::new(
io::ErrorKind::InvalidData,
format!(
"User config file {} has no parent directory",
user_file.as_path().display()
),
)
})?;
let user_config =
resolve_relative_paths_in_config_toml(user_config, user_config_parent)?;
layers.push(ConfigLayerEntry::new(
ConfigLayerSource::User { file: user_file },
user_config,
));
}
Err(e) => {
if e.kind() != io::ErrorKind::NotFound {
return Err(io::Error::new(
e.kind(),
format!(
"Failed to read user config file {}: {e}",
user_file.as_path().display(),
),
));
}
}
}
if let Some(cwd) = cwd {
let mut merged_so_far = TomlValue::Table(toml::map::Map::new());
@@ -204,52 +199,6 @@ pub async fn load_config_layers_state(
ConfigLayerStack::new(layers, config_requirements_toml.try_into()?)
}
/// Attempts to load a config.toml file from `config_toml`.
/// - If the file exists and is valid TOML, passes the parsed `toml::Value` to
/// `create_entry` and returns the resulting layer entry.
/// - If the file does not exist, uses an empty `Table` with `create_entry` and
/// returns the resulting layer entry.
/// - If there is an error reading the file or parsing the TOML, returns an
/// error.
async fn load_config_toml_for_required_layer(
config_toml: impl AsRef<Path>,
create_entry: impl FnOnce(TomlValue) -> ConfigLayerEntry,
) -> io::Result<ConfigLayerEntry> {
let toml_file = config_toml.as_ref();
let toml_value = match tokio::fs::read_to_string(toml_file).await {
Ok(contents) => {
let config: TomlValue = toml::from_str(&contents).map_err(|e| {
io::Error::new(
io::ErrorKind::InvalidData,
format!("Error parsing config file {}: {e}", toml_file.display()),
)
})?;
let config_parent = toml_file.parent().ok_or_else(|| {
io::Error::new(
io::ErrorKind::InvalidData,
format!(
"Config file {} has no parent directory",
toml_file.display()
),
)
})?;
resolve_relative_paths_in_config_toml(config, config_parent)
}
Err(e) => {
if e.kind() == io::ErrorKind::NotFound {
Ok(TomlValue::Table(toml::map::Map::new()))
} else {
Err(io::Error::new(
e.kind(),
format!("Failed to read config file {}: {e}", toml_file.display()),
))
}
}
}?;
Ok(create_entry(toml_value))
}
/// If available, apply requirements from `/etc/codex/requirements.toml` to
/// `config_requirements_toml` by filling in any unset fields.
async fn load_requirements_toml(

View File

@@ -12,14 +12,11 @@ use std::collections::HashMap;
use std::path::PathBuf;
use toml::Value as TomlValue;
/// LoaderOverrides overrides managed configuration inputs (primarily for tests).
#[derive(Debug, Default, Clone)]
pub struct LoaderOverrides {
pub managed_config_path: Option<PathBuf>,
//TODO(gt): Add a macos_ prefix to this field and remove the target_os check.
#[cfg(target_os = "macos")]
pub managed_preferences_base64: Option<String>,
pub macos_managed_config_requirements_base64: Option<String>,
}
#[derive(Debug, Clone, PartialEq)]
@@ -53,28 +50,9 @@ impl ConfigLayerEntry {
config: serde_json::to_value(&self.config).unwrap_or(JsonValue::Null),
}
}
// Get the `.codex/` folder associated with this config layer, if any.
pub fn config_folder(&self) -> Option<AbsolutePathBuf> {
match &self.name {
ConfigLayerSource::Mdm { .. } => None,
ConfigLayerSource::System { file } => file.parent(),
ConfigLayerSource::User { file } => file.parent(),
ConfigLayerSource::Project { dot_codex_folder } => Some(dot_codex_folder.clone()),
ConfigLayerSource::SessionFlags => None,
ConfigLayerSource::LegacyManagedConfigTomlFromFile { .. } => None,
ConfigLayerSource::LegacyManagedConfigTomlFromMdm => None,
}
}
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum ConfigLayerStackOrdering {
LowestPrecedenceFirst,
HighestPrecedenceFirst,
}
#[derive(Debug, Clone, Default, PartialEq)]
#[derive(Debug, Clone)]
pub struct ConfigLayerStack {
/// Layers are listed from lowest precedence (base) to highest (top), so
/// later entries in the Vec override earlier ones.
@@ -178,16 +156,7 @@ impl ConfigLayerStack {
/// Returns the highest-precedence to lowest-precedence layers, so
/// `ConfigLayerSource::SessionFlags` would be first, if present.
pub fn layers_high_to_low(&self) -> Vec<&ConfigLayerEntry> {
self.get_layers(ConfigLayerStackOrdering::HighestPrecedenceFirst)
}
/// Returns the highest-precedence to lowest-precedence layers, so
/// `ConfigLayerSource::SessionFlags` would be first, if present.
pub fn get_layers(&self, ordering: ConfigLayerStackOrdering) -> Vec<&ConfigLayerEntry> {
match ordering {
ConfigLayerStackOrdering::HighestPrecedenceFirst => self.layers.iter().rev().collect(),
ConfigLayerStackOrdering::LowestPrecedenceFirst => self.layers.iter().collect(),
}
self.layers.iter().rev().collect()
}
}

View File

@@ -9,8 +9,6 @@ use crate::config_loader::config_requirements::ConfigRequirementsToml;
use crate::config_loader::fingerprint::version_for_toml;
use crate::config_loader::load_requirements_toml;
use codex_protocol::protocol::AskForApproval;
#[cfg(target_os = "macos")]
use codex_protocol::protocol::SandboxPolicy;
use codex_utils_absolute_path::AbsolutePathBuf;
use pretty_assertions::assert_eq;
use tempfile::tempdir;
@@ -45,7 +43,6 @@ extra = true
managed_config_path: Some(managed_path),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
};
let cwd = AbsolutePathBuf::try_from(tmp.path()).expect("cwd");
@@ -76,12 +73,10 @@ extra = true
async fn returns_empty_when_all_layers_missing() {
let tmp = tempdir().expect("tempdir");
let managed_path = tmp.path().join("managed_config.toml");
let overrides = LoaderOverrides {
managed_config_path: Some(managed_path),
#[cfg(target_os = "macos")]
managed_preferences_base64: None,
macos_managed_config_requirements_base64: None,
};
let cwd = AbsolutePathBuf::try_from(tmp.path()).expect("cwd");
@@ -93,24 +88,9 @@ async fn returns_empty_when_all_layers_missing() {
)
.await
.expect("load layers");
let user_layer = layers
.get_user_layer()
.expect("expected a user layer even when CODEX_HOME/config.toml does not exist");
assert_eq!(
&ConfigLayerEntry {
name: super::ConfigLayerSource::User {
file: AbsolutePathBuf::resolve_path_against_base(CONFIG_TOML_FILE, tmp.path())
.expect("resolve user config.toml path")
},
config: TomlValue::Table(toml::map::Map::new()),
version: version_for_toml(&TomlValue::Table(toml::map::Map::new())),
},
user_layer,
);
assert_eq!(
user_layer.config,
TomlValue::Table(toml::map::Map::new()),
"expected empty config for user layer when config.toml does not exist"
assert!(
layers.get_user_layer().is_none(),
"no user layer when CODEX_HOME/config.toml does not exist"
);
let binding = layers.effective_config();
@@ -124,10 +104,9 @@ async fn returns_empty_when_all_layers_missing() {
.iter()
.filter(|layer| matches!(layer.name, super::ConfigLayerSource::System { .. }))
.count();
let expected_system_layers = if cfg!(unix) { 1 } else { 0 };
assert_eq!(
num_system_layers, expected_system_layers,
"system layer should be present only on unix"
num_system_layers, 0,
"managed config layer should be absent when file missing"
);
#[cfg(not(target_os = "macos"))]
@@ -146,6 +125,12 @@ async fn returns_empty_when_all_layers_missing() {
async fn managed_preferences_take_highest_precedence() {
use base64::Engine;
let managed_payload = r#"
[nested]
value = "managed"
flag = false
"#;
let encoded = base64::prelude::BASE64_STANDARD.encode(managed_payload.as_bytes());
let tmp = tempdir().expect("tempdir");
let managed_path = tmp.path().join("managed_config.toml");
@@ -167,17 +152,7 @@ flag = true
let overrides = LoaderOverrides {
managed_config_path: Some(managed_path),
managed_preferences_base64: Some(
base64::prelude::BASE64_STANDARD.encode(
r#"
[nested]
value = "managed"
flag = false
"#
.as_bytes(),
),
),
macos_managed_config_requirements_base64: None,
managed_preferences_base64: Some(encoded),
};
let cwd = AbsolutePathBuf::try_from(tmp.path()).expect("cwd");
@@ -201,108 +176,6 @@ flag = false
assert_eq!(nested.get("flag"), Some(&TomlValue::Boolean(false)));
}
#[cfg(target_os = "macos")]
#[tokio::test]
async fn managed_preferences_requirements_are_applied() -> anyhow::Result<()> {
use base64::Engine;
let tmp = tempdir()?;
let state = load_config_layers_state(
tmp.path(),
Some(AbsolutePathBuf::try_from(tmp.path())?),
&[] as &[(String, TomlValue)],
LoaderOverrides {
managed_config_path: Some(tmp.path().join("managed_config.toml")),
managed_preferences_base64: Some(String::new()),
macos_managed_config_requirements_base64: Some(
base64::prelude::BASE64_STANDARD.encode(
r#"
allowed_approval_policies = ["never"]
allowed_sandbox_modes = ["read-only"]
"#
.as_bytes(),
),
),
},
)
.await?;
assert_eq!(
state.requirements().approval_policy.value(),
AskForApproval::Never
);
assert_eq!(
*state.requirements().sandbox_policy.get(),
SandboxPolicy::ReadOnly
);
assert!(
state
.requirements()
.approval_policy
.can_set(&AskForApproval::OnRequest)
.is_err()
);
assert!(
state
.requirements()
.sandbox_policy
.can_set(&SandboxPolicy::WorkspaceWrite {
writable_roots: Vec::new(),
network_access: false,
exclude_tmpdir_env_var: false,
exclude_slash_tmp: false,
})
.is_err()
);
Ok(())
}
#[cfg(target_os = "macos")]
#[tokio::test]
async fn managed_preferences_requirements_take_precedence() -> anyhow::Result<()> {
use base64::Engine;
let tmp = tempdir()?;
let managed_path = tmp.path().join("managed_config.toml");
tokio::fs::write(&managed_path, "approval_policy = \"on-request\"\n").await?;
let state = load_config_layers_state(
tmp.path(),
Some(AbsolutePathBuf::try_from(tmp.path())?),
&[] as &[(String, TomlValue)],
LoaderOverrides {
managed_config_path: Some(managed_path),
managed_preferences_base64: Some(String::new()),
macos_managed_config_requirements_base64: Some(
base64::prelude::BASE64_STANDARD.encode(
r#"
allowed_approval_policies = ["never"]
"#
.as_bytes(),
),
),
},
)
.await?;
assert_eq!(
state.requirements().approval_policy.value(),
AskForApproval::Never
);
assert!(
state
.requirements()
.approval_policy
.can_set(&AskForApproval::OnRequest)
.is_err()
);
Ok(())
}
#[tokio::test(flavor = "current_thread")]
async fn load_requirements_toml_produces_expected_constraints() -> anyhow::Result<()> {
let tmp = tempdir()?;

View File

@@ -5,9 +5,6 @@ use crate::truncate::approx_token_count;
use crate::truncate::approx_tokens_from_byte_count;
use crate::truncate::truncate_function_output_items_with_policy;
use crate::truncate::truncate_text;
use crate::user_instructions::SkillInstructions;
use crate::user_instructions::UserInstructions;
use crate::user_shell_command::is_user_shell_command_text;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::FunctionCallOutputPayload;
@@ -155,39 +152,6 @@ impl ContextManager {
}
}
/// Drop the last `num_turns` user turns from this history.
///
/// "User turns" are identified as `ResponseItem::Message` entries whose role is `"user"`.
///
/// This mirrors thread-rollback semantics:
/// - `num_turns == 0` is a no-op
/// - if there are no user turns, this is a no-op
/// - if `num_turns` exceeds the number of user turns, all user turns are dropped while
/// preserving any items that occurred before the first user message.
pub(crate) fn drop_last_n_user_turns(&mut self, num_turns: u32) {
if num_turns == 0 {
return;
}
// Keep behavior consistent with call sites that previously operated on `get_history()`:
// normalize first (call/output invariants), then truncate based on the normalized view.
let snapshot = self.get_history();
let user_positions = user_message_positions(&snapshot);
let Some(&first_user_idx) = user_positions.first() else {
self.replace(snapshot);
return;
};
let n_from_end = usize::try_from(num_turns).unwrap_or(usize::MAX);
let cut_idx = if n_from_end >= user_positions.len() {
first_user_idx
} else {
user_positions[user_positions.len() - n_from_end]
};
self.replace(snapshot[..cut_idx].to_vec());
}
pub(crate) fn update_token_info(
&mut self,
usage: &TokenUsage,
@@ -327,56 +291,6 @@ fn estimate_reasoning_length(encoded_len: usize) -> usize {
.saturating_sub(650)
}
fn is_session_prefix(text: &str) -> bool {
let trimmed = text.trim_start();
let lowered = trimmed.to_ascii_lowercase();
lowered.starts_with("<environment_context>")
}
fn is_user_turn_boundary(item: &ResponseItem) -> bool {
let ResponseItem::Message { role, content, .. } = item else {
return false;
};
if role != "user" {
return false;
}
if UserInstructions::is_user_instructions(content)
|| SkillInstructions::is_skill_instructions(content)
{
return false;
}
for content_item in content {
match content_item {
ContentItem::InputText { text } => {
if is_session_prefix(text) || is_user_shell_command_text(text) {
return false;
}
}
ContentItem::OutputText { text } => {
if is_session_prefix(text) {
return false;
}
}
ContentItem::InputImage { .. } => {}
}
}
true
}
fn user_message_positions(items: &[ResponseItem]) -> Vec<usize> {
let mut positions = Vec::new();
for (idx, item) in items.iter().enumerate() {
if is_user_turn_boundary(item) {
positions.push(idx);
}
}
positions
}
#[cfg(test)]
#[path = "history_tests.rs"]
mod tests;

View File

@@ -43,16 +43,6 @@ fn user_msg(text: &str) -> ResponseItem {
}
}
fn user_input_text_msg(text: &str) -> ResponseItem {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: text.to_string(),
}],
}
}
fn reasoning_msg(text: &str) -> ResponseItem {
ResponseItem::Reasoning {
id: String::new(),
@@ -237,127 +227,6 @@ fn remove_first_item_handles_local_shell_pair() {
assert_eq!(h.contents(), vec![]);
}
#[test]
fn drop_last_n_user_turns_preserves_prefix() {
let items = vec![
assistant_msg("session prefix item"),
user_msg("u1"),
assistant_msg("a1"),
user_msg("u2"),
assistant_msg("a2"),
];
let mut history = create_history_with_items(items);
history.drop_last_n_user_turns(1);
assert_eq!(
history.get_history(),
vec![
assistant_msg("session prefix item"),
user_msg("u1"),
assistant_msg("a1"),
]
);
let mut history = create_history_with_items(vec![
assistant_msg("session prefix item"),
user_msg("u1"),
assistant_msg("a1"),
user_msg("u2"),
assistant_msg("a2"),
]);
history.drop_last_n_user_turns(99);
assert_eq!(
history.get_history(),
vec![assistant_msg("session prefix item")]
);
}
#[test]
fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let items = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
user_input_text_msg(
"<skill>\n<name>demo</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>",
),
user_input_text_msg("<user_shell_command>echo 42</user_shell_command>"),
user_input_text_msg("turn 1 user"),
assistant_msg("turn 1 assistant"),
user_input_text_msg("turn 2 user"),
assistant_msg("turn 2 assistant"),
];
let mut history = create_history_with_items(items);
history.drop_last_n_user_turns(1);
let expected_prefix_and_first_turn = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
user_input_text_msg(
"<skill>\n<name>demo</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>",
),
user_input_text_msg("<user_shell_command>echo 42</user_shell_command>"),
user_input_text_msg("turn 1 user"),
assistant_msg("turn 1 assistant"),
];
assert_eq!(history.get_history(), expected_prefix_and_first_turn);
let expected_prefix_only = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
user_input_text_msg(
"<skill>\n<name>demo</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>",
),
user_input_text_msg("<user_shell_command>echo 42</user_shell_command>"),
];
let mut history = create_history_with_items(vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
user_input_text_msg(
"<skill>\n<name>demo</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>",
),
user_input_text_msg("<user_shell_command>echo 42</user_shell_command>"),
user_input_text_msg("turn 1 user"),
assistant_msg("turn 1 assistant"),
user_input_text_msg("turn 2 user"),
assistant_msg("turn 2 assistant"),
]);
history.drop_last_n_user_turns(2);
assert_eq!(history.get_history(), expected_prefix_only);
let mut history = create_history_with_items(vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
user_input_text_msg(
"<skill>\n<name>demo</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>",
),
user_input_text_msg("<user_shell_command>echo 42</user_shell_command>"),
user_input_text_msg("turn 1 user"),
assistant_msg("turn 1 assistant"),
user_input_text_msg("turn 2 user"),
assistant_msg("turn 2 assistant"),
]);
history.drop_last_n_user_turns(3);
assert_eq!(history.get_history(), expected_prefix_only);
}
#[test]
fn remove_first_item_handles_custom_tool_pair() {
let items = vec![

View File

@@ -3,7 +3,6 @@ use crate::AuthManager;
use crate::CodexAuth;
#[cfg(any(test, feature = "test-support"))]
use crate::ModelProviderInfo;
use crate::agent::AgentControl;
use crate::codex::Codex;
use crate::codex::CodexSpawnOk;
use crate::codex::INITIAL_SUBMIT_ID;
@@ -16,12 +15,12 @@ use crate::protocol::Event;
use crate::protocol::EventMsg;
use crate::protocol::SessionConfiguredEvent;
use crate::rollout::RolloutRecorder;
use crate::rollout::truncation;
use crate::skills::SkillsManager;
use codex_protocol::ConversationId;
use codex_protocol::items::TurnItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::openai_models::ModelPreset;
use codex_protocol::protocol::InitialHistory;
use codex_protocol::protocol::Op;
use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::SessionSource;
use std::collections::HashMap;
@@ -42,34 +41,24 @@ pub struct NewConversation {
/// [`ConversationManager`] is responsible for creating conversations and
/// maintaining them in memory.
pub struct ConversationManager {
state: Arc<ConversationManagerState>,
#[cfg(any(test, feature = "test-support"))]
_test_codex_home_guard: Option<TempDir>,
}
/// Shared, `Arc`-owned state for [`ConversationManager`]. This `Arc` is required to have a single
/// `Arc` reference that can be downgraded to by `AgentControl` while preventing every single
/// function to require an `Arc<&Self>`.
pub(crate) struct ConversationManagerState {
conversations: Arc<RwLock<HashMap<ConversationId, Arc<CodexConversation>>>>,
auth_manager: Arc<AuthManager>,
models_manager: Arc<ModelsManager>,
skills_manager: Arc<SkillsManager>,
session_source: SessionSource,
#[cfg(any(test, feature = "test-support"))]
_test_codex_home_guard: Option<TempDir>,
}
impl ConversationManager {
pub fn new(auth_manager: Arc<AuthManager>, session_source: SessionSource) -> Self {
let skills_manager = Arc::new(SkillsManager::new(auth_manager.codex_home().to_path_buf()));
Self {
state: Arc::new(ConversationManagerState {
conversations: Arc::new(RwLock::new(HashMap::new())),
models_manager: Arc::new(ModelsManager::new(auth_manager.clone())),
skills_manager: Arc::new(SkillsManager::new(
auth_manager.codex_home().to_path_buf(),
)),
auth_manager,
session_source,
}),
conversations: Arc::new(RwLock::new(HashMap::new())),
auth_manager: auth_manager.clone(),
session_source,
models_manager: Arc::new(ModelsManager::new(auth_manager)),
skills_manager,
#[cfg(any(test, feature = "test-support"))]
_test_codex_home_guard: None,
}
@@ -94,165 +83,40 @@ impl ConversationManager {
provider: ModelProviderInfo,
codex_home: PathBuf,
) -> Self {
let auth_manager = AuthManager::from_auth_for_testing_with_home(auth, codex_home);
let auth_manager = crate::AuthManager::from_auth_for_testing_with_home(auth, codex_home);
let skills_manager = Arc::new(SkillsManager::new(auth_manager.codex_home().to_path_buf()));
Self {
state: Arc::new(ConversationManagerState {
conversations: Arc::new(RwLock::new(HashMap::new())),
models_manager: Arc::new(ModelsManager::with_provider(
auth_manager.clone(),
provider,
)),
skills_manager: Arc::new(SkillsManager::new(
auth_manager.codex_home().to_path_buf(),
)),
auth_manager,
session_source: SessionSource::Exec,
}),
conversations: Arc::new(RwLock::new(HashMap::new())),
auth_manager: auth_manager.clone(),
session_source: SessionSource::Exec,
models_manager: Arc::new(ModelsManager::with_provider(auth_manager, provider)),
skills_manager,
_test_codex_home_guard: None,
}
}
pub fn session_source(&self) -> SessionSource {
self.state.session_source.clone()
self.session_source.clone()
}
pub fn skills_manager(&self) -> Arc<SkillsManager> {
self.state.skills_manager.clone()
}
pub fn get_models_manager(&self) -> Arc<ModelsManager> {
self.state.models_manager.clone()
}
pub async fn list_models(&self, config: &Config) -> Vec<ModelPreset> {
self.state.models_manager.list_models(config).await
}
pub async fn get_conversation(
&self,
conversation_id: ConversationId,
) -> CodexResult<Arc<CodexConversation>> {
self.state.get_conversation(conversation_id).await
self.skills_manager.clone()
}
pub async fn new_conversation(&self, config: Config) -> CodexResult<NewConversation> {
self.state
.spawn_conversation(
config,
InitialHistory::New,
Arc::clone(&self.state.auth_manager),
self.agent_control(),
)
.await
}
pub async fn resume_conversation_from_rollout(
&self,
config: Config,
rollout_path: PathBuf,
auth_manager: Arc<AuthManager>,
) -> CodexResult<NewConversation> {
let initial_history = RolloutRecorder::get_rollout_history(&rollout_path).await?;
self.resume_conversation_with_history(config, initial_history, auth_manager)
.await
}
pub async fn resume_conversation_with_history(
&self,
config: Config,
initial_history: InitialHistory,
auth_manager: Arc<AuthManager>,
) -> CodexResult<NewConversation> {
self.state
.spawn_conversation(config, initial_history, auth_manager, self.agent_control())
.await
}
/// Removes the conversation from the manager's internal map, though the conversation is stored
/// as `Arc<CodexConversation>`, it is possible that other references to it exist elsewhere.
/// Returns the conversation if the conversation was found and removed.
pub async fn remove_conversation(
&self,
conversation_id: &ConversationId,
) -> Option<Arc<CodexConversation>> {
self.state
.conversations
.write()
.await
.remove(conversation_id)
}
/// Fork an existing conversation by taking messages up to the given position (not including
/// the message at the given position) and starting a new conversation with identical
/// configuration (unless overridden by the caller's `config`). The new conversation will have
/// a fresh id.
pub async fn fork_conversation(
&self,
nth_user_message: usize,
config: Config,
path: PathBuf,
) -> CodexResult<NewConversation> {
let history = RolloutRecorder::get_rollout_history(&path).await?;
let history = truncate_before_nth_user_message(history, nth_user_message);
self.state
.spawn_conversation(
config,
history,
Arc::clone(&self.state.auth_manager),
self.agent_control(),
)
.await
}
fn agent_control(&self) -> AgentControl {
AgentControl::new(Arc::downgrade(&self.state))
}
}
impl ConversationManagerState {
pub(crate) async fn get_conversation(
&self,
conversation_id: ConversationId,
) -> CodexResult<Arc<CodexConversation>> {
let conversations = self.conversations.read().await;
conversations
.get(&conversation_id)
.cloned()
.ok_or_else(|| CodexErr::ConversationNotFound(conversation_id))
}
pub(crate) async fn send_op(
&self,
conversation_id: ConversationId,
op: Op,
) -> CodexResult<String> {
self.get_conversation(conversation_id)
.await?
.submit(op)
.await
}
#[allow(dead_code)] // Used by upcoming multi-agent tooling.
pub(crate) async fn spawn_new_conversation(
&self,
config: Config,
agent_control: AgentControl,
) -> CodexResult<NewConversation> {
self.spawn_conversation(
config,
InitialHistory::New,
Arc::clone(&self.auth_manager),
agent_control,
self.auth_manager.clone(),
self.models_manager.clone(),
)
.await
}
pub(crate) async fn spawn_conversation(
async fn spawn_conversation(
&self,
config: Config,
initial_history: InitialHistory,
auth_manager: Arc<AuthManager>,
agent_control: AgentControl,
models_manager: Arc<ModelsManager>,
) -> CodexResult<NewConversation> {
let CodexSpawnOk {
codex,
@@ -260,11 +124,10 @@ impl ConversationManagerState {
} = Codex::spawn(
config,
auth_manager,
Arc::clone(&self.models_manager),
Arc::clone(&self.skills_manager),
initial_history,
models_manager,
self.skills_manager.clone(),
InitialHistory::New,
self.session_source.clone(),
agent_control,
)
.await?;
self.finalize_spawn(codex, conversation_id).await
@@ -275,6 +138,9 @@ impl ConversationManagerState {
codex: Codex,
conversation_id: ConversationId,
) -> CodexResult<NewConversation> {
// The first event must be `SessionInitialized`. Validate and forward it
// to the caller so that they can display it in the conversation
// history.
let event = codex.next_event().await?;
let session_configured = match event {
Event {
@@ -301,13 +167,129 @@ impl ConversationManagerState {
session_configured,
})
}
pub async fn get_conversation(
&self,
conversation_id: ConversationId,
) -> CodexResult<Arc<CodexConversation>> {
let conversations = self.conversations.read().await;
conversations
.get(&conversation_id)
.cloned()
.ok_or_else(|| CodexErr::ConversationNotFound(conversation_id))
}
pub async fn resume_conversation_from_rollout(
&self,
config: Config,
rollout_path: PathBuf,
auth_manager: Arc<AuthManager>,
) -> CodexResult<NewConversation> {
let initial_history = RolloutRecorder::get_rollout_history(&rollout_path).await?;
self.resume_conversation_with_history(config, initial_history, auth_manager)
.await
}
pub async fn resume_conversation_with_history(
&self,
config: Config,
initial_history: InitialHistory,
auth_manager: Arc<AuthManager>,
) -> CodexResult<NewConversation> {
let CodexSpawnOk {
codex,
conversation_id,
} = Codex::spawn(
config,
auth_manager,
self.models_manager.clone(),
self.skills_manager.clone(),
initial_history,
self.session_source.clone(),
)
.await?;
self.finalize_spawn(codex, conversation_id).await
}
/// Removes the conversation from the manager's internal map, though the
/// conversation is stored as `Arc<CodexConversation>`, it is possible that
/// other references to it exist elsewhere. Returns the conversation if the
/// conversation was found and removed.
pub async fn remove_conversation(
&self,
conversation_id: &ConversationId,
) -> Option<Arc<CodexConversation>> {
self.conversations.write().await.remove(conversation_id)
}
/// Fork an existing conversation by taking messages up to the given position
/// (not including the message at the given position) and starting a new
/// conversation with identical configuration (unless overridden by the
/// caller's `config`). The new conversation will have a fresh id.
pub async fn fork_conversation(
&self,
nth_user_message: usize,
config: Config,
path: PathBuf,
) -> CodexResult<NewConversation> {
// Compute the prefix up to the cut point.
let history = RolloutRecorder::get_rollout_history(&path).await?;
let history = truncate_before_nth_user_message(history, nth_user_message);
// Spawn a new conversation with the computed initial history.
let auth_manager = self.auth_manager.clone();
let CodexSpawnOk {
codex,
conversation_id,
} = Codex::spawn(
config,
auth_manager,
self.models_manager.clone(),
self.skills_manager.clone(),
history,
self.session_source.clone(),
)
.await?;
self.finalize_spawn(codex, conversation_id).await
}
pub async fn list_models(&self, config: &Config) -> Vec<ModelPreset> {
self.models_manager.list_models(config).await
}
pub fn get_models_manager(&self) -> Arc<ModelsManager> {
self.models_manager.clone()
}
}
/// Return a prefix of `items` obtained by cutting strictly before the nth user message
/// (0-based) and all items that follow it.
fn truncate_before_nth_user_message(history: InitialHistory, n: usize) -> InitialHistory {
// Work directly on rollout items, and cut the vector at the nth user message input.
let items: Vec<RolloutItem> = history.get_rollout_items();
let rolled = truncation::truncate_rollout_before_nth_user_message_from_start(&items, n);
// Find indices of user message inputs in rollout order.
let mut user_positions: Vec<usize> = Vec::new();
for (idx, item) in items.iter().enumerate() {
if let RolloutItem::ResponseItem(item @ ResponseItem::Message { .. }) = item
&& matches!(
crate::event_mapping::parse_turn_item(item),
Some(TurnItem::UserMessage(_))
)
{
user_positions.push(idx);
}
}
// If fewer than or equal to n user messages exist, treat as empty (out of range).
if user_positions.len() <= n {
return InitialHistory::New;
}
// Cut strictly before the nth user message (do not keep the nth itself).
let cut_idx = user_positions[n];
let rolled: Vec<RolloutItem> = items.into_iter().take(cut_idx).collect();
if rolled.is_empty() {
InitialHistory::New
@@ -363,13 +345,14 @@ mod tests {
},
ResponseItem::FunctionCall {
id: None,
call_id: "c1".to_string(),
name: "tool".to_string(),
arguments: "{}".to_string(),
call_id: "c1".to_string(),
},
assistant_msg("a4"),
];
// Wrap as InitialHistory::Forked with response items only.
let initial: Vec<RolloutItem> = items
.iter()
.cloned()

View File

@@ -6,8 +6,6 @@ use std::sync::Arc;
use arc_swap::ArcSwap;
use crate::command_safety::is_dangerous_command::requires_initial_appoval;
use crate::config_loader::ConfigLayerStack;
use crate::config_loader::ConfigLayerStackOrdering;
use codex_execpolicy::AmendError;
use codex_execpolicy::Decision;
use codex_execpolicy::Error as ExecPolicyRuleError;
@@ -28,10 +26,11 @@ use crate::features::Feature;
use crate::features::Features;
use crate::sandboxing::SandboxPermissions;
use crate::tools::sandboxing::ExecApprovalRequirement;
use shlex::try_join as shlex_try_join;
const FORBIDDEN_REASON: &str = "execpolicy forbids this command";
const PROMPT_CONFLICT_REASON: &str =
"approval required by policy, but AskForApproval is set to Never";
"execpolicy requires approval for this command, but AskForApproval is set to Never";
const PROMPT_REASON: &str = "execpolicy requires approval for this command";
const RULES_DIR_NAME: &str = "rules";
const RULE_EXTENSION: &str = "rules";
const DEFAULT_POLICY_FILE: &str = "default.rules";
@@ -95,9 +94,9 @@ impl ExecPolicyManager {
pub(crate) async fn load(
features: &Features,
config_stack: &ConfigLayerStack,
codex_home: &Path,
) -> Result<Self, ExecPolicyError> {
let policy = load_exec_policy_for_features(features, config_stack).await?;
let policy = load_exec_policy_for_features(features, codex_home).await?;
Ok(Self::new(Arc::new(policy)))
}
@@ -127,7 +126,7 @@ impl ExecPolicyManager {
match evaluation.decision {
Decision::Forbidden => ExecApprovalRequirement::Forbidden {
reason: derive_forbidden_reason(command, &evaluation),
reason: FORBIDDEN_REASON.to_string(),
},
Decision::Prompt => {
if matches!(approval_policy, AskForApproval::Never) {
@@ -136,7 +135,7 @@ impl ExecPolicyManager {
}
} else {
ExecApprovalRequirement::NeedsApproval {
reason: derive_prompt_reason(command, &evaluation),
reason: derive_prompt_reason(&evaluation),
proposed_execpolicy_amendment: if features.enabled(Feature::ExecPolicy) {
try_derive_execpolicy_amendment_for_prompt_rules(
&evaluation.matched_rules,
@@ -195,28 +194,18 @@ impl Default for ExecPolicyManager {
async fn load_exec_policy_for_features(
features: &Features,
config_stack: &ConfigLayerStack,
codex_home: &Path,
) -> Result<Policy, ExecPolicyError> {
if !features.enabled(Feature::ExecPolicy) {
Ok(Policy::empty())
} else {
load_exec_policy(config_stack).await
load_exec_policy(codex_home).await
}
}
pub async fn load_exec_policy(config_stack: &ConfigLayerStack) -> Result<Policy, ExecPolicyError> {
// Iterate the layers in increasing order of precedence, adding the *.rules
// from each layer, so that higher-precedence layers can override
// rules defined in lower-precedence ones.
let mut policy_paths = Vec::new();
for layer in config_stack.get_layers(ConfigLayerStackOrdering::LowestPrecedenceFirst) {
if let Some(config_folder) = layer.config_folder() {
#[expect(clippy::expect_used)]
let policy_dir = config_folder.join(RULES_DIR_NAME).expect("safe join");
let layer_policy_paths = collect_policy_files(&policy_dir).await?;
policy_paths.extend(layer_policy_paths);
}
}
pub async fn load_exec_policy(codex_home: &Path) -> Result<Policy, ExecPolicyError> {
let policy_dir = codex_home.join(RULES_DIR_NAME);
let policy_paths = collect_policy_files(&policy_dir).await?;
let mut parser = PolicyParser::new();
for policy_path in &policy_paths {
@@ -237,7 +226,11 @@ pub async fn load_exec_policy(config_stack: &ConfigLayerStack) -> Result<Policy,
}
let policy = parser.build();
tracing::debug!("loaded execpolicy from {} files", policy_paths.len());
tracing::debug!(
"loaded execpolicy from {} files in {}",
policy_paths.len(),
policy_dir.display()
);
Ok(policy)
}
@@ -298,73 +291,18 @@ fn try_derive_execpolicy_amendment_for_allow_rules(
})
}
/// Only return a reason when a policy rule drove the prompt decision.
fn derive_prompt_reason(command_args: &[String], evaluation: &Evaluation) -> Option<String> {
let command = render_shlex_command(command_args);
let most_specific_prompt = evaluation
.matched_rules
.iter()
.filter_map(|rule_match| match rule_match {
RuleMatch::PrefixRuleMatch {
matched_prefix,
decision: Decision::Prompt,
justification,
..
} => Some((matched_prefix.len(), justification.as_deref())),
_ => None,
})
.max_by_key(|(matched_prefix_len, _)| *matched_prefix_len);
match most_specific_prompt {
Some((_matched_prefix_len, Some(justification))) => {
Some(format!("`{command}` requires approval: {justification}"))
/// Only return PROMPT_REASON when an execpolicy rule drove the prompt decision.
fn derive_prompt_reason(evaluation: &Evaluation) -> Option<String> {
evaluation.matched_rules.iter().find_map(|rule_match| {
if is_policy_match(rule_match) && rule_match.decision() == Decision::Prompt {
Some(PROMPT_REASON.to_string())
} else {
None
}
Some((_matched_prefix_len, None)) => {
Some(format!("`{command}` requires approval by policy"))
}
None => None,
}
})
}
fn render_shlex_command(args: &[String]) -> String {
shlex_try_join(args.iter().map(String::as_str)).unwrap_or_else(|_| args.join(" "))
}
/// Derive a string explaining why the command was forbidden. If `justification`
/// is set by the user, this can contain instructions with recommended
/// alternatives, for example.
fn derive_forbidden_reason(command_args: &[String], evaluation: &Evaluation) -> String {
let command = render_shlex_command(command_args);
let most_specific_forbidden = evaluation
.matched_rules
.iter()
.filter_map(|rule_match| match rule_match {
RuleMatch::PrefixRuleMatch {
matched_prefix,
decision: Decision::Forbidden,
justification,
..
} => Some((matched_prefix, justification.as_deref())),
_ => None,
})
.max_by_key(|(matched_prefix, _)| matched_prefix.len());
match most_specific_forbidden {
Some((_matched_prefix, Some(justification))) => {
format!("`{command}` rejected: {justification}")
}
Some((matched_prefix, None)) => {
let prefix = render_shlex_command(matched_prefix);
format!("`{command}` rejected: policy forbids commands starting with `{prefix}`")
}
None => format!("`{command}` rejected: blocked by policy"),
}
}
async fn collect_policy_files(dir: impl AsRef<Path>) -> Result<Vec<PathBuf>, ExecPolicyError> {
let dir = dir.as_ref();
async fn collect_policy_files(dir: &Path) -> Result<Vec<PathBuf>, ExecPolicyError> {
let mut read_dir = match fs::read_dir(dir).await {
Ok(read_dir) => read_dir,
Err(err) if err.kind() == ErrorKind::NotFound => return Ok(Vec::new()),
@@ -407,51 +345,28 @@ async fn collect_policy_files(dir: impl AsRef<Path>) -> Result<Vec<PathBuf>, Exe
policy_paths.sort();
tracing::debug!(
"loaded {} .rules files in {}",
policy_paths.len(),
dir.display()
);
Ok(policy_paths)
}
#[cfg(test)]
mod tests {
use super::*;
use crate::config_loader::ConfigLayerEntry;
use crate::config_loader::ConfigLayerStack;
use crate::config_loader::ConfigRequirements;
use crate::features::Feature;
use crate::features::Features;
use codex_app_server_protocol::ConfigLayerSource;
use codex_protocol::protocol::AskForApproval;
use codex_protocol::protocol::SandboxPolicy;
use codex_utils_absolute_path::AbsolutePathBuf;
use pretty_assertions::assert_eq;
use std::fs;
use std::path::Path;
use std::sync::Arc;
use tempfile::tempdir;
use toml::Value as TomlValue;
fn config_stack_for_dot_codex_folder(dot_codex_folder: &Path) -> ConfigLayerStack {
let dot_codex_folder = AbsolutePathBuf::from_absolute_path(dot_codex_folder)
.expect("absolute dot_codex_folder");
let layer = ConfigLayerEntry::new(
ConfigLayerSource::Project { dot_codex_folder },
TomlValue::Table(Default::default()),
);
ConfigLayerStack::new(vec![layer], ConfigRequirements::default()).expect("ConfigLayerStack")
}
#[tokio::test]
async fn returns_empty_policy_when_feature_disabled() {
let mut features = Features::with_defaults();
features.disable(Feature::ExecPolicy);
let temp_dir = tempdir().expect("create temp dir");
let config_stack = config_stack_for_dot_codex_folder(temp_dir.path());
let manager = ExecPolicyManager::load(&features, &config_stack)
let manager = ExecPolicyManager::load(&features, temp_dir.path())
.await
.expect("manager result");
let policy = manager.current();
@@ -485,7 +400,6 @@ mod tests {
#[tokio::test]
async fn loads_policies_from_policy_subdirectory() {
let temp_dir = tempdir().expect("create temp dir");
let config_stack = config_stack_for_dot_codex_folder(temp_dir.path());
let policy_dir = temp_dir.path().join(RULES_DIR_NAME);
fs::create_dir_all(&policy_dir).expect("create policy dir");
fs::write(
@@ -494,7 +408,7 @@ mod tests {
)
.expect("write policy file");
let policy = load_exec_policy(&config_stack)
let policy = load_exec_policy(temp_dir.path())
.await
.expect("policy result");
let command = [vec!["rm".to_string()]];
@@ -503,8 +417,7 @@ mod tests {
decision: Decision::Forbidden,
matched_rules: vec![RuleMatch::PrefixRuleMatch {
matched_prefix: vec!["rm".to_string()],
decision: Decision::Forbidden,
justification: None,
decision: Decision::Forbidden
}],
},
policy.check_multiple(command.iter(), &|_| Decision::Allow)
@@ -514,14 +427,13 @@ mod tests {
#[tokio::test]
async fn ignores_policies_outside_policy_dir() {
let temp_dir = tempdir().expect("create temp dir");
let config_stack = config_stack_for_dot_codex_folder(temp_dir.path());
fs::write(
temp_dir.path().join("root.rules"),
r#"prefix_rule(pattern=["ls"], decision="prompt")"#,
)
.expect("write policy file");
let policy = load_exec_policy(&config_stack)
let policy = load_exec_policy(temp_dir.path())
.await
.expect("policy result");
let command = [vec!["ls".to_string()]];
@@ -537,71 +449,6 @@ mod tests {
);
}
#[tokio::test]
async fn loads_policies_from_multiple_config_layers() -> anyhow::Result<()> {
let user_dir = tempdir()?;
let project_dir = tempdir()?;
let user_policy_dir = user_dir.path().join(RULES_DIR_NAME);
fs::create_dir_all(&user_policy_dir)?;
fs::write(
user_policy_dir.join("user.rules"),
r#"prefix_rule(pattern=["rm"], decision="forbidden")"#,
)?;
let project_policy_dir = project_dir.path().join(RULES_DIR_NAME);
fs::create_dir_all(&project_policy_dir)?;
fs::write(
project_policy_dir.join("project.rules"),
r#"prefix_rule(pattern=["ls"], decision="prompt")"#,
)?;
let user_config_toml =
AbsolutePathBuf::from_absolute_path(user_dir.path().join("config.toml"))?;
let project_dot_codex_folder = AbsolutePathBuf::from_absolute_path(project_dir.path())?;
let layers = vec![
ConfigLayerEntry::new(
ConfigLayerSource::User {
file: user_config_toml,
},
TomlValue::Table(Default::default()),
),
ConfigLayerEntry::new(
ConfigLayerSource::Project {
dot_codex_folder: project_dot_codex_folder,
},
TomlValue::Table(Default::default()),
),
];
let config_stack = ConfigLayerStack::new(layers, ConfigRequirements::default())?;
let policy = load_exec_policy(&config_stack).await?;
assert_eq!(
Evaluation {
decision: Decision::Forbidden,
matched_rules: vec![RuleMatch::PrefixRuleMatch {
matched_prefix: vec!["rm".to_string()],
decision: Decision::Forbidden,
justification: None,
}],
},
policy.check_multiple([vec!["rm".to_string()]].iter(), &|_| Decision::Allow)
);
assert_eq!(
Evaluation {
decision: Decision::Prompt,
matched_rules: vec![RuleMatch::PrefixRuleMatch {
matched_prefix: vec!["ls".to_string()],
decision: Decision::Prompt,
justification: None,
}],
},
policy.check_multiple([vec!["ls".to_string()]].iter(), &|_| Decision::Allow)
);
Ok(())
}
#[tokio::test]
async fn evaluates_bash_lc_inner_commands() {
let policy_src = r#"
@@ -616,7 +463,7 @@ prefix_rule(pattern=["rm"], decision="forbidden")
let forbidden_script = vec![
"bash".to_string(),
"-lc".to_string(),
"rm -rf /some/important/folder".to_string(),
"rm -rf /tmp".to_string(),
];
let manager = ExecPolicyManager::new(policy);
@@ -633,45 +480,7 @@ prefix_rule(pattern=["rm"], decision="forbidden")
assert_eq!(
requirement,
ExecApprovalRequirement::Forbidden {
reason: "`bash -lc 'rm -rf /some/important/folder'` rejected: policy forbids commands starting with `rm`".to_string()
}
);
}
#[tokio::test]
async fn justification_is_included_in_forbidden_exec_approval_requirement() {
let policy_src = r#"
prefix_rule(
pattern=["rm"],
decision="forbidden",
justification="destructive command",
)
"#;
let mut parser = PolicyParser::new();
parser
.parse("test.rules", policy_src)
.expect("parse policy");
let policy = Arc::new(parser.build());
let manager = ExecPolicyManager::new(policy);
let requirement = manager
.create_exec_approval_requirement_for_command(
&Features::with_defaults(),
&[
"rm".to_string(),
"-rf".to_string(),
"/some/important/folder".to_string(),
],
AskForApproval::OnRequest,
&SandboxPolicy::DangerFullAccess,
SandboxPermissions::UseDefault,
)
.await;
assert_eq!(
requirement,
ExecApprovalRequirement::Forbidden {
reason: "`rm -rf /some/important/folder` rejected: destructive command".to_string()
reason: FORBIDDEN_REASON.to_string()
}
);
}
@@ -700,7 +509,7 @@ prefix_rule(
assert_eq!(
requirement,
ExecApprovalRequirement::NeedsApproval {
reason: Some("`rm` requires approval by policy".to_string()),
reason: Some(PROMPT_REASON.to_string()),
proposed_execpolicy_amendment: None,
}
);
@@ -918,7 +727,7 @@ prefix_rule(
assert_eq!(
requirement,
ExecApprovalRequirement::NeedsApproval {
reason: Some("`rm` requires approval by policy".to_string()),
reason: Some(PROMPT_REASON.to_string()),
proposed_execpolicy_amendment: None,
}
);

View File

@@ -72,11 +72,8 @@ pub enum Feature {
UnifiedExec,
/// Include the freeform apply_patch tool.
ApplyPatchFreeform,
/// Allow the model to request web searches that fetch live content.
/// Allow the model to request web searches.
WebSearchRequest,
/// Allow the model to request web searches that fetch cached content.
/// Takes precedence over `WebSearchRequest`.
WebSearchCached,
/// Gate the execpolicy enforcement for shell/unified exec.
ExecPolicy,
/// Enable Windows sandbox (restricted token) on Windows.
@@ -258,10 +255,6 @@ impl Features {
features
}
pub fn enabled_features(&self) -> Vec<Feature> {
self.enabled.iter().copied().collect()
}
}
/// Keys accepted in `[features]` tables.
@@ -333,12 +326,6 @@ pub const FEATURES: &[FeatureSpec] = &[
stage: Stage::Stable,
default_enabled: false,
},
FeatureSpec {
id: Feature::WebSearchCached,
key: "web_search_cached",
stage: Stage::Experimental,
default_enabled: false,
},
// Beta program. Rendered in the `/experimental` menu for users.
FeatureSpec {
id: Feature::UnifiedExec,

View File

@@ -15,7 +15,6 @@ pub mod codex;
mod codex_conversation;
mod compact_remote;
pub use codex_conversation::CodexConversation;
mod agent;
mod codex_delegate;
mod command_safety;
pub mod config;

View File

@@ -79,60 +79,26 @@ pub const DEFAULT_STARTUP_TIMEOUT: Duration = Duration::from_secs(10);
/// Default timeout for individual tool calls.
const DEFAULT_TOOL_TIMEOUT: Duration = Duration::from_secs(60);
/// The Responses API requires tool names to match `^[a-zA-Z0-9_-]+$`.
/// MCP server/tool names are user-controlled, so sanitize the fully-qualified
/// name we expose to the model by replacing any disallowed character with `_`.
fn sanitize_responses_api_tool_name(name: &str) -> String {
let mut sanitized = String::with_capacity(name.len());
for c in name.chars() {
if c.is_ascii_alphanumeric() || c == '_' || c == '-' {
sanitized.push(c);
} else {
sanitized.push('_');
}
}
if sanitized.is_empty() {
"_".to_string()
} else {
sanitized
}
}
fn sha1_hex(s: &str) -> String {
let mut hasher = Sha1::new();
hasher.update(s.as_bytes());
let sha1 = hasher.finalize();
format!("{sha1:x}")
}
fn qualify_tools<I>(tools: I) -> HashMap<String, ToolInfo>
where
I: IntoIterator<Item = ToolInfo>,
{
let mut used_names = HashSet::new();
let mut seen_raw_names = HashSet::new();
let mut qualified_tools = HashMap::new();
for tool in tools {
let qualified_name_raw = format!(
let mut qualified_name = format!(
"mcp{}{}{}{}",
MCP_TOOL_NAME_DELIMITER, tool.server_name, MCP_TOOL_NAME_DELIMITER, tool.tool_name
);
if !seen_raw_names.insert(qualified_name_raw.clone()) {
warn!("skipping duplicated tool {}", qualified_name_raw);
continue;
}
// Start from a "pretty" name (sanitized), then deterministically disambiguate on
// collisions by appending a hash of the *raw* (unsanitized) qualified name. This
// ensures tools like `foo.bar` and `foo_bar` don't collapse to the same key.
let mut qualified_name = sanitize_responses_api_tool_name(&qualified_name_raw);
// Enforce length constraints early; use the raw name for the hash input so the
// output remains stable even when sanitization changes.
if qualified_name.len() > MAX_TOOL_NAME_LENGTH {
let sha1_str = sha1_hex(&qualified_name_raw);
let mut hasher = Sha1::new();
hasher.update(qualified_name.as_bytes());
let sha1 = hasher.finalize();
let sha1_str = format!("{sha1:x}");
// Truncate to make room for the hash suffix
let prefix_len = MAX_TOOL_NAME_LENGTH - sha1_str.len();
qualified_name = format!("{}{}", &qualified_name[..prefix_len], sha1_str);
}
@@ -1069,28 +1035,6 @@ mod tests {
);
}
#[test]
fn test_qualify_tools_sanitizes_invalid_characters() {
let tools = vec![create_test_tool("server.one", "tool.two")];
let qualified_tools = qualify_tools(tools);
assert_eq!(qualified_tools.len(), 1);
let (qualified_name, tool) = qualified_tools.into_iter().next().expect("one tool");
assert_eq!(qualified_name, "mcp__server_one__tool_two");
// The key is sanitized for OpenAI, but we keep original parts for the actual MCP call.
assert_eq!(tool.server_name, "server.one");
assert_eq!(tool.tool_name, "tool.two");
assert!(
qualified_name
.chars()
.all(|c| c.is_ascii_alphanumeric() || c == '_' || c == '-'),
"qualified name must be Responses API compatible: {qualified_name:?}"
);
}
#[test]
fn tool_filter_allows_by_default() {
let filter = ToolFilter::default();

View File

@@ -77,7 +77,7 @@ impl ModelsManager {
}
/// Fetch the latest remote models, using the on-disk cache when still fresh.
pub async fn refresh_available_models_with_cache(&self, config: &Config) -> CoreResult<()> {
pub async fn refresh_available_models(&self, config: &Config) -> CoreResult<()> {
if !config.features.enabled(Feature::RemoteModels)
|| self.auth_manager.get_auth_mode() == Some(AuthMode::ApiKey)
{
@@ -86,17 +86,7 @@ impl ModelsManager {
if self.try_load_cache().await {
return Ok(());
}
self.refresh_available_models_no_cache(config.features.enabled(Feature::RemoteModels))
.await
}
pub(crate) async fn refresh_available_models_no_cache(
&self,
remote_models_feature: bool,
) -> CoreResult<()> {
if !remote_models_feature || self.auth_manager.get_auth_mode() == Some(AuthMode::ApiKey) {
return Ok(());
}
let auth = self.auth_manager.auth();
let api_provider = self.provider.to_api_provider(Some(AuthMode::ChatGPT))?;
let api_auth = auth_provider_from_auth(auth.clone(), &self.provider).await?;
@@ -104,11 +94,13 @@ impl ModelsManager {
let client = ModelsClient::new(transport, api_provider, api_auth);
let client_version = format_client_version_to_whole();
let (models, etag) = client
let ModelsResponse { models, etag } = client
.list_models(&client_version, HeaderMap::new())
.await
.map_err(map_api_error)?;
let etag = (!etag.is_empty()).then_some(etag);
self.apply_remote_models(models.clone()).await;
*self.etag.write().await = etag.clone();
self.persist_cache(&models, etag).await;
@@ -116,7 +108,7 @@ impl ModelsManager {
}
pub async fn list_models(&self, config: &Config) -> Vec<ModelPreset> {
if let Err(err) = self.refresh_available_models_with_cache(config).await {
if let Err(err) = self.refresh_available_models(config).await {
error!("failed to refresh available models: {err}");
}
let remote_models = self.remote_models(config).await;
@@ -143,7 +135,7 @@ impl ModelsManager {
if let Some(model) = model.as_ref() {
return model.to_string();
}
if let Err(err) = self.refresh_available_models_with_cache(config).await {
if let Err(err) = self.refresh_available_models(config).await {
error!("failed to refresh available models: {err}");
}
// if codex-auto-balanced exists & signed in with chatgpt mode, return it, otherwise return the default model
@@ -161,18 +153,6 @@ impl ModelsManager {
}
OPENAI_DEFAULT_API_MODEL.to_string()
}
pub async fn refresh_if_new_etag(&self, etag: String, remote_models_feature: bool) {
let current_etag = self.get_etag().await;
if current_etag.clone().is_some() && current_etag.as_deref() == Some(etag.as_str()) {
return;
}
if let Err(err) = self
.refresh_available_models_no_cache(remote_models_feature)
.await
{
error!("failed to refresh available models: {err}");
}
}
#[cfg(any(test, feature = "test-support"))]
pub fn get_model_offline(model: Option<&str>) -> String {
@@ -185,10 +165,6 @@ impl ModelsManager {
Self::find_family_for_model(model).with_config_overrides(config)
}
async fn get_etag(&self) -> Option<String> {
self.etag.read().await.clone()
}
/// Replace the cached remote models and rebuild the derived presets list.
async fn apply_remote_models(&self, models: Vec<ModelInfo>) {
*self.remote_models.write().await = models;
@@ -312,14 +288,26 @@ impl ModelsManager {
/// Convert a client version string to a whole version string (e.g. "1.2.3-alpha.4" -> "1.2.3")
fn format_client_version_to_whole() -> String {
format!(
"{}.{}.{}",
format_client_version_from_parts(
env!("CARGO_PKG_VERSION_MAJOR"),
env!("CARGO_PKG_VERSION_MINOR"),
env!("CARGO_PKG_VERSION_PATCH")
env!("CARGO_PKG_VERSION_PATCH"),
)
}
fn format_client_version_from_parts(major: &str, minor: &str, patch: &str) -> String {
const DEV_VERSION: &str = "0.0.0";
const FALLBACK_VERSION: &str = "99.99.99";
let normalized = format!("{major}.{minor}.{patch}");
if normalized == DEV_VERSION {
FALLBACK_VERSION.to_string()
} else {
normalized
}
}
#[cfg(test)]
mod tests {
use super::cache::ModelsCache;
@@ -366,6 +354,7 @@ mod tests {
"truncation_policy": {"mode": "bytes", "limit": 10_000},
"supports_parallel_tool_calls": false,
"context_window": null,
"reasoning_summary_format": "none",
"experimental_supported_tools": [],
}))
.expect("valid model")
@@ -400,6 +389,7 @@ mod tests {
&server,
ModelsResponse {
models: remote_models.clone(),
etag: String::new(),
},
)
.await;
@@ -417,7 +407,7 @@ mod tests {
let manager = ModelsManager::with_provider(auth_manager, provider);
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("refresh succeeds");
let cached_remote = manager.remote_models(&config).await;
@@ -456,6 +446,7 @@ mod tests {
&server,
ModelsResponse {
models: remote_models.clone(),
etag: String::new(),
},
)
.await;
@@ -476,7 +467,7 @@ mod tests {
let manager = ModelsManager::with_provider(auth_manager, provider);
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("first refresh succeeds");
assert_eq!(
@@ -487,7 +478,7 @@ mod tests {
// Second call should read from cache and avoid the network.
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("cached refresh succeeds");
assert_eq!(
@@ -510,6 +501,7 @@ mod tests {
&server,
ModelsResponse {
models: initial_models.clone(),
etag: String::new(),
},
)
.await;
@@ -530,7 +522,7 @@ mod tests {
let manager = ModelsManager::with_provider(auth_manager, provider);
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("initial refresh succeeds");
@@ -550,12 +542,13 @@ mod tests {
&server,
ModelsResponse {
models: updated_models.clone(),
etag: String::new(),
},
)
.await;
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("second refresh succeeds");
assert_eq!(
@@ -583,6 +576,7 @@ mod tests {
&server,
ModelsResponse {
models: initial_models,
etag: String::new(),
},
)
.await;
@@ -601,7 +595,7 @@ mod tests {
manager.cache_ttl = Duration::ZERO;
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("initial refresh succeeds");
@@ -611,12 +605,13 @@ mod tests {
&server,
ModelsResponse {
models: refreshed_models,
etag: String::new(),
},
)
.await;
manager
.refresh_available_models_with_cache(&config)
.refresh_available_models(&config)
.await
.expect("second refresh succeeds");

View File

@@ -3,6 +3,7 @@ use codex_protocol::openai_models::ApplyPatchToolType;
use codex_protocol::openai_models::ConfigShellToolType;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::openai_models::ReasoningSummaryFormat;
use crate::config::Config;
use crate::truncate::TruncationPolicy;
@@ -47,6 +48,9 @@ pub struct ModelFamily {
// 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,
/// Whether this model supports parallel tool calls when using the
/// Responses API.
pub supports_parallel_tool_calls: bool,
@@ -84,6 +88,9 @@ impl ModelFamily {
if let Some(supports_reasoning_summaries) = config.model_supports_reasoning_summaries {
self.supports_reasoning_summaries = supports_reasoning_summaries;
}
if let Some(reasoning_summary_format) = config.model_reasoning_summary_format.as_ref() {
self.reasoning_summary_format = reasoning_summary_format.clone();
}
if let Some(context_window) = config.model_context_window {
self.context_window = Some(context_window);
}
@@ -121,6 +128,7 @@ impl ModelFamily {
truncation_policy,
supports_parallel_tool_calls,
context_window,
reasoning_summary_format,
experimental_supported_tools,
} = model;
@@ -136,6 +144,7 @@ impl ModelFamily {
self.truncation_policy = truncation_policy.into();
self.supports_parallel_tool_calls = supports_parallel_tool_calls;
self.context_window = context_window;
self.reasoning_summary_format = reasoning_summary_format;
self.experimental_supported_tools = experimental_supported_tools;
}
@@ -166,6 +175,7 @@ macro_rules! model_family {
context_window: Some(CONTEXT_WINDOW_272K),
auto_compact_token_limit: None,
supports_reasoning_summaries: false,
reasoning_summary_format: ReasoningSummaryFormat::None,
supports_parallel_tool_calls: false,
apply_patch_tool_type: None,
base_instructions: BASE_INSTRUCTIONS.to_string(),
@@ -240,6 +250,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_CODEX_INSTRUCTIONS.to_string(),
experimental_supported_tools: vec![
"grep_files".to_string(),
@@ -259,6 +270,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_2_CODEX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: ConfigShellToolType::ShellCommand,
@@ -287,6 +299,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_2_CODEX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: ConfigShellToolType::ShellCommand,
@@ -299,6 +312,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_2_CODEX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: ConfigShellToolType::ShellCommand,
@@ -311,6 +325,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_1_CODEX_MAX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: ConfigShellToolType::ShellCommand,
@@ -326,6 +341,7 @@ pub(super) fn find_family_for_model(slug: &str) -> ModelFamily {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_CODEX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: ConfigShellToolType::ShellCommand,
@@ -400,6 +416,7 @@ fn derive_default_model_family(model: &str) -> ModelFamily {
context_window: None,
auto_compact_token_limit: None,
supports_reasoning_summaries: false,
reasoning_summary_format: ReasoningSummaryFormat::None,
supports_parallel_tool_calls: false,
apply_patch_tool_type: None,
base_instructions: BASE_INSTRUCTIONS.to_string(),
@@ -443,6 +460,7 @@ mod tests {
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
context_window: None,
reasoning_summary_format: ReasoningSummaryFormat::None,
experimental_supported_tools: Vec::new(),
}
}
@@ -506,6 +524,7 @@ mod tests {
experimental_supported_tools: vec!["local".to_string()],
truncation_policy: TruncationPolicy::Bytes(10_000),
context_window: Some(100),
reasoning_summary_format: ReasoningSummaryFormat::None,
);
let updated = family.with_remote_overrides(vec![ModelInfo {
@@ -530,6 +549,7 @@ mod tests {
truncation_policy: TruncationPolicyConfig::tokens(2_000),
supports_parallel_tool_calls: true,
context_window: Some(400_000),
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
experimental_supported_tools: vec!["alpha".to_string(), "beta".to_string()],
}]);
@@ -548,6 +568,10 @@ mod tests {
assert_eq!(updated.truncation_policy, TruncationPolicy::Tokens(2_000));
assert!(updated.supports_parallel_tool_calls);
assert_eq!(updated.context_window, Some(400_000));
assert_eq!(
updated.reasoning_summary_format,
ReasoningSummaryFormat::Experimental
);
assert_eq!(
updated.experimental_supported_tools,
vec!["alpha".to_string(), "beta".to_string()]

View File

@@ -513,9 +513,9 @@ mod tests {
)
.unwrap_or_else(|_| cfg.codex_home.join("skills/pdf-processing/SKILL.md"));
let expected_path_str = expected_path.to_string_lossy().replace('\\', "/");
let usage_rules = "- Discovery: The list above is the skills available in this session (name + description + file path). Skill bodies live on disk at the listed paths.\n- Trigger rules: If the user names a skill (with `$SkillName` or plain text) OR the task clearly matches a skill's description shown above, you must use that skill for that turn. Multiple mentions mean use them all. Do not carry skills across turns unless re-mentioned.\n- Missing/blocked: If a named skill isn't in the list or the path can't be read, say so briefly and continue with the best fallback.\n- How to use a skill (progressive disclosure):\n 1) After deciding to use a skill, open its `SKILL.md`. Read only enough to follow the workflow.\n 2) If `SKILL.md` points to extra folders such as `references/`, load only the specific files needed for the request; don't bulk-load everything.\n 3) If `scripts/` exist, prefer running or patching them instead of retyping large code blocks.\n 4) If `assets/` or templates exist, reuse them instead of recreating from scratch.\n- Coordination and sequencing:\n - If multiple skills apply, choose the minimal set that covers the request and state the order you'll use them.\n - Announce which skill(s) you're using and why (one short line). If you skip an obvious skill, say why.\n- Context hygiene:\n - Keep context small: summarize long sections instead of pasting them; only load extra files when needed.\n - Avoid deep reference-chasing: prefer opening only files directly linked from `SKILL.md` unless you're blocked.\n - When variants exist (frameworks, providers, domains), pick only the relevant reference file(s) and note that choice.\n- Safety and fallback: If a skill can't be applied cleanly (missing files, unclear instructions), state the issue, pick the next-best approach, and continue.";
let usage_rules = "- Discovery: Available skills are listed in project docs and may also appear in a runtime \"## Skills\" section (name + description + file path). These are the sources of truth; skill bodies live on disk at the listed paths.\n- Trigger rules: If the user names a skill (with `$SkillName` or plain text) OR the task clearly matches a skill's description, you must use that skill for that turn. Multiple mentions mean use them all. Do not carry skills across turns unless re-mentioned.\n- Missing/blocked: If a named skill isn't in the list or the path can't be read, say so briefly and continue with the best fallback.\n- How to use a skill (progressive disclosure):\n 1) After deciding to use a skill, open its `SKILL.md`. Read only enough to follow the workflow.\n 2) If `SKILL.md` points to extra folders such as `references/`, load only the specific files needed for the request; don't bulk-load everything.\n 3) If `scripts/` exist, prefer running or patching them instead of retyping large code blocks.\n 4) If `assets/` or templates exist, reuse them instead of recreating from scratch.\n- Description as trigger: The YAML `description` in `SKILL.md` is the primary trigger signal; rely on it to decide applicability. If unsure, ask a brief clarification before proceeding.\n- Coordination and sequencing:\n - If multiple skills apply, choose the minimal set that covers the request and state the order you'll use them.\n - Announce which skill(s) you're using and why (one short line). If you skip an obvious skill, say why.\n- Context hygiene:\n - Keep context small: summarize long sections instead of pasting them; only load extra files when needed.\n - Avoid deeply nested references; prefer one-hop files explicitly linked from `SKILL.md`.\n - When variants exist (frameworks, providers, domains), pick only the relevant reference file(s) and note that choice.\n- Safety and fallback: If a skill can't be applied cleanly (missing files, unclear instructions), state the issue, pick the next-best approach, and continue.";
let expected = format!(
"base doc\n\n## Skills\nA skill is a set of local instructions to follow that is stored in a `SKILL.md` file. Below is the list of skills that can be used. Each entry includes a name, description, and file path so you can open the source for full instructions when using a specific skill.\n### Available skills\n- pdf-processing: extract from pdfs (file: {expected_path_str})\n### How to use skills\n{usage_rules}"
"base doc\n\n## Skills\nThese skills are discovered at startup from multiple local sources. Each entry includes a name, description, and file path so you can open the source for full instructions.\n- pdf-processing: extract from pdfs (file: {expected_path_str})\n{usage_rules}"
);
assert_eq!(res, expected);
}
@@ -537,9 +537,9 @@ mod tests {
dunce::canonicalize(cfg.codex_home.join("skills/linting/SKILL.md").as_path())
.unwrap_or_else(|_| cfg.codex_home.join("skills/linting/SKILL.md"));
let expected_path_str = expected_path.to_string_lossy().replace('\\', "/");
let usage_rules = "- Discovery: The list above is the skills available in this session (name + description + file path). Skill bodies live on disk at the listed paths.\n- Trigger rules: If the user names a skill (with `$SkillName` or plain text) OR the task clearly matches a skill's description shown above, you must use that skill for that turn. Multiple mentions mean use them all. Do not carry skills across turns unless re-mentioned.\n- Missing/blocked: If a named skill isn't in the list or the path can't be read, say so briefly and continue with the best fallback.\n- How to use a skill (progressive disclosure):\n 1) After deciding to use a skill, open its `SKILL.md`. Read only enough to follow the workflow.\n 2) If `SKILL.md` points to extra folders such as `references/`, load only the specific files needed for the request; don't bulk-load everything.\n 3) If `scripts/` exist, prefer running or patching them instead of retyping large code blocks.\n 4) If `assets/` or templates exist, reuse them instead of recreating from scratch.\n- Coordination and sequencing:\n - If multiple skills apply, choose the minimal set that covers the request and state the order you'll use them.\n - Announce which skill(s) you're using and why (one short line). If you skip an obvious skill, say why.\n- Context hygiene:\n - Keep context small: summarize long sections instead of pasting them; only load extra files when needed.\n - Avoid deep reference-chasing: prefer opening only files directly linked from `SKILL.md` unless you're blocked.\n - When variants exist (frameworks, providers, domains), pick only the relevant reference file(s) and note that choice.\n- Safety and fallback: If a skill can't be applied cleanly (missing files, unclear instructions), state the issue, pick the next-best approach, and continue.";
let usage_rules = "- Discovery: Available skills are listed in project docs and may also appear in a runtime \"## Skills\" section (name + description + file path). These are the sources of truth; skill bodies live on disk at the listed paths.\n- Trigger rules: If the user names a skill (with `$SkillName` or plain text) OR the task clearly matches a skill's description, you must use that skill for that turn. Multiple mentions mean use them all. Do not carry skills across turns unless re-mentioned.\n- Missing/blocked: If a named skill isn't in the list or the path can't be read, say so briefly and continue with the best fallback.\n- How to use a skill (progressive disclosure):\n 1) After deciding to use a skill, open its `SKILL.md`. Read only enough to follow the workflow.\n 2) If `SKILL.md` points to extra folders such as `references/`, load only the specific files needed for the request; don't bulk-load everything.\n 3) If `scripts/` exist, prefer running or patching them instead of retyping large code blocks.\n 4) If `assets/` or templates exist, reuse them instead of recreating from scratch.\n- Description as trigger: The YAML `description` in `SKILL.md` is the primary trigger signal; rely on it to decide applicability. If unsure, ask a brief clarification before proceeding.\n- Coordination and sequencing:\n - If multiple skills apply, choose the minimal set that covers the request and state the order you'll use them.\n - Announce which skill(s) you're using and why (one short line). If you skip an obvious skill, say why.\n- Context hygiene:\n - Keep context small: summarize long sections instead of pasting them; only load extra files when needed.\n - Avoid deeply nested references; prefer one-hop files explicitly linked from `SKILL.md`.\n - When variants exist (frameworks, providers, domains), pick only the relevant reference file(s) and note that choice.\n- Safety and fallback: If a skill can't be applied cleanly (missing files, unclear instructions), state the issue, pick the next-best approach, and continue.";
let expected = format!(
"## Skills\nA skill is a set of local instructions to follow that is stored in a `SKILL.md` file. Below is the list of skills that can be used. Each entry includes a name, description, and file path so you can open the source for full instructions when using a specific skill.\n### Available skills\n- linting: run clippy (file: {expected_path_str})\n### How to use skills\n{usage_rules}"
"## Skills\nThese skills are discovered at startup from multiple local sources. Each entry includes a name, description, and file path so you can open the source for full instructions.\n- linting: run clippy (file: {expected_path_str})\n{usage_rules}"
);
assert_eq!(res, expected);
}

View File

@@ -11,7 +11,6 @@ pub(crate) mod error;
pub mod list;
pub(crate) mod policy;
pub mod recorder;
pub(crate) mod truncation;
pub use codex_protocol::protocol::SessionMeta;
pub(crate) use error::map_session_init_error;

Some files were not shown because too many files have changed in this diff Show More