Compare commits

..

1 Commits

Author SHA1 Message Date
kevin zhao
9df6dbcb0f adding support for reading execpolicy from /etc/codex and .codex of root git repo 2025-12-08 10:51:50 -08:00
1122 changed files with 8246 additions and 90669 deletions

View File

@@ -1,2 +1 @@
iTerm
psuedo

View File

@@ -1,44 +0,0 @@
name: linux-code-sign
description: Sign Linux artifacts with cosign.
inputs:
target:
description: Target triple for the artifacts to sign.
required: true
artifacts-dir:
description: Absolute path to the directory containing built binaries to sign.
required: true
runs:
using: composite
steps:
- name: Install cosign
uses: sigstore/cosign-installer@v3.7.0
- name: Cosign Linux artifacts
shell: bash
env:
COSIGN_EXPERIMENTAL: "1"
COSIGN_YES: "true"
COSIGN_OIDC_CLIENT_ID: "sigstore"
COSIGN_OIDC_ISSUER: "https://oauth2.sigstore.dev/auth"
run: |
set -euo pipefail
dest="${{ inputs.artifacts-dir }}"
if [[ ! -d "$dest" ]]; then
echo "Destination $dest does not exist"
exit 1
fi
for binary in codex codex-responses-api-proxy; do
artifact="${dest}/${binary}"
if [[ ! -f "$artifact" ]]; then
echo "Binary $artifact not found"
exit 1
fi
cosign sign-blob \
--yes \
--bundle "${artifact}.sigstore" \
"$artifact"
done

View File

@@ -1,212 +0,0 @@
name: macos-code-sign
description: Configure, sign, notarize, and clean up macOS code signing artifacts.
inputs:
target:
description: Rust compilation target triple (e.g. aarch64-apple-darwin).
required: true
apple-certificate:
description: Base64-encoded Apple signing certificate (P12).
required: true
apple-certificate-password:
description: Password for the signing certificate.
required: true
apple-notarization-key-p8:
description: Base64-encoded Apple notarization key (P8).
required: true
apple-notarization-key-id:
description: Apple notarization key ID.
required: true
apple-notarization-issuer-id:
description: Apple notarization issuer ID.
required: true
runs:
using: composite
steps:
- name: Configure Apple code signing
shell: bash
env:
KEYCHAIN_PASSWORD: actions
APPLE_CERTIFICATE: ${{ inputs.apple-certificate }}
APPLE_CERTIFICATE_PASSWORD: ${{ inputs.apple-certificate-password }}
run: |
set -euo pipefail
if [[ -z "${APPLE_CERTIFICATE:-}" ]]; then
echo "APPLE_CERTIFICATE is required for macOS signing"
exit 1
fi
if [[ -z "${APPLE_CERTIFICATE_PASSWORD:-}" ]]; then
echo "APPLE_CERTIFICATE_PASSWORD is required for macOS signing"
exit 1
fi
cert_path="${RUNNER_TEMP}/apple_signing_certificate.p12"
echo "$APPLE_CERTIFICATE" | base64 -d > "$cert_path"
keychain_path="${RUNNER_TEMP}/codex-signing.keychain-db"
security create-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
security set-keychain-settings -lut 21600 "$keychain_path"
security unlock-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
keychain_args=()
cleanup_keychain() {
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "${keychain_args[@]}" || true
security default-keychain -s "${keychain_args[0]}" || true
else
security list-keychains -s || true
fi
if [[ -f "$keychain_path" ]]; then
security delete-keychain "$keychain_path" || true
fi
}
while IFS= read -r keychain; do
[[ -n "$keychain" ]] && keychain_args+=("$keychain")
done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "$keychain_path" "${keychain_args[@]}"
else
security list-keychains -s "$keychain_path"
fi
security default-keychain -s "$keychain_path"
security import "$cert_path" -k "$keychain_path" -P "$APPLE_CERTIFICATE_PASSWORD" -T /usr/bin/codesign -T /usr/bin/security
security set-key-partition-list -S apple-tool:,apple: -s -k "$KEYCHAIN_PASSWORD" "$keychain_path" > /dev/null
codesign_hashes=()
while IFS= read -r hash; do
[[ -n "$hash" ]] && codesign_hashes+=("$hash")
done < <(security find-identity -v -p codesigning "$keychain_path" \
| sed -n 's/.*\([0-9A-F]\{40\}\).*/\1/p' \
| sort -u)
if ((${#codesign_hashes[@]} == 0)); then
echo "No signing identities found in $keychain_path"
cleanup_keychain
rm -f "$cert_path"
exit 1
fi
if ((${#codesign_hashes[@]} > 1)); then
echo "Multiple signing identities found in $keychain_path:"
printf ' %s\n' "${codesign_hashes[@]}"
cleanup_keychain
rm -f "$cert_path"
exit 1
fi
APPLE_CODESIGN_IDENTITY="${codesign_hashes[0]}"
rm -f "$cert_path"
echo "APPLE_CODESIGN_IDENTITY=$APPLE_CODESIGN_IDENTITY" >> "$GITHUB_ENV"
echo "APPLE_CODESIGN_KEYCHAIN=$keychain_path" >> "$GITHUB_ENV"
echo "::add-mask::$APPLE_CODESIGN_IDENTITY"
- name: Sign macOS binaries
shell: bash
run: |
set -euo pipefail
if [[ -z "${APPLE_CODESIGN_IDENTITY:-}" ]]; then
echo "APPLE_CODESIGN_IDENTITY is required for macOS signing"
exit 1
fi
keychain_args=()
if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" && -f "${APPLE_CODESIGN_KEYCHAIN}" ]]; then
keychain_args+=(--keychain "${APPLE_CODESIGN_KEYCHAIN}")
fi
for binary in codex codex-responses-api-proxy; do
path="codex-rs/target/${{ inputs.target }}/release/${binary}"
codesign --force --options runtime --timestamp --sign "$APPLE_CODESIGN_IDENTITY" "${keychain_args[@]}" "$path"
done
- name: Notarize macOS binaries
shell: bash
env:
APPLE_NOTARIZATION_KEY_P8: ${{ inputs.apple-notarization-key-p8 }}
APPLE_NOTARIZATION_KEY_ID: ${{ inputs.apple-notarization-key-id }}
APPLE_NOTARIZATION_ISSUER_ID: ${{ inputs.apple-notarization-issuer-id }}
run: |
set -euo pipefail
for var in APPLE_NOTARIZATION_KEY_P8 APPLE_NOTARIZATION_KEY_ID APPLE_NOTARIZATION_ISSUER_ID; do
if [[ -z "${!var:-}" ]]; then
echo "$var is required for notarization"
exit 1
fi
done
notary_key_path="${RUNNER_TEMP}/notarytool.key.p8"
echo "$APPLE_NOTARIZATION_KEY_P8" | base64 -d > "$notary_key_path"
cleanup_notary() {
rm -f "$notary_key_path"
}
trap cleanup_notary EXIT
notarize_binary() {
local binary="$1"
local source_path="codex-rs/target/${{ inputs.target }}/release/${binary}"
local archive_path="${RUNNER_TEMP}/${binary}.zip"
if [[ ! -f "$source_path" ]]; then
echo "Binary $source_path not found"
exit 1
fi
rm -f "$archive_path"
ditto -c -k --keepParent "$source_path" "$archive_path"
submission_json=$(xcrun notarytool submit "$archive_path" \
--key "$notary_key_path" \
--key-id "$APPLE_NOTARIZATION_KEY_ID" \
--issuer "$APPLE_NOTARIZATION_ISSUER_ID" \
--output-format json \
--wait)
status=$(printf '%s\n' "$submission_json" | jq -r '.status // "Unknown"')
submission_id=$(printf '%s\n' "$submission_json" | jq -r '.id // ""')
if [[ -z "$submission_id" ]]; then
echo "Failed to retrieve submission ID for $binary"
exit 1
fi
echo "::notice title=Notarization::$binary submission ${submission_id} completed with status ${status}"
if [[ "$status" != "Accepted" ]]; then
echo "Notarization failed for ${binary} (submission ${submission_id}, status ${status})"
exit 1
fi
}
notarize_binary "codex"
notarize_binary "codex-responses-api-proxy"
- name: Remove signing keychain
if: ${{ always() }}
shell: bash
env:
APPLE_CODESIGN_KEYCHAIN: ${{ env.APPLE_CODESIGN_KEYCHAIN }}
run: |
set -euo pipefail
if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" ]]; then
keychain_args=()
while IFS= read -r keychain; do
[[ "$keychain" == "$APPLE_CODESIGN_KEYCHAIN" ]] && continue
[[ -n "$keychain" ]] && keychain_args+=("$keychain")
done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "${keychain_args[@]}"
security default-keychain -s "${keychain_args[0]}"
fi
if [[ -f "$APPLE_CODESIGN_KEYCHAIN" ]]; then
security delete-keychain "$APPLE_CODESIGN_KEYCHAIN"
fi
fi

View File

@@ -1,57 +0,0 @@
name: windows-code-sign
description: Sign Windows binaries with Azure Trusted Signing.
inputs:
target:
description: Target triple for the artifacts to sign.
required: true
client-id:
description: Azure Trusted Signing client ID.
required: true
tenant-id:
description: Azure tenant ID for Trusted Signing.
required: true
subscription-id:
description: Azure subscription ID for Trusted Signing.
required: true
endpoint:
description: Azure Trusted Signing endpoint.
required: true
account-name:
description: Azure Trusted Signing account name.
required: true
certificate-profile-name:
description: Certificate profile name for signing.
required: true
runs:
using: composite
steps:
- name: Azure login for Trusted Signing (OIDC)
uses: azure/login@v2
with:
client-id: ${{ inputs.client-id }}
tenant-id: ${{ inputs.tenant-id }}
subscription-id: ${{ inputs.subscription-id }}
- name: Sign Windows binaries with Azure Trusted Signing
uses: azure/trusted-signing-action@v0
with:
endpoint: ${{ inputs.endpoint }}
trusted-signing-account-name: ${{ inputs.account-name }}
certificate-profile-name: ${{ inputs.certificate-profile-name }}
exclude-environment-credential: true
exclude-workload-identity-credential: true
exclude-managed-identity-credential: true
exclude-shared-token-cache-credential: true
exclude-visual-studio-credential: true
exclude-visual-studio-code-credential: true
exclude-azure-cli-credential: false
exclude-azure-powershell-credential: true
exclude-azure-developer-cli-credential: true
exclude-interactive-browser-credential: true
cache-dependencies: false
files: |
${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex.exe
${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex-responses-api-proxy.exe
${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex-windows-sandbox-setup.exe
${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex-command-runner.exe

View File

@@ -55,30 +55,6 @@
"path": "codex-responses-api-proxy.exe"
}
}
},
"codex-command-runner": {
"platforms": {
"windows-x86_64": {
"regex": "^codex-command-runner-x86_64-pc-windows-msvc\\.exe\\.zst$",
"path": "codex-command-runner.exe"
},
"windows-aarch64": {
"regex": "^codex-command-runner-aarch64-pc-windows-msvc\\.exe\\.zst$",
"path": "codex-command-runner.exe"
}
}
},
"codex-windows-sandbox-setup": {
"platforms": {
"windows-x86_64": {
"regex": "^codex-windows-sandbox-setup-x86_64-pc-windows-msvc\\.exe\\.zst$",
"path": "codex-windows-sandbox-setup.exe"
},
"windows-aarch64": {
"regex": "^codex-windows-sandbox-setup-aarch64-pc-windows-msvc\\.exe\\.zst$",
"path": "codex-windows-sandbox-setup.exe"
}
}
}
}
}

View File

@@ -20,7 +20,7 @@ jobs:
run_install: false
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: 22
@@ -46,7 +46,7 @@ jobs:
echo "pack_output=$PACK_OUTPUT" >> "$GITHUB_OUTPUT"
- name: Upload staged npm package artifact
uses: actions/upload-artifact@v6
uses: actions/upload-artifact@v5
with:
name: codex-npm-staging
path: ${{ steps.stage_npm_package.outputs.pack_output }}

View File

@@ -166,7 +166,7 @@ jobs:
# avoid caching the large target dir on the gnu-dev job.
- name: Restore cargo home cache
id: cache_cargo_home_restore
uses: actions/cache/restore@v5
uses: actions/cache/restore@v4
with:
path: |
~/.cargo/bin/
@@ -207,7 +207,7 @@ jobs:
- name: Restore sccache cache (fallback)
if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
id: cache_sccache_restore
uses: actions/cache/restore@v5
uses: actions/cache/restore@v4
with:
path: ${{ github.workspace }}/.sccache/
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ steps.lockhash.outputs.hash }}-${{ github.run_id }}
@@ -226,7 +226,7 @@ jobs:
- if: ${{ matrix.target == 'x86_64-unknown-linux-musl' || matrix.target == 'aarch64-unknown-linux-musl'}}
name: Restore APT cache (musl)
id: cache_apt_restore
uses: actions/cache/restore@v5
uses: actions/cache/restore@v4
with:
path: |
/var/cache/apt
@@ -280,7 +280,7 @@ jobs:
- name: Save cargo home cache
if: always() && !cancelled() && steps.cache_cargo_home_restore.outputs.cache-hit != 'true'
continue-on-error: true
uses: actions/cache/save@v5
uses: actions/cache/save@v4
with:
path: |
~/.cargo/bin/
@@ -292,7 +292,7 @@ jobs:
- name: Save sccache cache (fallback)
if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
continue-on-error: true
uses: actions/cache/save@v5
uses: actions/cache/save@v4
with:
path: ${{ github.workspace }}/.sccache/
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ steps.lockhash.outputs.hash }}-${{ github.run_id }}
@@ -317,7 +317,7 @@ jobs:
- name: Save APT cache (musl)
if: always() && !cancelled() && (matrix.target == 'x86_64-unknown-linux-musl' || matrix.target == 'aarch64-unknown-linux-musl') && steps.cache_apt_restore.outputs.cache-hit != 'true'
continue-on-error: true
uses: actions/cache/save@v5
uses: actions/cache/save@v4
with:
path: |
/var/cache/apt
@@ -369,27 +369,6 @@ jobs:
steps:
- uses: actions/checkout@v6
# We have been running out of space when running this job on Linux for
# x86_64-unknown-linux-gnu, so remove some unnecessary dependencies.
- name: Remove unnecessary dependencies to save space
if: ${{ startsWith(matrix.runner, 'ubuntu') }}
shell: bash
run: |
set -euo pipefail
sudo rm -rf \
/usr/local/lib/android \
/usr/share/dotnet \
/usr/local/share/boost \
/usr/local/lib/node_modules \
/opt/ghc
sudo apt-get remove -y docker.io docker-compose podman buildah
# Some integration tests rely on DotSlash being installed.
# See https://github.com/openai/codex/pull/7617.
- name: Install DotSlash
uses: facebook/install-dotslash@v2
- uses: dtolnay/rust-toolchain@1.90
with:
targets: ${{ matrix.target }}
@@ -405,7 +384,7 @@ jobs:
- name: Restore cargo home cache
id: cache_cargo_home_restore
uses: actions/cache/restore@v5
uses: actions/cache/restore@v4
with:
path: |
~/.cargo/bin/
@@ -445,7 +424,7 @@ jobs:
- name: Restore sccache cache (fallback)
if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
id: cache_sccache_restore
uses: actions/cache/restore@v5
uses: actions/cache/restore@v4
with:
path: ${{ github.workspace }}/.sccache/
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ steps.lockhash.outputs.hash }}-${{ github.run_id }}
@@ -468,7 +447,7 @@ jobs:
- name: Save cargo home cache
if: always() && !cancelled() && steps.cache_cargo_home_restore.outputs.cache-hit != 'true'
continue-on-error: true
uses: actions/cache/save@v5
uses: actions/cache/save@v4
with:
path: |
~/.cargo/bin/
@@ -480,7 +459,7 @@ jobs:
- name: Save sccache cache (fallback)
if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
continue-on-error: true
uses: actions/cache/save@v5
uses: actions/cache/save@v4
with:
path: ${{ github.workspace }}/.sccache/
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ steps.lockhash.outputs.hash }}-${{ github.run_id }}

View File

@@ -1,51 +0,0 @@
name: rust-release-prepare
on:
workflow_dispatch:
schedule:
- cron: "0 */4 * * *"
concurrency:
group: ${{ github.workflow }}
cancel-in-progress: false
permissions:
contents: write
pull-requests: write
jobs:
prepare:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v6
with:
ref: main
fetch-depth: 0
- name: Update models.json
env:
OPENAI_API_KEY: ${{ secrets.CODEX_OPENAI_API_KEY }}
run: |
set -euo pipefail
client_version="99.99.99"
terminal_info="github-actions"
user_agent="codex_cli_rs/99.99.99 (Linux $(uname -r); $(uname -m)) ${terminal_info}"
base_url="${OPENAI_BASE_URL:-https://chatgpt.com/backend-api/codex}"
headers=(
-H "Authorization: Bearer ${OPENAI_API_KEY}"
-H "User-Agent: ${user_agent}"
)
url="${base_url%/}/models?client_version=${client_version}"
curl --http1.1 --fail --show-error --location "${headers[@]}" "${url}" | jq '.' > codex-rs/core/models.json
- name: Open pull request (if changed)
uses: peter-evans/create-pull-request@v7
with:
commit-message: "Update models.json"
title: "Update models.json"
body: "Automated update of models.json."
branch: "bot/update-models-json"
reviewers: "pakrym-oai,aibrahim-oai"
delete-branch: true

View File

@@ -50,9 +50,6 @@ jobs:
name: Build - ${{ matrix.runner }} - ${{ matrix.target }}
runs-on: ${{ matrix.runner }}
timeout-minutes: 30
permissions:
contents: read
id-token: write
defaults:
run:
working-directory: codex-rs
@@ -84,7 +81,7 @@ jobs:
with:
targets: ${{ matrix.target }}
- uses: actions/cache@v5
- uses: actions/cache@v4
with:
path: |
~/.cargo/bin/
@@ -101,43 +98,176 @@ jobs:
sudo apt-get install -y musl-tools pkg-config
- name: Cargo build
shell: bash
run: |
if [[ "${{ contains(matrix.target, 'windows') }}" == 'true' ]]; then
cargo build --target ${{ matrix.target }} --release --bin codex --bin codex-responses-api-proxy --bin codex-windows-sandbox-setup --bin codex-command-runner
else
cargo build --target ${{ matrix.target }} --release --bin codex --bin codex-responses-api-proxy
fi
- if: ${{ contains(matrix.target, 'linux') }}
name: Cosign Linux artifacts
uses: ./.github/actions/linux-code-sign
with:
target: ${{ matrix.target }}
artifacts-dir: ${{ github.workspace }}/codex-rs/target/${{ matrix.target }}/release
- if: ${{ contains(matrix.target, 'windows') }}
name: Sign Windows binaries with Azure Trusted Signing
uses: ./.github/actions/windows-code-sign
with:
target: ${{ matrix.target }}
client-id: ${{ secrets.AZURE_TRUSTED_SIGNING_CLIENT_ID }}
tenant-id: ${{ secrets.AZURE_TRUSTED_SIGNING_TENANT_ID }}
subscription-id: ${{ secrets.AZURE_TRUSTED_SIGNING_SUBSCRIPTION_ID }}
endpoint: ${{ secrets.AZURE_TRUSTED_SIGNING_ENDPOINT }}
account-name: ${{ secrets.AZURE_TRUSTED_SIGNING_ACCOUNT_NAME }}
certificate-profile-name: ${{ secrets.AZURE_TRUSTED_SIGNING_CERTIFICATE_PROFILE_NAME }}
run: cargo build --target ${{ matrix.target }} --release --bin codex --bin codex-responses-api-proxy
- if: ${{ matrix.runner == 'macos-15-xlarge' }}
name: MacOS code signing
uses: ./.github/actions/macos-code-sign
with:
target: ${{ matrix.target }}
apple-certificate: ${{ secrets.APPLE_CERTIFICATE_P12 }}
apple-certificate-password: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
apple-notarization-key-p8: ${{ secrets.APPLE_NOTARIZATION_KEY_P8 }}
apple-notarization-key-id: ${{ secrets.APPLE_NOTARIZATION_KEY_ID }}
apple-notarization-issuer-id: ${{ secrets.APPLE_NOTARIZATION_ISSUER_ID }}
name: Configure Apple code signing
shell: bash
env:
KEYCHAIN_PASSWORD: actions
APPLE_CERTIFICATE: ${{ secrets.APPLE_CERTIFICATE_P12 }}
APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
run: |
set -euo pipefail
if [[ -z "${APPLE_CERTIFICATE:-}" ]]; then
echo "APPLE_CERTIFICATE is required for macOS signing"
exit 1
fi
if [[ -z "${APPLE_CERTIFICATE_PASSWORD:-}" ]]; then
echo "APPLE_CERTIFICATE_PASSWORD is required for macOS signing"
exit 1
fi
cert_path="${RUNNER_TEMP}/apple_signing_certificate.p12"
echo "$APPLE_CERTIFICATE" | base64 -d > "$cert_path"
keychain_path="${RUNNER_TEMP}/codex-signing.keychain-db"
security create-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
security set-keychain-settings -lut 21600 "$keychain_path"
security unlock-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
keychain_args=()
cleanup_keychain() {
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "${keychain_args[@]}" || true
security default-keychain -s "${keychain_args[0]}" || true
else
security list-keychains -s || true
fi
if [[ -f "$keychain_path" ]]; then
security delete-keychain "$keychain_path" || true
fi
}
while IFS= read -r keychain; do
[[ -n "$keychain" ]] && keychain_args+=("$keychain")
done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "$keychain_path" "${keychain_args[@]}"
else
security list-keychains -s "$keychain_path"
fi
security default-keychain -s "$keychain_path"
security import "$cert_path" -k "$keychain_path" -P "$APPLE_CERTIFICATE_PASSWORD" -T /usr/bin/codesign -T /usr/bin/security
security set-key-partition-list -S apple-tool:,apple: -s -k "$KEYCHAIN_PASSWORD" "$keychain_path" > /dev/null
codesign_hashes=()
while IFS= read -r hash; do
[[ -n "$hash" ]] && codesign_hashes+=("$hash")
done < <(security find-identity -v -p codesigning "$keychain_path" \
| sed -n 's/.*\([0-9A-F]\{40\}\).*/\1/p' \
| sort -u)
if ((${#codesign_hashes[@]} == 0)); then
echo "No signing identities found in $keychain_path"
cleanup_keychain
rm -f "$cert_path"
exit 1
fi
if ((${#codesign_hashes[@]} > 1)); then
echo "Multiple signing identities found in $keychain_path:"
printf ' %s\n' "${codesign_hashes[@]}"
cleanup_keychain
rm -f "$cert_path"
exit 1
fi
APPLE_CODESIGN_IDENTITY="${codesign_hashes[0]}"
rm -f "$cert_path"
echo "APPLE_CODESIGN_IDENTITY=$APPLE_CODESIGN_IDENTITY" >> "$GITHUB_ENV"
echo "APPLE_CODESIGN_KEYCHAIN=$keychain_path" >> "$GITHUB_ENV"
echo "::add-mask::$APPLE_CODESIGN_IDENTITY"
- if: ${{ matrix.runner == 'macos-15-xlarge' }}
name: Sign macOS binaries
shell: bash
run: |
set -euo pipefail
if [[ -z "${APPLE_CODESIGN_IDENTITY:-}" ]]; then
echo "APPLE_CODESIGN_IDENTITY is required for macOS signing"
exit 1
fi
keychain_args=()
if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" && -f "${APPLE_CODESIGN_KEYCHAIN}" ]]; then
keychain_args+=(--keychain "${APPLE_CODESIGN_KEYCHAIN}")
fi
for binary in codex codex-responses-api-proxy; do
path="target/${{ matrix.target }}/release/${binary}"
codesign --force --options runtime --timestamp --sign "$APPLE_CODESIGN_IDENTITY" "${keychain_args[@]}" "$path"
done
- if: ${{ matrix.runner == 'macos-15-xlarge' }}
name: Notarize macOS binaries
shell: bash
env:
APPLE_NOTARIZATION_KEY_P8: ${{ secrets.APPLE_NOTARIZATION_KEY_P8 }}
APPLE_NOTARIZATION_KEY_ID: ${{ secrets.APPLE_NOTARIZATION_KEY_ID }}
APPLE_NOTARIZATION_ISSUER_ID: ${{ secrets.APPLE_NOTARIZATION_ISSUER_ID }}
run: |
set -euo pipefail
for var in APPLE_NOTARIZATION_KEY_P8 APPLE_NOTARIZATION_KEY_ID APPLE_NOTARIZATION_ISSUER_ID; do
if [[ -z "${!var:-}" ]]; then
echo "$var is required for notarization"
exit 1
fi
done
notary_key_path="${RUNNER_TEMP}/notarytool.key.p8"
echo "$APPLE_NOTARIZATION_KEY_P8" | base64 -d > "$notary_key_path"
cleanup_notary() {
rm -f "$notary_key_path"
}
trap cleanup_notary EXIT
notarize_binary() {
local binary="$1"
local source_path="target/${{ matrix.target }}/release/${binary}"
local archive_path="${RUNNER_TEMP}/${binary}.zip"
if [[ ! -f "$source_path" ]]; then
echo "Binary $source_path not found"
exit 1
fi
rm -f "$archive_path"
ditto -c -k --keepParent "$source_path" "$archive_path"
submission_json=$(xcrun notarytool submit "$archive_path" \
--key "$notary_key_path" \
--key-id "$APPLE_NOTARIZATION_KEY_ID" \
--issuer "$APPLE_NOTARIZATION_ISSUER_ID" \
--output-format json \
--wait)
status=$(printf '%s\n' "$submission_json" | jq -r '.status // "Unknown"')
submission_id=$(printf '%s\n' "$submission_json" | jq -r '.id // ""')
if [[ -z "$submission_id" ]]; then
echo "Failed to retrieve submission ID for $binary"
exit 1
fi
echo "::notice title=Notarization::$binary submission ${submission_id} completed with status ${status}"
if [[ "$status" != "Accepted" ]]; then
echo "Notarization failed for ${binary} (submission ${submission_id}, status ${status})"
exit 1
fi
}
notarize_binary "codex"
notarize_binary "codex-responses-api-proxy"
- name: Stage artifacts
shell: bash
@@ -148,18 +278,11 @@ jobs:
if [[ "${{ matrix.runner }}" == windows* ]]; then
cp target/${{ matrix.target }}/release/codex.exe "$dest/codex-${{ matrix.target }}.exe"
cp target/${{ matrix.target }}/release/codex-responses-api-proxy.exe "$dest/codex-responses-api-proxy-${{ matrix.target }}.exe"
cp target/${{ matrix.target }}/release/codex-windows-sandbox-setup.exe "$dest/codex-windows-sandbox-setup-${{ matrix.target }}.exe"
cp target/${{ matrix.target }}/release/codex-command-runner.exe "$dest/codex-command-runner-${{ matrix.target }}.exe"
else
cp target/${{ matrix.target }}/release/codex "$dest/codex-${{ matrix.target }}"
cp target/${{ matrix.target }}/release/codex-responses-api-proxy "$dest/codex-responses-api-proxy-${{ matrix.target }}"
fi
if [[ "${{ matrix.target }}" == *linux* ]]; then
cp target/${{ matrix.target }}/release/codex.sigstore "$dest/codex-${{ matrix.target }}.sigstore"
cp target/${{ matrix.target }}/release/codex-responses-api-proxy.sigstore "$dest/codex-responses-api-proxy-${{ matrix.target }}.sigstore"
fi
- if: ${{ matrix.runner == 'windows-11-arm' }}
name: Install zstd
shell: powershell
@@ -198,11 +321,6 @@ jobs:
continue
fi
# Don't try to compress signature bundles.
if [[ "$base" == *.sigstore ]]; then
continue
fi
# Create per-binary tar.gz
tar -C "$dest" -czf "$dest/${base}.tar.gz" "$base"
@@ -222,7 +340,30 @@ jobs:
zstd "${zstd_args[@]}" "$dest/$base"
done
- uses: actions/upload-artifact@v6
- name: Remove signing keychain
if: ${{ always() && matrix.runner == 'macos-15-xlarge' }}
shell: bash
env:
APPLE_CODESIGN_KEYCHAIN: ${{ env.APPLE_CODESIGN_KEYCHAIN }}
run: |
set -euo pipefail
if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" ]]; then
keychain_args=()
while IFS= read -r keychain; do
[[ "$keychain" == "$APPLE_CODESIGN_KEYCHAIN" ]] && continue
[[ -n "$keychain" ]] && keychain_args+=("$keychain")
done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
if ((${#keychain_args[@]} > 0)); then
security list-keychains -s "${keychain_args[@]}"
security default-keychain -s "${keychain_args[0]}"
fi
if [[ -f "$APPLE_CODESIGN_KEYCHAIN" ]]; then
security delete-keychain "$APPLE_CODESIGN_KEYCHAIN"
fi
fi
- uses: actions/upload-artifact@v5
with:
name: ${{ matrix.target }}
# Upload the per-binary .zst files as well as the new .tar.gz
@@ -258,7 +399,7 @@ jobs:
- name: Checkout repository
uses: actions/checkout@v6
- uses: actions/download-artifact@v7
- uses: actions/download-artifact@v4
with:
path: dist
@@ -306,7 +447,7 @@ jobs:
run_install: false
- name: Setup Node.js for npm packaging
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: 22
@@ -357,7 +498,7 @@ jobs:
steps:
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: 22
registry-url: "https://registry.npmjs.org"

View File

@@ -19,7 +19,7 @@ jobs:
run_install: false
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: 22
cache: pnpm

View File

@@ -30,7 +30,7 @@ jobs:
run_install: false
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: ${{ env.NODE_VERSION }}
cache: "pnpm"

View File

@@ -113,7 +113,7 @@ jobs:
cp "target/${{ matrix.target }}/release/codex-exec-mcp-server" "$dest/"
cp "target/${{ matrix.target }}/release/codex-execve-wrapper" "$dest/"
- uses: actions/upload-artifact@v6
- uses: actions/upload-artifact@v5
with:
name: shell-tool-mcp-rust-${{ matrix.target }}
path: artifacts/**
@@ -211,7 +211,7 @@ jobs:
mkdir -p "$dest"
cp bash "$dest/bash"
- uses: actions/upload-artifact@v6
- uses: actions/upload-artifact@v5
with:
name: shell-tool-mcp-bash-${{ matrix.target }}-${{ matrix.variant }}
path: artifacts/**
@@ -253,7 +253,7 @@ jobs:
mkdir -p "$dest"
cp bash "$dest/bash"
- uses: actions/upload-artifact@v6
- uses: actions/upload-artifact@v5
with:
name: shell-tool-mcp-bash-${{ matrix.target }}-${{ matrix.variant }}
path: artifacts/**
@@ -280,7 +280,7 @@ jobs:
run_install: false
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: ${{ env.NODE_VERSION }}
@@ -291,7 +291,7 @@ jobs:
run: pnpm --filter @openai/codex-shell-tool-mcp run build
- name: Download build artifacts
uses: actions/download-artifact@v7
uses: actions/download-artifact@v4
with:
path: artifacts
@@ -352,7 +352,7 @@ jobs:
filename=$(PACK_INFO="$pack_info" node -e 'const data = JSON.parse(process.env.PACK_INFO); console.log(data[0].filename);')
mv "dist/npm/${filename}" "dist/npm/codex-shell-tool-mcp-npm-${PACKAGE_VERSION}.tgz"
- uses: actions/upload-artifact@v6
- uses: actions/upload-artifact@v5
with:
name: codex-shell-tool-mcp-npm
path: dist/npm/codex-shell-tool-mcp-npm-${{ env.PACKAGE_VERSION }}.tgz
@@ -376,7 +376,7 @@ jobs:
run_install: false
- name: Setup Node.js
uses: actions/setup-node@v6
uses: actions/setup-node@v5
with:
node-version: ${{ env.NODE_VERSION }}
registry-url: https://registry.npmjs.org
@@ -386,7 +386,7 @@ jobs:
run: npm install -g npm@latest
- name: Download npm tarball
uses: actions/download-artifact@v7
uses: actions/download-artifact@v4
with:
name: codex-shell-tool-mcp-npm
path: dist/npm

View File

@@ -11,6 +11,7 @@ In the codex-rs folder where the rust code lives:
- Always collapse if statements per https://rust-lang.github.io/rust-clippy/master/index.html#collapsible_if
- Always inline format! args when possible per https://rust-lang.github.io/rust-clippy/master/index.html#uninlined_format_args
- Use method references over closures when possible per https://rust-lang.github.io/rust-clippy/master/index.html#redundant_closure_for_method_calls
- Do not use unsigned integer even if the number cannot be negative.
- When writing tests, prefer comparing the equality of entire objects over fields one by one.
- When making a change that adds or changes an API, ensure that the documentation in the `docs/` folder is up to date if applicable.
@@ -75,7 +76,6 @@ If you dont have the tool:
- Tests should use pretty_assertions::assert_eq for clearer diffs. Import this at the top of the test module if it isn't already.
- 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.
### Integration tests (core)

View File

@@ -95,14 +95,6 @@ function detectPackageManager() {
return "bun";
}
if (
__dirname.includes(".bun/install/global") ||
__dirname.includes(".bun\\install\\global")
) {
return "bun";
}
return userAgent ? "npm" : null;
}

View File

@@ -20,14 +20,9 @@ PACKAGE_NATIVE_COMPONENTS: dict[str, list[str]] = {
"codex-responses-api-proxy": ["codex-responses-api-proxy"],
"codex-sdk": ["codex"],
}
WINDOWS_ONLY_COMPONENTS: dict[str, list[str]] = {
"codex": ["codex-windows-sandbox-setup", "codex-command-runner"],
}
COMPONENT_DEST_DIR: dict[str, str] = {
"codex": "codex",
"codex-responses-api-proxy": "codex-responses-api-proxy",
"codex-windows-sandbox-setup": "codex",
"codex-command-runner": "codex",
"rg": "path",
}
@@ -108,7 +103,7 @@ def main() -> int:
"pointing to a directory containing pre-installed binaries."
)
copy_native_binaries(vendor_src, staging_dir, package, native_components)
copy_native_binaries(vendor_src, staging_dir, native_components)
if release_version:
staging_dir_str = str(staging_dir)
@@ -237,12 +232,7 @@ def stage_codex_sdk_sources(staging_dir: Path) -> None:
shutil.copy2(license_src, staging_dir / "LICENSE")
def copy_native_binaries(
vendor_src: Path,
staging_dir: Path,
package: str,
components: list[str],
) -> None:
def copy_native_binaries(vendor_src: Path, staging_dir: Path, components: list[str]) -> None:
vendor_src = vendor_src.resolve()
if not vendor_src.exists():
raise RuntimeError(f"Vendor source directory not found: {vendor_src}")
@@ -260,9 +250,6 @@ def copy_native_binaries(
if not target_dir.is_dir():
continue
if "windows" in target_dir.name:
components_set.update(WINDOWS_ONLY_COMPONENTS.get(package, []))
dest_target_dir = vendor_dest / target_dir.name
dest_target_dir.mkdir(parents=True, exist_ok=True)

View File

@@ -36,11 +36,8 @@ class BinaryComponent:
artifact_prefix: str # matches the artifact filename prefix (e.g. codex-<target>.zst)
dest_dir: str # directory under vendor/<target>/ where the binary is installed
binary_basename: str # executable name inside dest_dir (before optional .exe)
targets: tuple[str, ...] | None = None # limit installation to specific targets
WINDOWS_TARGETS = tuple(target for target in BINARY_TARGETS if "windows" in target)
BINARY_COMPONENTS = {
"codex": BinaryComponent(
artifact_prefix="codex",
@@ -52,18 +49,6 @@ BINARY_COMPONENTS = {
dest_dir="codex-responses-api-proxy",
binary_basename="codex-responses-api-proxy",
),
"codex-windows-sandbox-setup": BinaryComponent(
artifact_prefix="codex-windows-sandbox-setup",
dest_dir="codex",
binary_basename="codex-windows-sandbox-setup",
targets=WINDOWS_TARGETS,
),
"codex-command-runner": BinaryComponent(
artifact_prefix="codex-command-runner",
dest_dir="codex",
binary_basename="codex-command-runner",
targets=WINDOWS_TARGETS,
),
}
RG_TARGET_PLATFORM_PAIRS: list[tuple[str, str]] = [
@@ -94,8 +79,7 @@ def parse_args() -> argparse.Namespace:
choices=tuple(list(BINARY_COMPONENTS) + ["rg"]),
help=(
"Limit installation to the specified components."
" May be repeated. Defaults to codex, codex-windows-sandbox-setup,"
" codex-command-runner, and rg."
" May be repeated. Defaults to 'codex' and 'rg'."
),
)
parser.add_argument(
@@ -117,12 +101,7 @@ def main() -> int:
vendor_dir = codex_cli_root / VENDOR_DIR_NAME
vendor_dir.mkdir(parents=True, exist_ok=True)
components = args.components or [
"codex",
"codex-windows-sandbox-setup",
"codex-command-runner",
"rg",
]
components = args.components or ["codex", "rg"]
workflow_url = (args.workflow_url or DEFAULT_WORKFLOW_URL).strip()
if not workflow_url:
@@ -137,7 +116,8 @@ def main() -> int:
install_binary_components(
artifacts_dir,
vendor_dir,
[BINARY_COMPONENTS[name] for name in components if name in BINARY_COMPONENTS],
BINARY_TARGETS,
[name for name in components if name in BINARY_COMPONENTS],
)
if "rg" in components:
@@ -226,19 +206,23 @@ def _download_artifacts(workflow_id: str, dest_dir: Path) -> None:
def install_binary_components(
artifacts_dir: Path,
vendor_dir: Path,
selected_components: Sequence[BinaryComponent],
targets: Iterable[str],
component_names: Sequence[str],
) -> None:
selected_components = [BINARY_COMPONENTS[name] for name in component_names if name in BINARY_COMPONENTS]
if not selected_components:
return
for component in selected_components:
component_targets = list(component.targets or BINARY_TARGETS)
targets = list(targets)
if not targets:
return
for component in selected_components:
print(
f"Installing {component.binary_basename} binaries for targets: "
+ ", ".join(component_targets)
+ ", ".join(targets)
)
max_workers = min(len(component_targets), max(1, (os.cpu_count() or 1)))
max_workers = min(len(targets), max(1, (os.cpu_count() or 1)))
with ThreadPoolExecutor(max_workers=max_workers) as executor:
futures = {
executor.submit(
@@ -248,7 +232,7 @@ def install_binary_components(
target,
component,
): target
for target in component_targets
for target in targets
}
for future in as_completed(futures):
installed_path = future.result()

722
codex-rs/Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -34,8 +34,6 @@ members = [
"stdio-to-uds",
"otel",
"tui",
"tui2",
"utils/absolute-path",
"utils/git",
"utils/cache",
"utils/image",
@@ -90,8 +88,6 @@ codex-responses-api-proxy = { path = "responses-api-proxy" }
codex-rmcp-client = { path = "rmcp-client" }
codex-stdio-to-uds = { path = "stdio-to-uds" }
codex-tui = { path = "tui" }
codex-tui2 = { path = "tui2" }
codex-utils-absolute-path = { path = "utils/absolute-path" }
codex-utils-cache = { path = "utils/cache" }
codex-utils-image = { path = "utils/image" }
codex-utils-json-to-toml = { path = "utils/json-to-toml" }
@@ -100,7 +96,6 @@ codex-utils-readiness = { path = "utils/readiness" }
codex-utils-string = { path = "utils/string" }
codex-windows-sandbox = { path = "windows-sandbox-rs" }
core_test_support = { path = "core/tests/common" }
exec_server_test_support = { path = "exec-server/tests/common" }
mcp-types = { path = "mcp-types" }
mcp_test_support = { path = "mcp-server/tests/common" }
@@ -109,6 +104,7 @@ allocative = "0.3.3"
ansi-to-tui = "7.0.0"
anyhow = "1"
arboard = { version = "3", features = ["wayland-data-control"] }
askama = "0.14"
assert_cmd = "2"
assert_matches = "1.5.0"
async-channel = "2.3.1"
@@ -142,14 +138,14 @@ icu_provider = { version = "2.1", features = ["sync"] }
ignore = "0.4.23"
image = { version = "^0.25.9", default-features = false }
indexmap = "2.12.0"
insta = "1.44.3"
insta = "1.43.2"
itertools = "0.14.0"
keyring = { version = "3.6", default-features = false }
landlock = "0.4.1"
lazy_static = "1"
libc = "0.2.177"
log = "0.4"
lru = "0.16.2"
lru = "0.12.5"
maplit = "1.0.2"
mime_guess = "2.0.5"
multimap = "0.10.0"
@@ -162,7 +158,6 @@ 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"
@@ -180,7 +175,7 @@ reqwest = "0.12"
rmcp = { version = "0.10.0", default-features = false }
schemars = "0.8.22"
seccompiler = "0.5.0"
sentry = "0.46.0"
sentry = "0.34.0"
serde = "1"
serde_json = "1"
serde_with = "3.16"
@@ -190,7 +185,7 @@ sha1 = "0.10.6"
sha2 = "0.10"
shlex = "1.3.0"
similar = "2.7.0"
socket2 = "0.6.1"
socket2 = "0.6.0"
starlark = "0.13.0"
strum = "0.27.2"
strum_macros = "0.27.2"
@@ -227,7 +222,7 @@ vt100 = "0.16.2"
walkdir = "2.5.0"
webbrowser = "1.0"
which = "6"
wildmatch = "2.6.1"
wildmatch = "2.5.0"
wiremock = "0.6"
zeroize = "1.8.2"
@@ -293,6 +288,7 @@ opt-level = 0
# Uncomment to debug local changes.
# ratatui = { path = "../../ratatui" }
crossterm = { git = "https://github.com/nornagon/crossterm", branch = "nornagon/color-query" }
portable-pty = { git = "https://github.com/pakrym/wezterm", branch = "PSUEDOCONSOLE_INHERIT_CURSOR" }
ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.29.0-patch" }
# Uncomment to debug local changes.

View File

@@ -46,7 +46,7 @@ Use `codex mcp` to add/list/get/remove MCP server launchers defined in `config.t
### Notifications
You can enable notifications by configuring a script that is run whenever the agent finishes a turn. The [notify documentation](../docs/config.md#notify) includes a detailed example that explains how to get desktop notifications via [terminal-notifier](https://github.com/julienXX/terminal-notifier) on macOS. When Codex detects that it is running under WSL 2 inside Windows Terminal (`WT_SESSION` is set), the TUI automatically falls back to native Windows toast notifications so approval prompts and completed turns surface even though Windows Terminal does not implement OSC 9.
You can enable notifications by configuring a script that is run whenever the agent finishes a turn. The [notify documentation](../docs/config.md#notify) includes a detailed example that explains how to get desktop notifications via [terminal-notifier](https://github.com/julienXX/terminal-notifier) on macOS.
### `codex exec` to run Codex programmatically/non-interactively

View File

@@ -15,7 +15,6 @@ workspace = true
anyhow = { workspace = true }
clap = { workspace = true, features = ["derive"] }
codex-protocol = { workspace = true }
codex-utils-absolute-path = { workspace = true }
mcp-types = { workspace = true }
schemars = { workspace = true }
serde = { workspace = true, features = ["derive"] }

View File

@@ -31,7 +31,6 @@ use std::process::Command;
use ts_rs::TS;
const HEADER: &str = "// GENERATED CODE! DO NOT MODIFY BY HAND!\n\n";
const IGNORED_DEFINITIONS: &[&str] = &["Option<()>"];
#[derive(Clone)]
pub struct GeneratedSchema {
@@ -185,6 +184,7 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
"ServerNotification",
"ServerRequest",
];
const IGNORED_DEFINITIONS: &[&str] = &["Option<()>"];
let namespaced_types = collect_namespaced_types(&schemas);
let mut definitions = Map::new();
@@ -304,11 +304,8 @@ where
out_dir.join(format!("{file_stem}.json"))
};
if !IGNORED_DEFINITIONS.contains(&logical_name) {
write_pretty_json(out_path, &schema_value)
.with_context(|| format!("Failed to write JSON schema for {file_stem}"))?;
}
write_pretty_json(out_path, &schema_value)
.with_context(|| format!("Failed to write JSON schema for {file_stem}"))?;
let namespace = match raw_namespace {
Some("v1") | None => None,
Some(ns) => Some(ns.to_string()),

View File

@@ -117,9 +117,9 @@ client_request_definitions! {
params: v2::ThreadListParams,
response: v2::ThreadListResponse,
},
SkillsList => "skills/list" {
params: v2::SkillsListParams,
response: v2::SkillsListResponse,
ThreadCompact => "thread/compact" {
params: v2::ThreadCompactParams,
response: v2::ThreadCompactResponse,
},
TurnStart => "turn/start" {
params: v2::TurnStartParams,
@@ -139,14 +139,9 @@ client_request_definitions! {
response: v2::ModelListResponse,
},
McpServerOauthLogin => "mcpServer/oauth/login" {
params: v2::McpServerOauthLoginParams,
response: v2::McpServerOauthLoginResponse,
},
McpServerStatusList => "mcpServerStatus/list" {
params: v2::ListMcpServerStatusParams,
response: v2::ListMcpServerStatusResponse,
McpServersList => "mcpServers/list" {
params: v2::ListMcpServersParams,
response: v2::ListMcpServersResponse,
},
LoginAccount => "account/login/start" {
@@ -525,14 +520,10 @@ server_notification_definitions! {
TurnPlanUpdated => "turn/plan/updated" (v2::TurnPlanUpdatedNotification),
ItemStarted => "item/started" (v2::ItemStartedNotification),
ItemCompleted => "item/completed" (v2::ItemCompletedNotification),
/// This event is internal-only. Used by Codex Cloud.
RawResponseItemCompleted => "rawResponseItem/completed" (v2::RawResponseItemCompletedNotification),
AgentMessageDelta => "item/agentMessage/delta" (v2::AgentMessageDeltaNotification),
CommandExecutionOutputDelta => "item/commandExecution/outputDelta" (v2::CommandExecutionOutputDeltaNotification),
TerminalInteraction => "item/commandExecution/terminalInteraction" (v2::TerminalInteractionNotification),
FileChangeOutputDelta => "item/fileChange/outputDelta" (v2::FileChangeOutputDeltaNotification),
McpToolCallProgress => "item/mcpToolCall/progress" (v2::McpToolCallProgressNotification),
McpServerOauthLoginCompleted => "mcpServer/oauthLogin/completed" (v2::McpServerOauthLoginCompletedNotification),
AccountUpdated => "account/updated" (v2::AccountUpdatedNotification),
AccountRateLimitsUpdated => "account/rateLimits/updated" (v2::AccountRateLimitsUpdatedNotification),
ReasoningSummaryTextDelta => "item/reasoning/summaryTextDelta" (v2::ReasoningSummaryTextDeltaNotification),
@@ -656,6 +647,7 @@ mod tests {
command: vec!["echo".to_string(), "hello".to_string()],
cwd: PathBuf::from("/tmp"),
reason: Some("because tests".to_string()),
risk: None,
parsed_cmd: vec![ParsedCommand::Unknown {
cmd: "echo hello".to_string(),
}],
@@ -675,6 +667,7 @@ mod tests {
"command": ["echo", "hello"],
"cwd": "/tmp",
"reason": "because tests",
"risk": null,
"parsedCmd": [
{
"type": "unknown",

View File

@@ -13,10 +13,10 @@ use codex_protocol::protocol::AskForApproval;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::FileChange;
use codex_protocol::protocol::ReviewDecision;
use codex_protocol::protocol::SandboxCommandAssessment;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::TurnAbortReason;
use codex_utils_absolute_path::AbsolutePathBuf;
use schemars::JsonSchema;
use serde::Deserialize;
use serde::Serialize;
@@ -226,6 +226,7 @@ pub struct ExecCommandApprovalParams {
pub command: Vec<String>,
pub cwd: PathBuf,
pub reason: Option<String>,
pub risk: Option<SandboxCommandAssessment>,
pub parsed_cmd: Vec<ParsedCommand>,
}
@@ -360,7 +361,7 @@ pub struct Tools {
#[serde(rename_all = "camelCase")]
pub struct SandboxSettings {
#[serde(default)]
pub writable_roots: Vec<AbsolutePathBuf>,
pub writable_roots: Vec<PathBuf>,
pub network_access: Option<bool>,
pub exclude_tmpdir_env_var: Option<bool>,
pub exclude_slash_tmp: Option<bool>,

View File

@@ -3,11 +3,8 @@ use std::path::PathBuf;
use crate::protocol::common::AuthMode;
use codex_protocol::account::PlanType;
use codex_protocol::approvals::ExecPolicyAmendment as CoreExecPolicyAmendment;
use codex_protocol::config_types::ForcedLoginMethod;
use codex_protocol::approvals::SandboxCommandAssessment as CoreSandboxCommandAssessment;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::config_types::SandboxMode as CoreSandboxMode;
use codex_protocol::config_types::Verbosity;
use codex_protocol::items::AgentMessageContent as CoreAgentMessageContent;
use codex_protocol::items::TurnItem as CoreTurnItem;
use codex_protocol::models::ResponseItem;
@@ -15,19 +12,14 @@ use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::parse_command::ParsedCommand as CoreParsedCommand;
use codex_protocol::plan_tool::PlanItemArg as CorePlanItemArg;
use codex_protocol::plan_tool::StepStatus as CorePlanStepStatus;
use codex_protocol::protocol::AskForApproval as CoreAskForApproval;
use codex_protocol::protocol::CodexErrorInfo as CoreCodexErrorInfo;
use codex_protocol::protocol::CreditsSnapshot as CoreCreditsSnapshot;
use codex_protocol::protocol::RateLimitSnapshot as CoreRateLimitSnapshot;
use codex_protocol::protocol::RateLimitWindow as CoreRateLimitWindow;
use codex_protocol::protocol::SessionSource as CoreSessionSource;
use codex_protocol::protocol::SkillErrorInfo as CoreSkillErrorInfo;
use codex_protocol::protocol::SkillMetadata as CoreSkillMetadata;
use codex_protocol::protocol::SkillScope as CoreSkillScope;
use codex_protocol::protocol::TokenUsage as CoreTokenUsage;
use codex_protocol::protocol::TokenUsageInfo as CoreTokenUsageInfo;
use codex_protocol::user_input::UserInput as CoreUserInput;
use codex_utils_absolute_path::AbsolutePathBuf;
use mcp_types::ContentBlock as McpContentBlock;
use mcp_types::Resource as McpResource;
use mcp_types::ResourceTemplate as McpResourceTemplate;
@@ -130,68 +122,17 @@ impl From<CoreCodexErrorInfo> for CodexErrorInfo {
}
}
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "kebab-case")]
#[ts(rename_all = "kebab-case", export_to = "v2/")]
pub enum AskForApproval {
#[serde(rename = "untrusted")]
#[ts(rename = "untrusted")]
UnlessTrusted,
OnFailure,
OnRequest,
Never,
}
impl AskForApproval {
pub fn to_core(self) -> CoreAskForApproval {
match self {
AskForApproval::UnlessTrusted => CoreAskForApproval::UnlessTrusted,
AskForApproval::OnFailure => CoreAskForApproval::OnFailure,
AskForApproval::OnRequest => CoreAskForApproval::OnRequest,
AskForApproval::Never => CoreAskForApproval::Never,
}
v2_enum_from_core!(
pub enum AskForApproval from codex_protocol::protocol::AskForApproval {
UnlessTrusted, OnFailure, OnRequest, Never
}
}
);
impl From<CoreAskForApproval> for AskForApproval {
fn from(value: CoreAskForApproval) -> Self {
match value {
CoreAskForApproval::UnlessTrusted => AskForApproval::UnlessTrusted,
CoreAskForApproval::OnFailure => AskForApproval::OnFailure,
CoreAskForApproval::OnRequest => AskForApproval::OnRequest,
CoreAskForApproval::Never => AskForApproval::Never,
}
v2_enum_from_core!(
pub enum SandboxMode from codex_protocol::config_types::SandboxMode {
ReadOnly, WorkspaceWrite, DangerFullAccess
}
}
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "kebab-case")]
#[ts(rename_all = "kebab-case", export_to = "v2/")]
pub enum SandboxMode {
ReadOnly,
WorkspaceWrite,
DangerFullAccess,
}
impl SandboxMode {
pub fn to_core(self) -> CoreSandboxMode {
match self {
SandboxMode::ReadOnly => CoreSandboxMode::ReadOnly,
SandboxMode::WorkspaceWrite => CoreSandboxMode::WorkspaceWrite,
SandboxMode::DangerFullAccess => CoreSandboxMode::DangerFullAccess,
}
}
}
impl From<CoreSandboxMode> for SandboxMode {
fn from(value: CoreSandboxMode) -> Self {
match value {
CoreSandboxMode::ReadOnly => SandboxMode::ReadOnly,
CoreSandboxMode::WorkspaceWrite => SandboxMode::WorkspaceWrite,
CoreSandboxMode::DangerFullAccess => SandboxMode::DangerFullAccess,
}
}
}
);
v2_enum_from_core!(
pub enum ReviewDelivery from codex_protocol::protocol::ReviewDelivery {
@@ -218,72 +159,6 @@ pub enum ConfigLayerName {
User,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
pub struct SandboxWorkspaceWrite {
#[serde(default)]
pub writable_roots: Vec<PathBuf>,
#[serde(default)]
pub network_access: bool,
#[serde(default)]
pub exclude_tmpdir_env_var: bool,
#[serde(default)]
pub exclude_slash_tmp: bool,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
pub struct ToolsV2 {
#[serde(alias = "web_search_request")]
pub web_search: Option<bool>,
pub view_image: Option<bool>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
pub struct ProfileV2 {
pub model: Option<String>,
pub model_provider: Option<String>,
pub approval_policy: Option<AskForApproval>,
pub model_reasoning_effort: Option<ReasoningEffort>,
pub model_reasoning_summary: Option<ReasoningSummary>,
pub model_verbosity: Option<Verbosity>,
pub chatgpt_base_url: Option<String>,
#[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/")]
pub struct Config {
pub model: Option<String>,
pub review_model: Option<String>,
pub model_context_window: Option<i64>,
pub model_auto_compact_token_limit: Option<i64>,
pub model_provider: Option<String>,
pub approval_policy: Option<AskForApproval>,
pub sandbox_mode: Option<SandboxMode>,
pub sandbox_workspace_write: Option<SandboxWorkspaceWrite>,
pub forced_chatgpt_workspace_id: Option<String>,
pub forced_login_method: Option<ForcedLoginMethod>,
pub tools: Option<ToolsV2>,
pub profile: Option<String>,
#[serde(default)]
pub profiles: HashMap<String, ProfileV2>,
pub instructions: Option<String>,
pub developer_instructions: Option<String>,
pub compact_prompt: Option<String>,
pub model_reasoning_effort: Option<ReasoningEffort>,
pub model_reasoning_summary: Option<ReasoningSummary>,
pub model_verbosity: Option<Verbosity>,
#[serde(default, flatten)]
pub additional: HashMap<String, JsonValue>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -362,7 +237,7 @@ pub struct ConfigReadParams {
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ConfigReadResponse {
pub config: Config,
pub config: JsonValue,
pub origins: HashMap<String, ConfigLayerMetadata>,
#[serde(skip_serializing_if = "Option::is_none")]
pub layers: Option<Vec<ConfigLayer>>,
@@ -399,16 +274,19 @@ pub struct ConfigEdit {
pub merge_strategy: MergeStrategy,
}
v2_enum_from_core!(
pub enum CommandRiskLevel from codex_protocol::approvals::SandboxRiskLevel {
Low,
Medium,
High
}
);
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub enum ApprovalDecision {
Accept,
/// Approve and remember the approval for the session.
AcceptForSession,
AcceptWithExecpolicyAmendment {
execpolicy_amendment: ExecPolicyAmendment,
},
Decline,
Cancel,
}
@@ -424,7 +302,7 @@ pub enum SandboxPolicy {
#[ts(rename_all = "camelCase")]
WorkspaceWrite {
#[serde(default)]
writable_roots: Vec<AbsolutePathBuf>,
writable_roots: Vec<PathBuf>,
#[serde(default)]
network_access: bool,
#[serde(default)]
@@ -478,23 +356,28 @@ impl From<codex_protocol::protocol::SandboxPolicy> for SandboxPolicy {
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(transparent)]
#[ts(type = "Array<string>", export_to = "v2/")]
pub struct ExecPolicyAmendment {
pub command: Vec<String>,
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SandboxCommandAssessment {
pub description: String,
pub risk_level: CommandRiskLevel,
}
impl ExecPolicyAmendment {
pub fn into_core(self) -> CoreExecPolicyAmendment {
CoreExecPolicyAmendment::new(self.command)
impl SandboxCommandAssessment {
pub fn into_core(self) -> CoreSandboxCommandAssessment {
CoreSandboxCommandAssessment {
description: self.description,
risk_level: self.risk_level.to_core(),
}
}
}
impl From<CoreExecPolicyAmendment> for ExecPolicyAmendment {
fn from(value: CoreExecPolicyAmendment) -> Self {
impl From<CoreSandboxCommandAssessment> for SandboxCommandAssessment {
fn from(value: CoreSandboxCommandAssessment) -> Self {
Self {
command: value.command().to_vec(),
description: value.description,
risk_level: CommandRiskLevel::from(value.risk_level),
}
}
}
@@ -672,21 +555,10 @@ pub struct CancelLoginAccountParams {
pub login_id: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub enum CancelLoginAccountStatus {
Canceled,
NotFound,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct CancelLoginAccountResponse {
pub status: CancelLoginAccountStatus,
}
pub struct CancelLoginAccountResponse {}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
@@ -761,7 +633,7 @@ pub struct ModelListResponse {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ListMcpServerStatusParams {
pub struct ListMcpServersParams {
/// Opaque pagination cursor returned by a previous call.
pub cursor: Option<String>,
/// Optional page size; defaults to a server-defined value.
@@ -771,7 +643,7 @@ pub struct ListMcpServerStatusParams {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct McpServerStatus {
pub struct McpServer {
pub name: String,
pub tools: std::collections::HashMap<String, McpTool>,
pub resources: Vec<McpResource>,
@@ -782,33 +654,13 @@ pub struct McpServerStatus {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ListMcpServerStatusResponse {
pub data: Vec<McpServerStatus>,
pub struct ListMcpServersResponse {
pub data: Vec<McpServer>,
/// Opaque cursor to pass to the next call to continue after the last item.
/// If None, there are no more items to return.
pub next_cursor: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct McpServerOauthLoginParams {
pub name: String,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub scopes: Option<Vec<String>>,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub timeout_secs: Option<i64>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct McpServerOauthLoginResponse {
pub authorization_url: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -860,12 +712,6 @@ pub struct ThreadStartParams {
pub config: Option<HashMap<String, JsonValue>>,
pub base_instructions: Option<String>,
pub developer_instructions: Option<String>,
/// If true, opt into emitting raw response items on the event stream.
///
/// This is for internal use only (e.g. Codex Cloud).
/// (TODO): Figure out a better way to categorize internal / experimental events & protocols.
#[serde(default)]
pub experimental_raw_events: bool,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -967,85 +813,14 @@ pub struct ThreadListResponse {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillsListParams {
/// When empty, defaults to the current session working directory.
#[serde(default, skip_serializing_if = "Vec::is_empty")]
pub cwds: Vec<PathBuf>,
pub struct ThreadCompactParams {
pub thread_id: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillsListResponse {
pub data: Vec<SkillsListEntry>,
}
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "snake_case")]
#[ts(rename_all = "snake_case")]
#[ts(export_to = "v2/")]
pub enum SkillScope {
User,
Repo,
Public,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillMetadata {
pub name: String,
pub description: String,
pub path: PathBuf,
pub scope: SkillScope,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillErrorInfo {
pub path: PathBuf,
pub message: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillsListEntry {
pub cwd: PathBuf,
pub skills: Vec<SkillMetadata>,
pub errors: Vec<SkillErrorInfo>,
}
impl From<CoreSkillMetadata> for SkillMetadata {
fn from(value: CoreSkillMetadata) -> Self {
Self {
name: value.name,
description: value.description,
path: value.path,
scope: value.scope.into(),
}
}
}
impl From<CoreSkillScope> for SkillScope {
fn from(value: CoreSkillScope) -> Self {
match value {
CoreSkillScope::User => Self::User,
CoreSkillScope::Repo => Self::Repo,
CoreSkillScope::Public => Self::Public,
}
}
}
impl From<CoreSkillErrorInfo> for SkillErrorInfo {
fn from(value: CoreSkillErrorInfo) -> Self {
Self {
path: value.path,
message: value.message,
}
}
}
pub struct ThreadCompactResponse {}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
@@ -1589,15 +1364,6 @@ pub struct ItemCompletedNotification {
pub turn_id: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct RawResponseItemCompletedNotification {
pub thread_id: String,
pub turn_id: String,
pub item: ResponseItem,
}
// Item-specific progress notifications
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
@@ -1644,17 +1410,6 @@ pub struct ReasoningTextDeltaNotification {
pub content_index: i64,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct TerminalInteractionNotification {
pub thread_id: String,
pub turn_id: String,
pub item_id: String,
pub process_id: String,
pub stdin: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -1685,17 +1440,6 @@ pub struct McpToolCallProgressNotification {
pub message: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct McpServerOauthLoginCompletedNotification {
pub name: String,
pub success: bool,
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional)]
pub error: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -1722,8 +1466,17 @@ pub struct CommandExecutionRequestApprovalParams {
pub item_id: String,
/// Optional explanatory reason (e.g. request for network access).
pub reason: Option<String>,
/// Optional proposed execpolicy amendment to allow similar commands without prompting.
pub proposed_execpolicy_amendment: Option<ExecPolicyAmendment>,
/// Optional model-provided risk assessment describing the blocked command.
pub risk: Option<SandboxCommandAssessment>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct CommandExecutionRequestAcceptSettings {
/// If true, automatically approve this command for the duration of the session.
#[serde(default)]
pub for_session: bool,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1731,6 +1484,10 @@ pub struct CommandExecutionRequestApprovalParams {
#[ts(export_to = "v2/")]
pub struct CommandExecutionRequestApprovalResponse {
pub decision: ApprovalDecision,
/// Optional approval settings for when the decision is `accept`.
/// Ignored if the decision is `decline` or `cancel`.
#[serde(default)]
pub accept_settings: Option<CommandExecutionRequestAcceptSettings>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]

View File

@@ -21,6 +21,7 @@ use codex_app_server_protocol::ApprovalDecision;
use codex_app_server_protocol::AskForApproval;
use codex_app_server_protocol::ClientInfo;
use codex_app_server_protocol::ClientRequest;
use codex_app_server_protocol::CommandExecutionRequestAcceptSettings;
use codex_app_server_protocol::CommandExecutionRequestApprovalParams;
use codex_app_server_protocol::CommandExecutionRequestApprovalResponse;
use codex_app_server_protocol::FileChangeRequestApprovalParams;
@@ -553,10 +554,6 @@ impl CodexClient {
print!("{}", delta.delta);
std::io::stdout().flush().ok();
}
ServerNotification::TerminalInteraction(delta) => {
println!("[stdin sent: {}]", delta.stdin);
std::io::stdout().flush().ok();
}
ServerNotification::ItemStarted(payload) => {
println!("\n< item started: {:?}", payload.item);
}
@@ -756,7 +753,7 @@ impl CodexClient {
turn_id,
item_id,
reason,
proposed_execpolicy_amendment,
risk,
} = params;
println!(
@@ -765,12 +762,13 @@ impl CodexClient {
if let Some(reason) = reason.as_deref() {
println!("< reason: {reason}");
}
if let Some(execpolicy_amendment) = proposed_execpolicy_amendment.as_ref() {
println!("< proposed execpolicy amendment: {execpolicy_amendment:?}");
if let Some(risk) = risk.as_ref() {
println!("< risk assessment: {risk:?}");
}
let response = CommandExecutionRequestApprovalResponse {
decision: ApprovalDecision::Accept,
accept_settings: Some(CommandExecutionRequestAcceptSettings { for_session: false }),
};
self.send_server_request_response(request_id, &response)?;
println!("< approved commandExecution request for item {item_id}");

View File

@@ -26,11 +26,11 @@ codex-login = { workspace = true }
codex-protocol = { workspace = true }
codex-app-server-protocol = { workspace = true }
codex-feedback = { workspace = true }
codex-rmcp-client = { workspace = true }
codex-utils-json-to-toml = { workspace = true }
chrono = { workspace = true }
serde = { workspace = true, features = ["derive"] }
serde_json = { workspace = true }
sha2 = { workspace = true }
mcp-types = { workspace = true }
tempfile = { workspace = true }
toml = { workspace = true }
@@ -43,6 +43,7 @@ tokio = { workspace = true, features = [
] }
tracing = { workspace = true, features = ["log"] }
tracing-subscriber = { workspace = true, features = ["env-filter", "fmt"] }
opentelemetry-appender-tracing = { workspace = true }
uuid = { workspace = true, features = ["serde", "v7"] }
[dev-dependencies]

View File

@@ -3,7 +3,6 @@
`codex app-server` is the interface Codex uses to power rich interfaces such as the [Codex VS Code extension](https://marketplace.visualstudio.com/items?itemName=openai.chatgpt).
## Table of Contents
- [Protocol](#protocol)
- [Message Schema](#message-schema)
- [Core Primitives](#core-primitives)
@@ -29,7 +28,6 @@ codex app-server generate-json-schema --out DIR
## Core Primitives
The API exposes three top level primitives representing an interaction between a user and Codex:
- **Thread**: A conversation between a user and the Codex agent. Each thread contains multiple turns.
- **Turn**: One turn of the conversation, typically starting with a user message and finishing with an agent message. Each turn contains multiple items.
- **Item**: Represents user inputs and agent outputs as part of the turn, persisted and used as the context for future conversations. Example items include user message, agent reasoning, agent message, shell command, file edit, etc.
@@ -51,23 +49,13 @@ Clients must send a single `initialize` request before invoking any other method
Applications building on top of `codex app-server` should identify themselves via the `clientInfo` parameter.
Example (from OpenAI's official VSCode extension):
```json
{
"method": "initialize",
"id": 0,
"params": {
"clientInfo": {
"name": "codex-vscode",
"title": "Codex VS Code Extension",
"version": "0.1.0"
}
}
}
{ "method": "initialize", "id": 0, "params": {
"clientInfo": { "name": "codex-vscode", "title": "Codex VS Code Extension", "version": "0.1.0" }
} }
```
## API Overview
- `thread/start` — create a new thread; emits `thread/started` and auto-subscribes you to turn/item events for that thread.
- `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.
@@ -77,9 +65,6 @@ Example (from OpenAI's official VSCode extension):
- `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.
- `command/exec` — run a single command under the server sandbox without starting a thread/turn (handy for utilities and validation).
- `model/list` — list available models (with reasoning effort options).
- `skills/list` — list skills for one or more `cwd` values.
- `mcpServer/oauth/login` — start an OAuth login for a configured MCP server; returns an `authorization_url` and later emits `mcpServer/oauthLogin/completed` once the browser flow finishes.
- `mcpServerStatus/list` — enumerate configured MCP servers with their tools, resources, resource templates, and auth status; supports cursor+limit pagination.
- `feedback/upload` — submit a feedback report (classification + optional reason/logs and conversation_id); returns the tracking thread id.
- `command/exec` — run a single command under the server sandbox without starting a thread/turn (handy for utilities and validation).
- `config/read` — fetch the effective config on disk after resolving config layering.
@@ -120,7 +105,6 @@ To continue a stored session, call `thread/resume` with the `thread.id` you prev
### Example: List threads (with pagination & filters)
`thread/list` lets you render a history UI. Pass any combination of:
- `cursor` — opaque string from a prior response; omit for the first page.
- `limit` — server defaults to a reasonable page size if unset.
- `modelProviders` — restrict results to specific providers; unset, null, or an empty array will include all providers.
@@ -241,32 +225,22 @@ Codex streams the usual `turn/started` notification followed by an `item/started
with an `enteredReviewMode` item so clients can show progress:
```json
{
"method": "item/started",
"params": {
"item": {
"type": "enteredReviewMode",
"id": "turn_900",
"review": "current changes"
}
}
}
{ "method": "item/started", "params": { "item": {
"type": "enteredReviewMode",
"id": "turn_900",
"review": "current changes"
} } }
```
When the reviewer finishes, the server emits `item/started` and `item/completed`
containing an `exitedReviewMode` item with the final review text:
```json
{
"method": "item/completed",
"params": {
"item": {
"type": "exitedReviewMode",
"id": "turn_900",
"review": "Looks solid overall...\n\n- Prefer Stylize helpers — app.rs:10-20\n ..."
}
}
}
{ "method": "item/completed", "params": { "item": {
"type": "exitedReviewMode",
"id": "turn_900",
"review": "Looks solid overall...\n\n- Prefer Stylize helpers — app.rs:10-20\n ..."
} } }
```
The `review` string is plain text that already bundles the overall explanation plus a bullet list for each structured finding (matching `ThreadItem::ExitedReviewMode` in the generated schema). Use this notification to render the reviewer output in your client.
@@ -286,7 +260,6 @@ Run a standalone command (argv vector) in the servers sandbox without creatin
```
Notes:
- Empty `command` arrays are rejected.
- `sandboxPolicy` accepts the same shape used by `turn/start` (e.g., `dangerFullAccess`, `readOnly`, `workspaceWrite` with flags).
- When omitted, `timeoutMs` falls back to the server default.
@@ -309,7 +282,6 @@ Today both notifications carry an empty `items` array even when item events were
#### Items
`ThreadItem` is the tagged union carried in turn responses and `item/*` notifications. Currently we support events for the following items:
- `userMessage``{id, content}` where `content` is a list of user inputs (`text`, `image`, or `localImage`).
- `agentMessage``{id, text}` containing the accumulated agent reply.
- `reasoning``{id, summary, content}` where `summary` holds streamed reasoning summaries (applicable for most OpenAI models) and `content` holds raw reasoning blocks (applicable for e.g. open source models).
@@ -323,48 +295,37 @@ Today both notifications carry an empty `items` array even when item events were
- `compacted` - `{threadId, turnId}` when codex compacts the conversation history. This can happen automatically.
All items emit two shared lifecycle events:
- `item/started` — emits the full `item` when a new unit of work begins so the UI can render it immediately; the `item.id` in this payload matches the `itemId` used by deltas.
- `item/completed` — sends the final `item` once that work finishes (e.g., after a tool call or message completes); treat this as the authoritative state.
There are additional item-specific events:
#### agentMessage
- `item/agentMessage/delta` — appends streamed text for the agent message; concatenate `delta` values for the same `itemId` in order to reconstruct the full reply.
#### reasoning
- `item/reasoning/summaryTextDelta` — streams readable reasoning summaries; `summaryIndex` increments when a new summary section opens.
- `item/reasoning/summaryPartAdded` — marks the boundary between reasoning summary sections for an `itemId`; subsequent `summaryTextDelta` entries share the same `summaryIndex`.
- `item/reasoning/textDelta` — streams raw reasoning text (only applicable for e.g. open source models); use `contentIndex` to group deltas that belong together before showing them in the UI.
#### commandExecution
- `item/commandExecution/outputDelta` — streams stdout/stderr for the command; append deltas in order to render live output alongside `aggregatedOutput` in the final item.
Final `commandExecution` items include parsed `commandActions`, `status`, `exitCode`, and `durationMs` so the UI can summarize what ran and whether it succeeded.
Final `commandExecution` items include parsed `commandActions`, `status`, `exitCode`, and `durationMs` so the UI can summarize what ran and whether it succeeded.
#### fileChange
- `item/fileChange/outputDelta` - contains the tool call response of the underlying `apply_patch` tool call.
### 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? } }` payload as `turn.status: "failed"` and may precede that terminal notification.
`codexErrorInfo` maps to the `CodexErrorInfo` enum. Common values:
- `ContextWindowExceeded`
- `UsageLimitExceeded`
- `HttpConnectionFailed { httpStatusCode? }`: upstream HTTP failures including 4xx/5xx
- `ResponseStreamConnectionFailed { httpStatusCode? }`: failure to connect to the response SSE stream
- `ResponseStreamDisconnected { httpStatusCode? }`: disconnect of the response SSE stream in the middle of a turn before completion
- `ResponseTooManyFailedAttempts { httpStatusCode? }`
- `BadRequest`
- `Unauthorized`
- `SandboxError`
- `InternalServerError`
- `Other`: all unclassified errors
`codexErrorInfo` maps to the `CodexErrorInfo` enum. Common values:
- `ContextWindowExceeded`
- `UsageLimitExceeded`
- `HttpConnectionFailed { httpStatusCode? }`: upstream HTTP failures including 4xx/5xx
- `ResponseStreamConnectionFailed { httpStatusCode? }`: failure to connect to the response SSE stream
- `ResponseStreamDisconnected { httpStatusCode? }`: disconnect of the response SSE stream in the middle of a turn before completion
- `ResponseTooManyFailedAttempts { httpStatusCode? }`
- `BadRequest`
- `Unauthorized`
- `SandboxError`
- `InternalServerError`
- `Other`: all unclassified errors
When an upstream HTTP status is available (for example, from the Responses API or a provider), it is forwarded in `httpStatusCode` on the relevant `codexErrorInfo` variant.
@@ -378,7 +339,6 @@ Certain actions (shell commands or modifying files) may require explicit user ap
### Command execution approvals
Order of messages:
1. `item/started` — shows the pending `commandExecution` item with `command`, `cwd`, and other fields so you can render the proposed action.
2. `item/commandExecution/requestApproval` (request) — carries the same `itemId`, `threadId`, `turnId`, optionally `reason` or `risk`, plus `parsedCmd` for friendly display.
3. Client response — `{ "decision": "accept", "acceptSettings": { "forSession": false } }` or `{ "decision": "decline" }`.
@@ -387,7 +347,6 @@ Order of messages:
### File change approvals
Order of messages:
1. `item/started` — emits a `fileChange` item with `changes` (diff chunk summaries) and `status: "inProgress"`. Show the proposed edits and paths to the user.
2. `item/fileChange/requestApproval` (request) — includes `itemId`, `threadId`, `turnId`, and an optional `reason`.
3. Client response — `{ "decision": "accept" }` or `{ "decision": "decline" }`.
@@ -400,7 +359,6 @@ UI guidance for IDEs: surface an approval dialog as soon as the request arrives.
The JSON-RPC auth/account surface exposes request/response methods plus server-initiated notifications (no `id`). Use these to determine auth state, start or cancel logins, logout, and inspect ChatGPT rate limits.
### API Overview
- `account/read` — fetch current account info; optionally refresh tokens.
- `account/login/start` — begin login (`apiKey` or `chatgpt`).
- `account/login/completed` (notify) — emitted when a login attempt finishes (success or error).
@@ -408,19 +366,15 @@ The JSON-RPC auth/account surface exposes request/response methods plus server-i
- `account/logout` — sign out; triggers `account/updated`.
- `account/updated` (notify) — emitted whenever auth mode changes (`authMode`: `apikey`, `chatgpt`, or `null`).
- `account/rateLimits/read` — fetch ChatGPT rate limits; updates arrive via `account/rateLimits/updated` (notify).
- `account/rateLimits/updated` (notify) — emitted whenever a user's ChatGPT rate limits change.
- `mcpServer/oauthLogin/completed` (notify) — emitted after a `mcpServer/oauth/login` flow finishes for a server; payload includes `{ name, success, error? }`.
### 1) Check auth state
Request:
```json
{ "method": "account/read", "id": 1, "params": { "refreshToken": false } }
```
Response examples:
```json
{ "id": 1, "result": { "account": null, "requiresOpenaiAuth": false } } // No OpenAI auth needed (e.g., OSS/local models)
{ "id": 1, "result": { "account": null, "requiresOpenaiAuth": true } } // OpenAI auth required (typical for OpenAI-hosted models)
@@ -429,7 +383,6 @@ Response examples:
```
Field notes:
- `refreshToken` (bool): set `true` to force a token refresh.
- `requiresOpenaiAuth` reflects the active provider; when `false`, Codex can run without OpenAI credentials.
@@ -437,11 +390,7 @@ Field notes:
1. Send:
```json
{
"method": "account/login/start",
"id": 2,
"params": { "type": "apiKey", "apiKey": "sk-…" }
}
{ "method": "account/login/start", "id": 2, "params": { "type": "apiKey", "apiKey": "sk-…" } }
```
2. Expect:
```json
@@ -491,7 +440,6 @@ Field notes:
```
Field notes:
- `usedPercent` is current usage within the OpenAI quota window.
- `windowDurationMins` is the quota window length.
- `resetsAt` is a Unix timestamp (seconds) for the next reset.

View File

@@ -18,7 +18,6 @@ use codex_app_server_protocol::ContextCompactedNotification;
use codex_app_server_protocol::ErrorNotification;
use codex_app_server_protocol::ExecCommandApprovalParams;
use codex_app_server_protocol::ExecCommandApprovalResponse;
use codex_app_server_protocol::ExecPolicyAmendment as V2ExecPolicyAmendment;
use codex_app_server_protocol::FileChangeOutputDeltaNotification;
use codex_app_server_protocol::FileChangeRequestApprovalParams;
use codex_app_server_protocol::FileChangeRequestApprovalResponse;
@@ -31,13 +30,12 @@ use codex_app_server_protocol::McpToolCallResult;
use codex_app_server_protocol::McpToolCallStatus;
use codex_app_server_protocol::PatchApplyStatus;
use codex_app_server_protocol::PatchChangeKind as V2PatchChangeKind;
use codex_app_server_protocol::RawResponseItemCompletedNotification;
use codex_app_server_protocol::ReasoningSummaryPartAddedNotification;
use codex_app_server_protocol::ReasoningSummaryTextDeltaNotification;
use codex_app_server_protocol::ReasoningTextDeltaNotification;
use codex_app_server_protocol::SandboxCommandAssessment as V2SandboxCommandAssessment;
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::ThreadTokenUsage;
use codex_app_server_protocol::ThreadTokenUsageUpdatedNotification;
@@ -180,7 +178,8 @@ pub(crate) async fn apply_bespoke_event_handling(
command,
cwd,
reason,
proposed_execpolicy_amendment,
risk,
proposed_execpolicy_amendment: _,
parsed_cmd,
}) => match api_version {
ApiVersion::V1 => {
@@ -190,6 +189,7 @@ pub(crate) async fn apply_bespoke_event_handling(
command,
cwd,
reason,
risk,
parsed_cmd,
};
let rx = outgoing
@@ -207,8 +207,6 @@ pub(crate) async fn apply_bespoke_event_handling(
.map(V2ParsedCommand::from)
.collect::<Vec<_>>();
let command_string = shlex_join(&command);
let proposed_execpolicy_amendment_v2 =
proposed_execpolicy_amendment.map(V2ExecPolicyAmendment::from);
let params = CommandExecutionRequestApprovalParams {
thread_id: conversation_id.to_string(),
@@ -217,7 +215,7 @@ pub(crate) async fn apply_bespoke_event_handling(
// and emit the corresponding EventMsg, we repurpose the call_id as the item_id.
item_id: item_id.clone(),
reason,
proposed_execpolicy_amendment: proposed_execpolicy_amendment_v2,
risk: risk.map(V2SandboxCommandAssessment::from),
};
let rx = outgoing
.send_request(ServerRequestPayload::CommandExecutionRequestApproval(
@@ -452,16 +450,6 @@ pub(crate) async fn apply_bespoke_event_handling(
.send_server_notification(ServerNotification::ItemCompleted(completed))
.await;
}
EventMsg::RawResponseItem(raw_response_item_event) => {
maybe_emit_raw_response_item_completed(
api_version,
conversation_id,
&event_turn_id,
raw_response_item_event.item,
outgoing.as_ref(),
)
.await;
}
EventMsg::PatchApplyBegin(patch_begin_event) => {
// Until we migrate the core to be aware of a first class FileChangeItem
// and emit the corresponding EventMsg, we repurpose the call_id as the item_id.
@@ -581,20 +569,6 @@ pub(crate) async fn apply_bespoke_event_handling(
.await;
}
}
EventMsg::TerminalInteraction(terminal_event) => {
let item_id = terminal_event.call_id.clone();
let notification = TerminalInteractionNotification {
thread_id: conversation_id.to_string(),
turn_id: event_turn_id.clone(),
item_id,
process_id: terminal_event.process_id,
stdin: terminal_event.stdin,
};
outgoing
.send_server_notification(ServerNotification::TerminalInteraction(notification))
.await;
}
EventMsg::ExecCommandEnd(exec_command_end_event) => {
let ExecCommandEndEvent {
call_id,
@@ -831,27 +805,6 @@ async fn complete_command_execution_item(
.await;
}
async fn maybe_emit_raw_response_item_completed(
api_version: ApiVersion,
conversation_id: ConversationId,
turn_id: &str,
item: codex_protocol::models::ResponseItem,
outgoing: &OutgoingMessageSender,
) {
let ApiVersion::V2 = api_version else {
return;
};
let notification = RawResponseItemCompletedNotification {
thread_id: conversation_id.to_string(),
turn_id: turn_id.to_string(),
item,
};
outgoing
.send_server_notification(ServerNotification::RawResponseItemCompleted(notification))
.await;
}
async fn find_and_remove_turn_summary(
conversation_id: ConversationId,
turn_summary_store: &TurnSummaryStore,
@@ -1094,11 +1047,7 @@ async fn on_file_change_request_approval_response(
});
let (decision, completion_status) = match response.decision {
ApprovalDecision::Accept
| ApprovalDecision::AcceptForSession
| ApprovalDecision::AcceptWithExecpolicyAmendment { .. } => {
(ReviewDecision::Approved, None)
}
ApprovalDecision::Accept => (ReviewDecision::Approved, None),
ApprovalDecision::Decline => {
(ReviewDecision::Denied, Some(PatchApplyStatus::Declined))
}
@@ -1160,27 +1109,25 @@ async fn on_command_execution_request_approval_response(
error!("failed to deserialize CommandExecutionRequestApprovalResponse: {err}");
CommandExecutionRequestApprovalResponse {
decision: ApprovalDecision::Decline,
accept_settings: None,
}
});
let decision = response.decision;
let CommandExecutionRequestApprovalResponse {
decision,
accept_settings,
} = response;
let (decision, completion_status) = match decision {
ApprovalDecision::Accept => (ReviewDecision::Approved, None),
ApprovalDecision::AcceptForSession => (ReviewDecision::ApprovedForSession, None),
ApprovalDecision::AcceptWithExecpolicyAmendment {
execpolicy_amendment,
} => (
ReviewDecision::ApprovedExecpolicyAmendment {
proposed_execpolicy_amendment: execpolicy_amendment.into_core(),
},
None,
),
ApprovalDecision::Decline => (
let (decision, completion_status) = match (decision, accept_settings) {
(ApprovalDecision::Accept, Some(settings)) if settings.for_session => {
(ReviewDecision::ApprovedForSession, None)
}
(ApprovalDecision::Accept, _) => (ReviewDecision::Approved, None),
(ApprovalDecision::Decline, _) => (
ReviewDecision::Denied,
Some(CommandExecutionStatus::Declined),
),
ApprovalDecision::Cancel => (
(ApprovalDecision::Cancel, _) => (
ReviewDecision::Abort,
Some(CommandExecutionStatus::Declined),
),
@@ -1242,7 +1189,7 @@ async fn construct_mcp_tool_call_notification(
}
}
/// similar to handle_mcp_tool_call_end in exec
/// simiilar to handle_mcp_tool_call_end in exec
async fn construct_mcp_tool_call_end_notification(
end_event: McpToolCallEndEvent,
thread_id: String,

View File

@@ -19,7 +19,6 @@ use codex_app_server_protocol::AuthMode;
use codex_app_server_protocol::AuthStatusChangeNotification;
use codex_app_server_protocol::CancelLoginAccountParams;
use codex_app_server_protocol::CancelLoginAccountResponse;
use codex_app_server_protocol::CancelLoginAccountStatus;
use codex_app_server_protocol::CancelLoginChatGptResponse;
use codex_app_server_protocol::ClientRequest;
use codex_app_server_protocol::CommandExecParams;
@@ -46,8 +45,8 @@ use codex_app_server_protocol::InterruptConversationParams;
use codex_app_server_protocol::JSONRPCErrorError;
use codex_app_server_protocol::ListConversationsParams;
use codex_app_server_protocol::ListConversationsResponse;
use codex_app_server_protocol::ListMcpServerStatusParams;
use codex_app_server_protocol::ListMcpServerStatusResponse;
use codex_app_server_protocol::ListMcpServersParams;
use codex_app_server_protocol::ListMcpServersResponse;
use codex_app_server_protocol::LoginAccountParams;
use codex_app_server_protocol::LoginApiKeyParams;
use codex_app_server_protocol::LoginApiKeyResponse;
@@ -55,10 +54,7 @@ use codex_app_server_protocol::LoginChatGptCompleteNotification;
use codex_app_server_protocol::LoginChatGptResponse;
use codex_app_server_protocol::LogoutAccountResponse;
use codex_app_server_protocol::LogoutChatGptResponse;
use codex_app_server_protocol::McpServerOauthLoginCompletedNotification;
use codex_app_server_protocol::McpServerOauthLoginParams;
use codex_app_server_protocol::McpServerOauthLoginResponse;
use codex_app_server_protocol::McpServerStatus;
use codex_app_server_protocol::McpServer;
use codex_app_server_protocol::ModelListParams;
use codex_app_server_protocol::ModelListResponse;
use codex_app_server_protocol::NewConversationParams;
@@ -81,8 +77,6 @@ use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::SessionConfiguredNotification;
use codex_app_server_protocol::SetDefaultModelParams;
use codex_app_server_protocol::SetDefaultModelResponse;
use codex_app_server_protocol::SkillsListParams;
use codex_app_server_protocol::SkillsListResponse;
use codex_app_server_protocol::Thread;
use codex_app_server_protocol::ThreadArchiveParams;
use codex_app_server_protocol::ThreadArchiveResponse;
@@ -119,9 +113,9 @@ use codex_core::auth::CLIENT_ID;
use codex_core::auth::login_with_api_key;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use codex_core::config::ConfigService;
use codex_core::config::ConfigToml;
use codex_core::config::edit::ConfigEditsBuilder;
use codex_core::config::types::McpServerTransportConfig;
use codex_core::config_loader::load_config_as_toml;
use codex_core::default_client::get_codex_user_agent;
use codex_core::exec::ExecParams;
use codex_core::exec_env::create_env;
@@ -138,7 +132,6 @@ use codex_core::protocol::ReviewRequest;
use codex_core::protocol::ReviewTarget as CoreReviewTarget;
use codex_core::protocol::SessionConfiguredEvent;
use codex_core::read_head_for_summary;
use codex_core::sandboxing::SandboxPermissions;
use codex_feedback::CodexFeedback;
use codex_login::ServerOptions as LoginServerOptions;
use codex_login::ShutdownHandle;
@@ -154,7 +147,6 @@ use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::SessionMetaLine;
use codex_protocol::protocol::USER_MESSAGE_BEGIN;
use codex_protocol::user_input::UserInput as CoreInputItem;
use codex_rmcp_client::perform_oauth_login_return_url;
use codex_utils_json_to_toml::json_to_toml;
use std::collections::HashMap;
use std::collections::HashSet;
@@ -169,7 +161,6 @@ use std::time::Duration;
use tokio::select;
use tokio::sync::Mutex;
use tokio::sync::oneshot;
use toml::Value as TomlValue;
use tracing::error;
use tracing::info;
use tracing::warn;
@@ -187,9 +178,6 @@ pub(crate) struct TurnSummary {
pub(crate) type TurnSummaryStore = Arc<Mutex<HashMap<ConversationId, TurnSummary>>>;
const THREAD_LIST_DEFAULT_LIMIT: usize = 25;
const THREAD_LIST_MAX_LIMIT: usize = 100;
// Duration before a ChatGPT login attempt is abandoned.
const LOGIN_CHATGPT_TIMEOUT: Duration = Duration::from_secs(10 * 60);
struct ActiveLogin {
@@ -197,11 +185,6 @@ struct ActiveLogin {
login_id: Uuid,
}
#[derive(Clone, Copy, Debug)]
enum CancelLoginError {
NotFound(Uuid),
}
impl Drop for ActiveLogin {
fn drop(&mut self) {
self.shutdown_handle.shutdown();
@@ -215,7 +198,6 @@ pub(crate) struct CodexMessageProcessor {
outgoing: Arc<OutgoingMessageSender>,
codex_linux_sandbox_exe: Option<PathBuf>,
config: Arc<Config>,
cli_overrides: Vec<(String, TomlValue)>,
conversation_listeners: HashMap<Uuid, oneshot::Sender<()>>,
active_login: Arc<Mutex<Option<ActiveLogin>>>,
// Queue of pending interrupt requests per conversation. We reply when TurnAborted arrives.
@@ -262,7 +244,6 @@ impl CodexMessageProcessor {
outgoing: Arc<OutgoingMessageSender>,
codex_linux_sandbox_exe: Option<PathBuf>,
config: Arc<Config>,
cli_overrides: Vec<(String, TomlValue)>,
feedback: CodexFeedback,
) -> Self {
Self {
@@ -271,7 +252,6 @@ impl CodexMessageProcessor {
outgoing,
codex_linux_sandbox_exe,
config,
cli_overrides,
conversation_listeners: HashMap::new(),
active_login: Arc::new(Mutex::new(None)),
pending_interrupts: Arc::new(Mutex::new(HashMap::new())),
@@ -281,16 +261,6 @@ impl CodexMessageProcessor {
}
}
async fn load_latest_config(&self) -> Result<Config, JSONRPCErrorError> {
Config::load_with_cli_overrides(self.cli_overrides.clone(), ConfigOverrides::default())
.await
.map_err(|err| JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to reload config: {err}"),
data: None,
})
}
fn review_request_from_target(
target: ApiReviewTarget,
) -> Result<(ReviewRequest, String), JSONRPCErrorError> {
@@ -368,8 +338,12 @@ impl CodexMessageProcessor {
ClientRequest::ThreadList { request_id, params } => {
self.thread_list(request_id, params).await;
}
ClientRequest::SkillsList { request_id, params } => {
self.skills_list(request_id, params).await;
ClientRequest::ThreadCompact {
request_id,
params: _,
} => {
self.send_unimplemented_error(request_id, "thread/compact")
.await;
}
ClientRequest::TurnStart { request_id, params } => {
self.turn_start(request_id, params).await;
@@ -395,11 +369,8 @@ impl CodexMessageProcessor {
ClientRequest::ModelList { request_id, params } => {
self.list_models(request_id, params).await;
}
ClientRequest::McpServerOauthLogin { request_id, params } => {
self.mcp_server_oauth_login(request_id, params).await;
}
ClientRequest::McpServerStatusList { request_id, params } => {
self.list_mcp_server_status(request_id, params).await;
ClientRequest::McpServersList { request_id, params } => {
self.list_mcp_servers(request_id, params).await;
}
ClientRequest::LoginAccount { request_id, params } => {
self.login_v2(request_id, params).await;
@@ -508,6 +479,15 @@ impl CodexMessageProcessor {
}
}
async fn send_unimplemented_error(&self, request_id: RequestId, method: &str) {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("{method} is not implemented yet"),
data: None,
};
self.outgoing.send_error(request_id, error).await;
}
async fn login_v2(&mut self, request_id: RequestId, params: LoginAccountParams) {
match params {
LoginAccountParams::ApiKey { api_key } => {
@@ -822,7 +802,7 @@ impl CodexMessageProcessor {
async fn cancel_login_chatgpt_common(
&mut self,
login_id: Uuid,
) -> std::result::Result<(), CancelLoginError> {
) -> std::result::Result<(), JSONRPCErrorError> {
let mut guard = self.active_login.lock().await;
if guard.as_ref().map(|l| l.login_id) == Some(login_id) {
if let Some(active) = guard.take() {
@@ -830,7 +810,11 @@ impl CodexMessageProcessor {
}
Ok(())
} else {
Err(CancelLoginError::NotFound(login_id))
Err(JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: format!("login id not found: {login_id}"),
data: None,
})
}
}
@@ -841,12 +825,7 @@ impl CodexMessageProcessor {
.send_response(request_id, CancelLoginChatGptResponse {})
.await;
}
Err(CancelLoginError::NotFound(missing_login_id)) => {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: format!("login id not found: {missing_login_id}"),
data: None,
};
Err(error) => {
self.outgoing.send_error(request_id, error).await;
}
}
@@ -855,14 +834,16 @@ impl CodexMessageProcessor {
async fn cancel_login_v2(&mut self, request_id: RequestId, params: CancelLoginAccountParams) {
let login_id = params.login_id;
match Uuid::parse_str(&login_id) {
Ok(uuid) => {
let status = match self.cancel_login_chatgpt_common(uuid).await {
Ok(()) => CancelLoginAccountStatus::Canceled,
Err(CancelLoginError::NotFound(_)) => CancelLoginAccountStatus::NotFound,
};
let response = CancelLoginAccountResponse { status };
self.outgoing.send_response(request_id, response).await;
}
Ok(uuid) => match self.cancel_login_chatgpt_common(uuid).await {
Ok(()) => {
self.outgoing
.send_response(request_id, CancelLoginAccountResponse {})
.await;
}
Err(error) => {
self.outgoing.send_error(request_id, error).await;
}
},
Err(_) => {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
@@ -1096,13 +1077,12 @@ impl CodexMessageProcessor {
}
async fn get_user_saved_config(&self, request_id: RequestId) {
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,
let toml_value = match load_config_as_toml(&self.config.codex_home).await {
Ok(val) => val,
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: err.to_string(),
message: format!("failed to load config.toml: {err}"),
data: None,
};
self.outgoing.send_error(request_id, error).await;
@@ -1110,6 +1090,21 @@ impl CodexMessageProcessor {
}
};
let cfg: ConfigToml = match toml_value.try_into() {
Ok(cfg) => cfg,
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to parse config.toml: {err}"),
data: None,
};
self.outgoing.send_error(request_id, error).await;
return;
}
};
let user_saved_config: UserSavedConfig = cfg.into();
let response = GetUserSavedConfigResponse {
config: user_saved_config,
};
@@ -1174,7 +1169,7 @@ impl CodexMessageProcessor {
cwd,
expiration: timeout_ms.into(),
env,
sandbox_permissions: SandboxPermissions::UseDefault,
with_escalated_permissions: None,
justification: None,
arg0: None,
};
@@ -1254,7 +1249,7 @@ impl CodexMessageProcessor {
let mut cli_overrides = cli_overrides.unwrap_or_default();
if cfg!(windows) && self.config.features.enabled(Feature::WindowsSandbox) {
cli_overrides.insert(
"features.experimental_windows_sandbox".to_string(),
"features.enable_experimental_windows_sandbox".to_string(),
serde_json::json!(true),
);
}
@@ -1373,13 +1368,9 @@ impl CodexMessageProcessor {
};
// Auto-attach a conversation listener when starting a thread.
// Use the same behavior as the v1 API, with opt-in support for raw item events.
// Use the same behavior as the v1 API with experimental_raw_events=false.
if let Err(err) = self
.attach_conversation_listener(
conversation_id,
params.experimental_raw_events,
ApiVersion::V2,
)
.attach_conversation_listener(conversation_id, false, ApiVersion::V2)
.await
{
tracing::warn!(
@@ -1494,12 +1485,10 @@ impl CodexMessageProcessor {
model_providers,
} = params;
let requested_page_size = limit
.map(|value| value as usize)
.unwrap_or(THREAD_LIST_DEFAULT_LIMIT)
.clamp(1, THREAD_LIST_MAX_LIMIT);
let page_size = limit.unwrap_or(25).max(1) as usize;
let (summaries, next_cursor) = match self
.list_conversations_common(requested_page_size, cursor, model_providers)
.list_conversations_common(page_size, cursor, model_providers)
.await
{
Ok(r) => r,
@@ -1510,6 +1499,7 @@ impl CodexMessageProcessor {
};
let data = summaries.into_iter().map(summary_to_thread).collect();
let response = ThreadListResponse { data, next_cursor };
self.outgoing.send_response(request_id, response).await;
}
@@ -1787,12 +1777,10 @@ impl CodexMessageProcessor {
cursor,
model_providers,
} = params;
let requested_page_size = page_size
.unwrap_or(THREAD_LIST_DEFAULT_LIMIT)
.clamp(1, THREAD_LIST_MAX_LIMIT);
let page_size = page_size.unwrap_or(25).max(1);
match self
.list_conversations_common(requested_page_size, cursor, model_providers)
.list_conversations_common(page_size, cursor, model_providers)
.await
{
Ok((items, next_cursor)) => {
@@ -1807,15 +1795,12 @@ impl CodexMessageProcessor {
async fn list_conversations_common(
&self,
requested_page_size: usize,
page_size: usize,
cursor: Option<String>,
model_providers: Option<Vec<String>>,
) -> Result<(Vec<ConversationSummary>, Option<String>), JSONRPCErrorError> {
let mut cursor_obj: Option<RolloutCursor> = cursor.as_ref().and_then(|s| parse_cursor(s));
let mut last_cursor = cursor_obj.clone();
let mut remaining = requested_page_size;
let mut items = Vec::with_capacity(requested_page_size);
let mut next_cursor: Option<String> = None;
let cursor_obj: Option<RolloutCursor> = cursor.as_ref().and_then(|s| parse_cursor(s));
let cursor_ref = cursor_obj.as_ref();
let model_provider_filter = match model_providers {
Some(providers) => {
@@ -1829,76 +1814,55 @@ impl CodexMessageProcessor {
};
let fallback_provider = self.config.model_provider_id.clone();
while remaining > 0 {
let page_size = remaining.min(THREAD_LIST_MAX_LIMIT);
let page = RolloutRecorder::list_conversations(
&self.config.codex_home,
page_size,
cursor_obj.as_ref(),
INTERACTIVE_SESSION_SOURCES,
model_provider_filter.as_deref(),
fallback_provider.as_str(),
)
.await
.map_err(|err| JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to list conversations: {err}"),
data: None,
})?;
let mut filtered = page
.items
.into_iter()
.filter_map(|it| {
let session_meta_line = it.head.first().and_then(|first| {
serde_json::from_value::<SessionMetaLine>(first.clone()).ok()
})?;
extract_conversation_summary(
it.path,
&it.head,
&session_meta_line.meta,
session_meta_line.git.as_ref(),
fallback_provider.as_str(),
)
})
.collect::<Vec<_>>();
if filtered.len() > remaining {
filtered.truncate(remaining);
let page = match RolloutRecorder::list_conversations(
&self.config.codex_home,
page_size,
cursor_ref,
INTERACTIVE_SESSION_SOURCES,
model_provider_filter.as_deref(),
fallback_provider.as_str(),
)
.await
{
Ok(p) => p,
Err(err) => {
return Err(JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to list conversations: {err}"),
data: None,
});
}
items.extend(filtered);
remaining = requested_page_size.saturating_sub(items.len());
};
// Encode RolloutCursor into the JSON-RPC string form returned to clients.
let next_cursor_value = page.next_cursor.clone();
next_cursor = next_cursor_value
.as_ref()
.and_then(|cursor| serde_json::to_value(cursor).ok())
.and_then(|value| value.as_str().map(str::to_owned));
if remaining == 0 {
break;
}
let items = page
.items
.into_iter()
.filter_map(|it| {
let session_meta_line = it.head.first().and_then(|first| {
serde_json::from_value::<SessionMetaLine>(first.clone()).ok()
})?;
extract_conversation_summary(
it.path,
&it.head,
&session_meta_line.meta,
session_meta_line.git.as_ref(),
fallback_provider.as_str(),
)
})
.collect::<Vec<_>>();
match next_cursor_value {
Some(cursor_val) if remaining > 0 => {
// Break if our pagination would reuse the same cursor again; this avoids
// an infinite loop when filtering drops everything on the page.
if last_cursor.as_ref() == Some(&cursor_val) {
next_cursor = None;
break;
}
last_cursor = Some(cursor_val.clone());
cursor_obj = Some(cursor_val);
}
_ => break,
}
}
// Encode next_cursor as a plain string
let next_cursor = page
.next_cursor
.and_then(|cursor| serde_json::to_value(&cursor).ok())
.and_then(|value| value.as_str().map(str::to_owned));
Ok((items, next_cursor))
}
async fn list_models(&self, request_id: RequestId, params: ModelListParams) {
let ModelListParams { limit, cursor } = params;
let models = supported_models(self.conversation_manager.clone(), &self.config).await;
let models = supported_models(self.conversation_manager.clone()).await;
let total = models.len();
if total == 0 {
@@ -1952,128 +1916,13 @@ impl CodexMessageProcessor {
self.outgoing.send_response(request_id, response).await;
}
async fn mcp_server_oauth_login(
&self,
request_id: RequestId,
params: McpServerOauthLoginParams,
) {
let config = match self.load_latest_config().await {
Ok(config) => config,
Err(error) => {
self.outgoing.send_error(request_id, error).await;
return;
}
};
if !config.features.enabled(Feature::RmcpClient) {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: "OAuth login is only supported when [features].rmcp_client is true in config.toml".to_string(),
data: None,
};
self.outgoing.send_error(request_id, error).await;
return;
}
let McpServerOauthLoginParams {
name,
scopes,
timeout_secs,
} = params;
let Some(server) = config.mcp_servers.get(&name) else {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: format!("No MCP server named '{name}' found."),
data: None,
};
self.outgoing.send_error(request_id, error).await;
return;
};
let (url, http_headers, env_http_headers) = match &server.transport {
McpServerTransportConfig::StreamableHttp {
url,
http_headers,
env_http_headers,
..
} => (url.clone(), http_headers.clone(), env_http_headers.clone()),
_ => {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
message: "OAuth login is only supported for streamable HTTP servers."
.to_string(),
data: None,
};
self.outgoing.send_error(request_id, error).await;
return;
}
};
match perform_oauth_login_return_url(
&name,
&url,
config.mcp_oauth_credentials_store_mode,
http_headers,
env_http_headers,
scopes.as_deref().unwrap_or_default(),
timeout_secs,
)
.await
{
Ok(handle) => {
let authorization_url = handle.authorization_url().to_string();
let notification_name = name.clone();
let outgoing = Arc::clone(&self.outgoing);
tokio::spawn(async move {
let (success, error) = match handle.wait().await {
Ok(()) => (true, None),
Err(err) => (false, Some(err.to_string())),
};
let notification = ServerNotification::McpServerOauthLoginCompleted(
McpServerOauthLoginCompletedNotification {
name: notification_name,
success,
error,
},
);
outgoing.send_server_notification(notification).await;
});
let response = McpServerOauthLoginResponse { authorization_url };
self.outgoing.send_response(request_id, response).await;
}
Err(err) => {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to login to MCP server '{name}': {err}"),
data: None,
};
self.outgoing.send_error(request_id, error).await;
}
}
}
async fn list_mcp_server_status(
&self,
request_id: RequestId,
params: ListMcpServerStatusParams,
) {
let config = match self.load_latest_config().await {
Ok(config) => config,
Err(error) => {
self.outgoing.send_error(request_id, error).await;
return;
}
};
let snapshot = collect_mcp_snapshot(&config).await;
async fn list_mcp_servers(&self, request_id: RequestId, params: ListMcpServersParams) {
let snapshot = collect_mcp_snapshot(self.config.as_ref()).await;
let tools_by_server = group_tools_by_server(&snapshot.tools);
let mut server_names: Vec<String> = config
let mut server_names: Vec<String> = self
.config
.mcp_servers
.keys()
.cloned()
@@ -2115,9 +1964,9 @@ impl CodexMessageProcessor {
let end = start.saturating_add(effective_limit).min(total);
let data: Vec<McpServerStatus> = server_names[start..end]
let data: Vec<McpServer> = server_names[start..end]
.iter()
.map(|name| McpServerStatus {
.map(|name| McpServer {
name: name.clone(),
tools: tools_by_server.get(name).cloned().unwrap_or_default(),
resources: snapshot.resources.get(name).cloned().unwrap_or_default(),
@@ -2141,7 +1990,7 @@ impl CodexMessageProcessor {
None
};
let response = ListMcpServerStatusResponse { data, next_cursor };
let response = ListMcpServersResponse { data, next_cursor };
self.outgoing.send_response(request_id, response).await;
}
@@ -2179,7 +2028,7 @@ impl CodexMessageProcessor {
let mut cli_overrides = cli_overrides.unwrap_or_default();
if cfg!(windows) && self.config.features.enabled(Feature::WindowsSandbox) {
cli_overrides.insert(
"features.experimental_windows_sandbox".to_string(),
"features.enable_experimental_windows_sandbox".to_string(),
serde_json::json!(true),
);
}
@@ -2612,42 +2461,6 @@ impl CodexMessageProcessor {
.await;
}
async fn skills_list(&self, request_id: RequestId, params: SkillsListParams) {
let SkillsListParams { cwds } = params;
let cwds = if cwds.is_empty() {
vec![self.config.cwd.clone()]
} else {
cwds
};
let data = if self.config.features.enabled(Feature::Skills) {
let skills_manager = self.conversation_manager.skills_manager();
cwds.into_iter()
.map(|cwd| {
let outcome = skills_manager.skills_for_cwd(&cwd);
let errors = errors_to_info(&outcome.errors);
let skills = skills_to_info(&outcome.skills);
codex_app_server_protocol::SkillsListEntry {
cwd,
skills,
errors,
}
})
.collect()
} else {
cwds.into_iter()
.map(|cwd| codex_app_server_protocol::SkillsListEntry {
cwd,
skills: Vec::new(),
errors: Vec::new(),
})
.collect()
};
self.outgoing
.send_response(request_id, SkillsListResponse { data })
.await;
}
async fn interrupt_conversation(
&mut self,
request_id: RequestId,
@@ -2856,7 +2669,7 @@ impl CodexMessageProcessor {
})?;
let mut config = self.config.as_ref().clone();
config.model = Some(self.config.review_model.clone());
config.model = self.config.review_model.clone();
let NewConversation {
conversation_id,
@@ -3293,32 +3106,6 @@ impl CodexMessageProcessor {
}
}
fn skills_to_info(
skills: &[codex_core::skills::SkillMetadata],
) -> Vec<codex_app_server_protocol::SkillMetadata> {
skills
.iter()
.map(|skill| codex_app_server_protocol::SkillMetadata {
name: skill.name.clone(),
description: skill.description.clone(),
path: skill.path.clone(),
scope: skill.scope.into(),
})
.collect()
}
fn errors_to_info(
errors: &[codex_core::skills::SkillError],
) -> Vec<codex_app_server_protocol::SkillErrorInfo> {
errors
.iter()
.map(|err| codex_app_server_protocol::SkillErrorInfo {
path: err.path.clone(),
message: err.message.clone(),
})
.collect()
}
async fn derive_config_from_params(
overrides: ConfigOverrides,
cli_overrides: Option<std::collections::HashMap<String, serde_json::Value>>,

File diff suppressed because it is too large Load Diff

View File

@@ -3,6 +3,7 @@
use codex_common::CliConfigOverrides;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use opentelemetry_appender_tracing::layer::OpenTelemetryTracingBridge;
use std::io::ErrorKind;
use std::io::Result as IoResult;
use std::path::PathBuf;
@@ -102,7 +103,6 @@ pub async fn run_main(
// control the log level with `RUST_LOG`.
let stderr_fmt = tracing_subscriber::fmt::layer()
.with_writer(std::io::stderr)
.with_span_events(tracing_subscriber::fmt::format::FmtSpan::FULL)
.with_filter(EnvFilter::from_default_env());
let feedback_layer = tracing_subscriber::fmt::layer()
@@ -111,15 +111,14 @@ pub async fn run_main(
.with_target(false)
.with_filter(Targets::new().with_default(Level::TRACE));
let otel_logger_layer = otel.as_ref().and_then(|o| o.logger_layer());
let otel_tracing_layer = otel.as_ref().and_then(|o| o.tracing_layer());
let _ = tracing_subscriber::registry()
.with(stderr_fmt)
.with(feedback_layer)
.with(otel_logger_layer)
.with(otel_tracing_layer)
.with(otel.as_ref().map(|provider| {
OpenTelemetryTracingBridge::new(&provider.logger).with_filter(
tracing_subscriber::filter::filter_fn(codex_core::otel_init::codex_export_filter),
)
}))
.try_init();
// Task: process incoming messages.

View File

@@ -59,7 +59,6 @@ impl MessageProcessor {
outgoing.clone(),
codex_linux_sandbox_exe,
Arc::clone(&config),
cli_overrides.clone(),
feedback,
);
let config_api = ConfigApi::new(config.codex_home.clone(), cli_overrides);

View File

@@ -3,16 +3,12 @@ use std::sync::Arc;
use codex_app_server_protocol::Model;
use codex_app_server_protocol::ReasoningEffortOption;
use codex_core::ConversationManager;
use codex_core::config::Config;
use codex_protocol::openai_models::ModelPreset;
use codex_protocol::openai_models::ReasoningEffortPreset;
pub async fn supported_models(
conversation_manager: Arc<ConversationManager>,
config: &Config,
) -> Vec<Model> {
pub async fn supported_models(conversation_manager: Arc<ConversationManager>) -> Vec<Model> {
conversation_manager
.list_models(config)
.list_models()
.await
.into_iter()
.map(model_from_preset)

View File

@@ -13,7 +13,7 @@ assert_cmd = { workspace = true }
base64 = { workspace = true }
chrono = { workspace = true }
codex-app-server-protocol = { workspace = true }
codex-core = { workspace = true, features = ["test-support"] }
codex-core = { workspace = true }
codex-protocol = { workspace = true }
serde = { workspace = true }
serde_json = { workspace = true }

View File

@@ -1,7 +1,6 @@
mod auth_fixtures;
mod mcp_process;
mod mock_model_server;
mod models_cache;
mod responses;
mod rollout;
@@ -12,16 +11,9 @@ pub use auth_fixtures::write_chatgpt_auth;
use codex_app_server_protocol::JSONRPCResponse;
pub use core_test_support::format_with_current_shell;
pub use core_test_support::format_with_current_shell_display;
pub use core_test_support::format_with_current_shell_display_non_login;
pub use core_test_support::format_with_current_shell_non_login;
pub use core_test_support::test_path_buf_with_windows;
pub use core_test_support::test_tmp_path;
pub use core_test_support::test_tmp_path_buf;
pub use mcp_process::McpProcess;
pub use mock_model_server::create_mock_chat_completions_server;
pub use mock_model_server::create_mock_chat_completions_server_unchecked;
pub use models_cache::write_models_cache;
pub use models_cache::write_models_cache_with_models;
pub use responses::create_apply_patch_sse_response;
pub use responses::create_exec_command_sse_response;
pub use responses::create_final_assistant_message_sse_response;

View File

@@ -1,85 +0,0 @@
use chrono::DateTime;
use chrono::Utc;
use codex_core::openai_models::model_presets::all_model_presets;
use codex_protocol::openai_models::ClientVersion;
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;
/// Convert a ModelPreset to ModelInfo for cache storage.
fn preset_to_info(preset: &ModelPreset, priority: i32) -> ModelInfo {
ModelInfo {
slug: preset.id.clone(),
display_name: preset.display_name.clone(),
description: Some(preset.description.clone()),
default_reasoning_level: preset.default_reasoning_effort,
supported_reasoning_levels: preset.supported_reasoning_efforts.clone(),
shell_type: ConfigShellToolType::ShellCommand,
visibility: if preset.show_in_picker {
ModelVisibility::List
} else {
ModelVisibility::Hide
},
minimal_client_version: ClientVersion(0, 1, 0),
supported_in_api: true,
priority,
upgrade: preset.upgrade.as_ref().map(|u| u.id.clone()),
base_instructions: None,
supports_reasoning_summaries: false,
support_verbosity: false,
default_verbosity: None,
apply_patch_tool_type: None,
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
context_window: None,
reasoning_summary_format: ReasoningSummaryFormat::None,
experimental_supported_tools: Vec::new(),
}
}
/// Write a models_cache.json file to the codex home directory.
/// This prevents ModelsManager from making network requests to refresh models.
/// The cache will be treated as fresh (within TTL) and used instead of fetching from the network.
/// Uses the built-in model presets from ModelsManager, converted to ModelInfo format.
pub fn write_models_cache(codex_home: &Path) -> std::io::Result<()> {
// Get all presets and filter for show_in_picker (same as builtin_model_presets does)
let presets: Vec<&ModelPreset> = all_model_presets()
.iter()
.filter(|preset| preset.show_in_picker)
.collect();
// Convert presets to ModelInfo, assigning priorities (higher = earlier in list)
// Priority is used for sorting, so first model gets highest priority
let models: Vec<ModelInfo> = presets
.iter()
.enumerate()
.map(|(idx, preset)| {
// Higher priority = earlier in list, so reverse the index
let priority = (presets.len() - idx) as i32;
preset_to_info(preset, priority)
})
.collect();
write_models_cache_with_models(codex_home, models)
}
/// Write a models_cache.json file with specific models.
/// Useful when tests need specific models to be available.
pub fn write_models_cache_with_models(
codex_home: &Path,
models: Vec<ModelInfo>,
) -> std::io::Result<()> {
let cache_path = codex_home.join("models_cache.json");
// DateTime<Utc> serializes to RFC3339 format by default with serde
let fetched_at: DateTime<Utc> = Utc::now();
let cache = json!({
"fetched_at": fetched_at,
"etag": null,
"models": models
});
std::fs::write(cache_path, serde_json::to_string_pretty(&cache)?)
}

View File

@@ -271,6 +271,7 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
command: format_with_current_shell("python3 -c 'print(42)'"),
cwd: working_directory.clone(),
reason: None,
risk: None,
parsed_cmd: vec![ParsedCommand::Unknown {
cmd: "python3 -c 'print(42)'".to_string()
}],
@@ -410,7 +411,7 @@ async fn test_send_user_turn_updates_sandbox_and_cwd_between_turns() -> Result<(
cwd: first_cwd.clone(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::WorkspaceWrite {
writable_roots: vec![first_cwd.try_into()?],
writable_roots: vec![first_cwd.clone()],
network_access: false,
exclude_tmpdir_env_var: false,
exclude_slash_tmp: false,

View File

@@ -1,6 +1,5 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::test_tmp_path;
use app_test_support::to_response;
use codex_app_server_protocol::GetUserSavedConfigResponse;
use codex_app_server_protocol::JSONRPCResponse;
@@ -24,12 +23,10 @@ use tokio::time::timeout;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
fn create_config_toml(codex_home: &Path) -> std::io::Result<()> {
let writable_root = test_tmp_path();
let config_toml = codex_home.join("config.toml");
std::fs::write(
config_toml,
format!(
r#"
r#"
model = "gpt-5.1-codex-max"
approval_policy = "on-request"
sandbox_mode = "workspace-write"
@@ -41,7 +38,7 @@ forced_chatgpt_workspace_id = "12345678-0000-0000-0000-000000000000"
forced_login_method = "chatgpt"
[sandbox_workspace_write]
writable_roots = [{}]
writable_roots = ["/tmp"]
network_access = true
exclude_tmpdir_env_var = true
exclude_slash_tmp = true
@@ -59,8 +56,6 @@ model_verbosity = "medium"
model_provider = "openai"
chatgpt_base_url = "https://api.chatgpt.com"
"#,
serde_json::json!(writable_root)
),
)
}
@@ -80,13 +75,12 @@ async fn get_config_toml_parses_all_fields() -> Result<()> {
.await??;
let config: GetUserSavedConfigResponse = to_response(resp)?;
let writable_root = test_tmp_path();
let expected = GetUserSavedConfigResponse {
config: UserSavedConfig {
approval_policy: Some(AskForApproval::OnRequest),
sandbox_mode: Some(SandboxMode::WorkspaceWrite),
sandbox_settings: Some(SandboxSettings {
writable_roots: vec![writable_root],
writable_roots: vec!["/tmp".into()],
network_access: Some(true),
exclude_tmpdir_env_var: Some(true),
exclude_slash_tmp: Some(true),

View File

@@ -358,81 +358,3 @@ async fn test_list_and_resume_conversations() -> Result<()> {
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn list_conversations_fetches_through_filtered_pages() -> Result<()> {
let codex_home = TempDir::new()?;
// Only the last 3 conversations match the provider filter; request 3 and
// ensure pagination keeps fetching past non-matching pages.
let cases = [
(
"2025-03-04T12-00-00",
"2025-03-04T12:00:00Z",
"skip_provider",
),
(
"2025-03-03T12-00-00",
"2025-03-03T12:00:00Z",
"skip_provider",
),
(
"2025-03-02T12-00-00",
"2025-03-02T12:00:00Z",
"target_provider",
),
(
"2025-03-01T12-00-00",
"2025-03-01T12:00:00Z",
"target_provider",
),
(
"2025-02-28T12-00-00",
"2025-02-28T12:00:00Z",
"target_provider",
),
];
for (ts_file, ts_rfc, provider) in cases {
create_fake_rollout(
codex_home.path(),
ts_file,
ts_rfc,
"Hello",
Some(provider),
None,
)?;
}
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let req_id = mcp
.send_list_conversations_request(ListConversationsParams {
page_size: Some(3),
cursor: None,
model_providers: Some(vec!["target_provider".to_string()]),
})
.await?;
let resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(req_id)),
)
.await??;
let ListConversationsResponse { items, next_cursor } =
to_response::<ListConversationsResponse>(resp)?;
assert_eq!(
items.len(),
3,
"should fetch across pages to satisfy the limit"
);
assert!(
items
.iter()
.all(|item| item.model_provider == "target_provider")
);
assert_eq!(next_cursor, None);
Ok(())
}

View File

@@ -1,6 +1,8 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::to_response;
use codex_app_server_protocol::CancelLoginChatGptParams;
use codex_app_server_protocol::CancelLoginChatGptResponse;
use codex_app_server_protocol::GetAuthStatusParams;
use codex_app_server_protocol::GetAuthStatusResponse;
use codex_app_server_protocol::JSONRPCError;
@@ -12,6 +14,7 @@ use codex_core::auth::AuthCredentialsStoreMode;
use codex_login::login_with_api_key;
use serial_test::serial;
use std::path::Path;
use std::time::Duration;
use tempfile::TempDir;
use tokio::time::timeout;
@@ -84,6 +87,48 @@ async fn logout_chatgpt_removes_auth() -> Result<()> {
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
// Serialize tests that launch the login server since it binds to a fixed port.
#[serial(login_port)]
async fn login_and_cancel_chatgpt() -> Result<()> {
let codex_home = TempDir::new()?;
create_config_toml(codex_home.path())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let login_id = mcp.send_login_chat_gpt_request().await?;
let login_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(login_id)),
)
.await??;
let login: LoginChatGptResponse = to_response(login_resp)?;
let cancel_id = mcp
.send_cancel_login_chat_gpt_request(CancelLoginChatGptParams {
login_id: login.login_id,
})
.await?;
let cancel_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(cancel_id)),
)
.await??;
let _ok: CancelLoginChatGptResponse = to_response(cancel_resp)?;
// Optionally observe the completion notification; do not fail if it races.
let maybe_note = timeout(
Duration::from_secs(2),
mcp.read_stream_until_notification_message("codex/event/login_chat_gpt_complete"),
)
.await;
if maybe_note.is_err() {
eprintln!("warning: did not observe login_chat_gpt_complete notification after cancel");
}
Ok(())
}
fn create_config_toml_forced_login(codex_home: &Path, forced_method: &str) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
let contents = format!(

View File

@@ -241,7 +241,7 @@ async fn login_account_chatgpt_rejected_when_forced_api() -> Result<()> {
#[tokio::test]
// Serialize tests that launch the login server since it binds to a fixed port.
#[serial(login_port)]
async fn login_account_chatgpt_start_can_be_cancelled() -> Result<()> {
async fn login_account_chatgpt_start() -> Result<()> {
let codex_home = TempDir::new()?;
create_config_toml(codex_home.path(), CreateConfigTomlParams::default())?;

View File

@@ -1,9 +1,6 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::test_path_buf_with_windows;
use app_test_support::test_tmp_path_buf;
use app_test_support::to_response;
use codex_app_server_protocol::AskForApproval;
use codex_app_server_protocol::ConfigBatchWriteParams;
use codex_app_server_protocol::ConfigEdit;
use codex_app_server_protocol::ConfigLayerName;
@@ -15,8 +12,6 @@ use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::MergeStrategy;
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 pretty_assertions::assert_eq;
use serde_json::json;
@@ -62,7 +57,7 @@ sandbox_mode = "workspace-write"
layers,
} = to_response(resp)?;
assert_eq!(config.model.as_deref(), Some("gpt-user"));
assert_eq!(config.get("model"), Some(&json!("gpt-user")));
assert_eq!(
origins.get("model").expect("origin").name,
ConfigLayerName::User
@@ -76,97 +71,31 @@ sandbox_mode = "workspace-write"
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn config_read_includes_tools() -> Result<()> {
async fn config_read_includes_system_layer_and_overrides() -> Result<()> {
let codex_home = TempDir::new()?;
write_config(
&codex_home,
r#"
model = "gpt-user"
[tools]
web_search = true
view_image = false
"#,
)?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let request_id = mcp
.send_config_read_request(ConfigReadParams {
include_layers: true,
})
.await?;
let resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
)
.await??;
let ConfigReadResponse {
config,
origins,
layers,
} = to_response(resp)?;
let tools = config.tools.expect("tools present");
assert_eq!(
tools,
ToolsV2 {
web_search: Some(true),
view_image: Some(false),
}
);
assert_eq!(
origins.get("tools.web_search").expect("origin").name,
ConfigLayerName::User
);
assert_eq!(
origins.get("tools.view_image").expect("origin").name,
ConfigLayerName::User
);
let layers = layers.expect("layers present");
assert_eq!(layers.len(), 2);
assert_eq!(layers[0].name, ConfigLayerName::SessionFlags);
assert_eq!(layers[1].name, ConfigLayerName::User);
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn config_read_includes_system_layer_and_overrides() -> Result<()> {
let codex_home = TempDir::new()?;
let user_dir = test_path_buf_with_windows("/user", Some(r"C:\Users\user"));
let system_dir = test_path_buf_with_windows("/system", Some(r"C:\System"));
write_config(
&codex_home,
&format!(
r#"
model = "gpt-user"
approval_policy = "on-request"
sandbox_mode = "workspace-write"
[sandbox_workspace_write]
writable_roots = [{}]
writable_roots = ["/user"]
network_access = true
"#,
serde_json::json!(user_dir)
),
)?;
let managed_path = codex_home.path().join("managed_config.toml");
std::fs::write(
&managed_path,
format!(
r#"
r#"
model = "gpt-system"
approval_policy = "never"
[sandbox_workspace_write]
writable_roots = [{}]
writable_roots = ["/system"]
"#,
serde_json::json!(system_dir.clone())
),
)?;
let managed_path_str = managed_path.display().to_string();
@@ -194,29 +123,30 @@ writable_roots = [{}]
layers,
} = to_response(resp)?;
assert_eq!(config.model.as_deref(), Some("gpt-system"));
assert_eq!(config.get("model"), Some(&json!("gpt-system")));
assert_eq!(
origins.get("model").expect("origin").name,
ConfigLayerName::System
);
assert_eq!(config.approval_policy, Some(AskForApproval::Never));
assert_eq!(config.get("approval_policy"), Some(&json!("never")));
assert_eq!(
origins.get("approval_policy").expect("origin").name,
ConfigLayerName::System
);
assert_eq!(config.sandbox_mode, Some(SandboxMode::WorkspaceWrite));
assert_eq!(config.get("sandbox_mode"), Some(&json!("workspace-write")));
assert_eq!(
origins.get("sandbox_mode").expect("origin").name,
ConfigLayerName::User
);
let sandbox = config
.sandbox_workspace_write
.as_ref()
.expect("sandbox workspace write");
assert_eq!(sandbox.writable_roots, vec![system_dir]);
assert_eq!(
config
.get("sandbox_workspace_write")
.and_then(|v| v.get("writable_roots")),
Some(&json!(["/system"]))
);
assert_eq!(
origins
.get("sandbox_workspace_write.writable_roots.0")
@@ -225,7 +155,12 @@ writable_roots = [{}]
ConfigLayerName::System
);
assert!(sandbox.network_access);
assert_eq!(
config
.get("sandbox_workspace_write")
.and_then(|v| v.get("network_access")),
Some(&json!(true))
);
assert_eq!(
origins
.get("sandbox_workspace_write.network_access")
@@ -307,7 +242,7 @@ model = "gpt-old"
)
.await??;
let verify: ConfigReadResponse = to_response(verify_resp)?;
assert_eq!(verify.config.model.as_deref(), Some("gpt-new"));
assert_eq!(verify.config.get("model"), Some(&json!("gpt-new")));
Ok(())
}
@@ -359,7 +294,6 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let writable_root = test_tmp_path_buf();
let batch_id = mcp
.send_config_batch_write_request(ConfigBatchWriteParams {
file_path: Some(codex_home.path().join("config.toml").display().to_string()),
@@ -372,7 +306,7 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
ConfigEdit {
key_path: "sandbox_workspace_write".to_string(),
value: json!({
"writable_roots": [writable_root.clone()],
"writable_roots": ["/tmp"],
"network_access": false
}),
merge_strategy: MergeStrategy::Replace,
@@ -408,14 +342,22 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
)
.await??;
let read: ConfigReadResponse = to_response(read_resp)?;
assert_eq!(read.config.sandbox_mode, Some(SandboxMode::WorkspaceWrite));
let sandbox = read
.config
.sandbox_workspace_write
.as_ref()
.expect("sandbox workspace write");
assert_eq!(sandbox.writable_roots, vec![writable_root]);
assert!(!sandbox.network_access);
assert_eq!(
read.config.get("sandbox_mode"),
Some(&json!("workspace-write"))
);
assert_eq!(
read.config
.get("sandbox_workspace_write")
.and_then(|v| v.get("writable_roots")),
Some(&json!(["/tmp"]))
);
assert_eq!(
read.config
.get("sandbox_workspace_write")
.and_then(|v| v.get("network_access")),
Some(&json!(false))
);
Ok(())
}

View File

@@ -4,7 +4,6 @@ use anyhow::Result;
use anyhow::anyhow;
use app_test_support::McpProcess;
use app_test_support::to_response;
use app_test_support::write_models_cache;
use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::Model;
@@ -23,7 +22,6 @@ const INVALID_REQUEST_ERROR_CODE: i64 = -32600;
#[tokio::test]
async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
let codex_home = TempDir::new()?;
write_models_cache(codex_home.path())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;
@@ -64,7 +62,7 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
},
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::High,
description: "Greater reasoning depth for complex problems".to_string(),
description: "Maximizes reasoning depth for complex problems".to_string(),
},
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::XHigh,
@@ -116,39 +114,6 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
default_reasoning_effort: ReasoningEffort::Medium,
is_default: false,
},
Model {
id: "gpt-5.2".to_string(),
model: "gpt-5.2".to_string(),
display_name: "gpt-5.2".to_string(),
description:
"Latest frontier model with improvements across knowledge, reasoning and coding"
.to_string(),
supported_reasoning_efforts: vec![
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::Low,
description: "Balances speed with some reasoning; useful for straightforward \
queries and short explanations"
.to_string(),
},
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::Medium,
description: "Provides a solid balance of reasoning depth and latency for \
general-purpose tasks"
.to_string(),
},
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::High,
description: "Greater reasoning depth for complex or ambiguous problems"
.to_string(),
},
ReasoningEffortOption {
reasoning_effort: ReasoningEffort::XHigh,
description: "Extra high reasoning for complex problems".to_string(),
},
],
default_reasoning_effort: ReasoningEffort::Medium,
is_default: false,
},
Model {
id: "gpt-5.1".to_string(),
model: "gpt-5.1".to_string(),
@@ -186,7 +151,6 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
#[tokio::test]
async fn list_models_pagination_works() -> Result<()> {
let codex_home = TempDir::new()?;
write_models_cache(codex_home.path())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;
@@ -276,37 +240,14 @@ async fn list_models_pagination_works() -> Result<()> {
} = to_response::<ModelListResponse>(fourth_response)?;
assert_eq!(fourth_items.len(), 1);
assert_eq!(fourth_items[0].id, "gpt-5.2");
let fifth_cursor = fourth_cursor.ok_or_else(|| anyhow!("cursor for fifth page"))?;
let fifth_request = mcp
.send_list_models_request(ModelListParams {
limit: Some(1),
cursor: Some(fifth_cursor.clone()),
})
.await?;
let fifth_response: JSONRPCResponse = timeout(
DEFAULT_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(fifth_request)),
)
.await??;
let ModelListResponse {
data: fifth_items,
next_cursor: fifth_cursor,
} = to_response::<ModelListResponse>(fifth_response)?;
assert_eq!(fifth_items.len(), 1);
assert_eq!(fifth_items[0].id, "gpt-5.1");
assert!(fifth_cursor.is_none());
assert_eq!(fourth_items[0].id, "gpt-5.1");
assert!(fourth_cursor.is_none());
Ok(())
}
#[tokio::test]
async fn list_models_rejects_invalid_cursor() -> Result<()> {
let codex_home = TempDir::new()?;
write_models_cache(codex_home.path())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;

View File

@@ -6,96 +6,37 @@ use codex_app_server_protocol::GitInfo as ApiGitInfo;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::SessionSource;
use codex_app_server_protocol::ThreadListParams;
use codex_app_server_protocol::ThreadListResponse;
use codex_protocol::protocol::GitInfo as CoreGitInfo;
use std::path::Path;
use std::path::PathBuf;
use tempfile::TempDir;
use tokio::time::timeout;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
async fn init_mcp(codex_home: &Path) -> Result<McpProcess> {
let mut mcp = McpProcess::new(codex_home).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
Ok(mcp)
}
async fn list_threads(
mcp: &mut McpProcess,
cursor: Option<String>,
limit: Option<u32>,
providers: Option<Vec<String>>,
) -> Result<ThreadListResponse> {
let request_id = mcp
.send_thread_list_request(codex_app_server_protocol::ThreadListParams {
cursor,
limit,
model_providers: providers,
})
.await?;
let resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
)
.await??;
to_response::<ThreadListResponse>(resp)
}
fn create_fake_rollouts<F, G>(
codex_home: &Path,
count: usize,
provider_for_index: F,
timestamp_for_index: G,
preview: &str,
) -> Result<Vec<String>>
where
F: Fn(usize) -> &'static str,
G: Fn(usize) -> (String, String),
{
let mut ids = Vec::with_capacity(count);
for i in 0..count {
let (ts_file, ts_rfc) = timestamp_for_index(i);
ids.push(create_fake_rollout(
codex_home,
&ts_file,
&ts_rfc,
preview,
Some(provider_for_index(i)),
None,
)?);
}
Ok(ids)
}
fn timestamp_at(
year: i32,
month: u32,
day: u32,
hour: u32,
minute: u32,
second: u32,
) -> (String, String) {
(
format!("{year:04}-{month:02}-{day:02}T{hour:02}-{minute:02}-{second:02}"),
format!("{year:04}-{month:02}-{day:02}T{hour:02}:{minute:02}:{second:02}Z"),
)
}
#[tokio::test]
async fn thread_list_basic_empty() -> Result<()> {
let codex_home = TempDir::new()?;
create_minimal_config(codex_home.path())?;
let mut mcp = init_mcp(codex_home.path()).await?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let ThreadListResponse { data, next_cursor } = list_threads(
&mut mcp,
None,
Some(10),
Some(vec!["mock_provider".to_string()]),
// List threads in an empty CODEX_HOME; should return an empty page with nextCursor: null.
let list_id = mcp
.send_thread_list_request(ThreadListParams {
cursor: None,
limit: Some(10),
model_providers: Some(vec!["mock_provider".to_string()]),
})
.await?;
let list_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
)
.await?;
.await??;
let ThreadListResponse { data, next_cursor } = to_response::<ThreadListResponse>(list_resp)?;
assert!(data.is_empty());
assert_eq!(next_cursor, None);
@@ -145,19 +86,26 @@ async fn thread_list_pagination_next_cursor_none_on_last_page() -> Result<()> {
None,
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
// Page 1: limit 2 → expect next_cursor Some.
let page1_id = mcp
.send_thread_list_request(ThreadListParams {
cursor: None,
limit: Some(2),
model_providers: Some(vec!["mock_provider".to_string()]),
})
.await?;
let page1_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(page1_id)),
)
.await??;
let ThreadListResponse {
data: data1,
next_cursor: cursor1,
} = list_threads(
&mut mcp,
None,
Some(2),
Some(vec!["mock_provider".to_string()]),
)
.await?;
} = to_response::<ThreadListResponse>(page1_resp)?;
assert_eq!(data1.len(), 2);
for thread in &data1 {
assert_eq!(thread.preview, "Hello");
@@ -171,16 +119,22 @@ async fn thread_list_pagination_next_cursor_none_on_last_page() -> Result<()> {
let cursor1 = cursor1.expect("expected nextCursor on first page");
// Page 2: with cursor → expect next_cursor None when no more results.
let page2_id = mcp
.send_thread_list_request(ThreadListParams {
cursor: Some(cursor1),
limit: Some(2),
model_providers: Some(vec!["mock_provider".to_string()]),
})
.await?;
let page2_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(page2_id)),
)
.await??;
let ThreadListResponse {
data: data2,
next_cursor: cursor2,
} = list_threads(
&mut mcp,
Some(cursor1),
Some(2),
Some(vec!["mock_provider".to_string()]),
)
.await?;
} = to_response::<ThreadListResponse>(page2_resp)?;
assert!(data2.len() <= 2);
for thread in &data2 {
assert_eq!(thread.preview, "Hello");
@@ -219,16 +173,23 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
None,
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
// Filter to only other_provider; expect 1 item, nextCursor None.
let ThreadListResponse { data, next_cursor } = list_threads(
&mut mcp,
None,
Some(10),
Some(vec!["other_provider".to_string()]),
let list_id = mcp
.send_thread_list_request(ThreadListParams {
cursor: None,
limit: Some(10),
model_providers: Some(vec!["other_provider".to_string()]),
})
.await?;
let resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
)
.await?;
.await??;
let ThreadListResponse { data, next_cursor } = to_response::<ThreadListResponse>(resp)?;
assert_eq!(data.len(), 1);
assert_eq!(next_cursor, None);
let thread = &data[0];
@@ -244,146 +205,6 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
Ok(())
}
#[tokio::test]
async fn thread_list_fetches_until_limit_or_exhausted() -> Result<()> {
let codex_home = TempDir::new()?;
create_minimal_config(codex_home.path())?;
// Newest 16 conversations belong to a different provider; the older 8 are the
// only ones that match the filter. We request 8 so the server must keep
// paging past the first two pages to reach the desired count.
create_fake_rollouts(
codex_home.path(),
24,
|i| {
if i < 16 {
"skip_provider"
} else {
"target_provider"
}
},
|i| timestamp_at(2025, 3, 30 - i as u32, 12, 0, 0),
"Hello",
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
// Request 8 threads for the target provider; the matches only start on the
// third page so we rely on pagination to reach the limit.
let ThreadListResponse { data, next_cursor } = list_threads(
&mut mcp,
None,
Some(8),
Some(vec!["target_provider".to_string()]),
)
.await?;
assert_eq!(
data.len(),
8,
"should keep paging until the requested count is filled"
);
assert!(
data.iter()
.all(|thread| thread.model_provider == "target_provider"),
"all returned threads must match the requested provider"
);
assert_eq!(
next_cursor, None,
"once the requested count is satisfied on the final page, nextCursor should be None"
);
Ok(())
}
#[tokio::test]
async fn thread_list_enforces_max_limit() -> Result<()> {
let codex_home = TempDir::new()?;
create_minimal_config(codex_home.path())?;
create_fake_rollouts(
codex_home.path(),
105,
|_| "mock_provider",
|i| {
let month = 5 + (i / 28);
let day = (i % 28) + 1;
timestamp_at(2025, month as u32, day as u32, 0, 0, 0)
},
"Hello",
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
let ThreadListResponse { data, next_cursor } = list_threads(
&mut mcp,
None,
Some(200),
Some(vec!["mock_provider".to_string()]),
)
.await?;
assert_eq!(
data.len(),
100,
"limit should be clamped to the maximum page size"
);
assert!(
next_cursor.is_some(),
"when more than the maximum exist, nextCursor should continue pagination"
);
Ok(())
}
#[tokio::test]
async fn thread_list_stops_when_not_enough_filtered_results_exist() -> Result<()> {
let codex_home = TempDir::new()?;
create_minimal_config(codex_home.path())?;
// Only the last 7 conversations match the provider filter; we ask for 10 to
// ensure the server exhausts pagination without looping forever.
create_fake_rollouts(
codex_home.path(),
22,
|i| {
if i < 15 {
"skip_provider"
} else {
"target_provider"
}
},
|i| timestamp_at(2025, 4, 28 - i as u32, 8, 0, 0),
"Hello",
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
// Request more threads than exist after filtering; expect all matches to be
// returned with nextCursor None.
let ThreadListResponse { data, next_cursor } = list_threads(
&mut mcp,
None,
Some(10),
Some(vec!["target_provider".to_string()]),
)
.await?;
assert_eq!(
data.len(),
7,
"all available filtered threads should be returned"
);
assert!(
data.iter()
.all(|thread| thread.model_provider == "target_provider"),
"results should still respect the provider filter"
);
assert_eq!(
next_cursor, None,
"when results are exhausted before reaching the limit, nextCursor should be None"
);
Ok(())
}
#[tokio::test]
async fn thread_list_includes_git_info() -> Result<()> {
let codex_home = TempDir::new()?;
@@ -403,15 +224,22 @@ async fn thread_list_includes_git_info() -> Result<()> {
Some(git_info),
)?;
let mut mcp = init_mcp(codex_home.path()).await?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let ThreadListResponse { data, .. } = list_threads(
&mut mcp,
None,
Some(10),
Some(vec!["mock_provider".to_string()]),
let list_id = mcp
.send_thread_list_request(ThreadListParams {
cursor: None,
limit: Some(10),
model_providers: Some(vec!["mock_provider".to_string()]),
})
.await?;
let resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
)
.await?;
.await??;
let ThreadListResponse { data, .. } = to_response::<ThreadListResponse>(resp)?;
let thread = data
.iter()
.find(|t| t.id == conversation_id)

View File

@@ -427,6 +427,7 @@ async fn turn_start_exec_approval_decline_v2() -> Result<()> {
request_id,
serde_json::to_value(CommandExecutionRequestApprovalResponse {
decision: ApprovalDecision::Decline,
accept_settings: None,
})?,
)
.await?;
@@ -532,7 +533,7 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
cwd: Some(first_cwd.clone()),
approval_policy: Some(codex_app_server_protocol::AskForApproval::Never),
sandbox_policy: Some(codex_app_server_protocol::SandboxPolicy::WorkspaceWrite {
writable_roots: vec![first_cwd.try_into()?],
writable_roots: vec![first_cwd.clone()],
network_access: false,
exclude_tmpdir_env_var: false,
exclude_slash_tmp: false,

View File

@@ -1,813 +0,0 @@
use std::collections::HashMap;
use std::path::Path;
use std::sync::LazyLock;
use tree_sitter::Parser;
use tree_sitter::Query;
use tree_sitter::QueryCursor;
use tree_sitter::StreamingIterator;
use tree_sitter_bash::LANGUAGE as BASH;
use crate::ApplyPatchAction;
use crate::ApplyPatchArgs;
use crate::ApplyPatchError;
use crate::ApplyPatchFileChange;
use crate::ApplyPatchFileUpdate;
use crate::IoError;
use crate::MaybeApplyPatchVerified;
use crate::parser::Hunk;
use crate::parser::ParseError;
use crate::parser::parse_patch;
use crate::unified_diff_from_chunks;
use std::str::Utf8Error;
use tree_sitter::LanguageError;
const APPLY_PATCH_COMMANDS: [&str; 2] = ["apply_patch", "applypatch"];
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum ApplyPatchShell {
Unix,
PowerShell,
Cmd,
}
#[derive(Debug, PartialEq)]
pub enum MaybeApplyPatch {
Body(ApplyPatchArgs),
ShellParseError(ExtractHeredocError),
PatchParseError(ParseError),
NotApplyPatch,
}
#[derive(Debug, PartialEq)]
pub enum ExtractHeredocError {
CommandDidNotStartWithApplyPatch,
FailedToLoadBashGrammar(LanguageError),
HeredocNotUtf8(Utf8Error),
FailedToParsePatchIntoAst,
FailedToFindHeredocBody,
}
fn classify_shell_name(shell: &str) -> Option<String> {
std::path::Path::new(shell)
.file_stem()
.and_then(|name| name.to_str())
.map(str::to_ascii_lowercase)
}
fn classify_shell(shell: &str, flag: &str) -> Option<ApplyPatchShell> {
classify_shell_name(shell).and_then(|name| match name.as_str() {
"bash" | "zsh" | "sh" if matches!(flag, "-lc" | "-c") => Some(ApplyPatchShell::Unix),
"pwsh" | "powershell" if flag.eq_ignore_ascii_case("-command") => {
Some(ApplyPatchShell::PowerShell)
}
"cmd" if flag.eq_ignore_ascii_case("/c") => Some(ApplyPatchShell::Cmd),
_ => None,
})
}
fn can_skip_flag(shell: &str, flag: &str) -> bool {
classify_shell_name(shell).is_some_and(|name| {
matches!(name.as_str(), "pwsh" | "powershell") && flag.eq_ignore_ascii_case("-noprofile")
})
}
fn parse_shell_script(argv: &[String]) -> Option<(ApplyPatchShell, &str)> {
match argv {
[shell, flag, script] => classify_shell(shell, flag).map(|shell_type| {
let script = script.as_str();
(shell_type, script)
}),
[shell, skip_flag, flag, script] if can_skip_flag(shell, skip_flag) => {
classify_shell(shell, flag).map(|shell_type| {
let script = script.as_str();
(shell_type, script)
})
}
_ => None,
}
}
fn extract_apply_patch_from_shell(
shell: ApplyPatchShell,
script: &str,
) -> std::result::Result<(String, Option<String>), ExtractHeredocError> {
match shell {
ApplyPatchShell::Unix | ApplyPatchShell::PowerShell | ApplyPatchShell::Cmd => {
extract_apply_patch_from_bash(script)
}
}
}
// TODO: make private once we remove tests in lib.rs
pub fn maybe_parse_apply_patch(argv: &[String]) -> MaybeApplyPatch {
match argv {
// Direct invocation: apply_patch <patch>
[cmd, body] if APPLY_PATCH_COMMANDS.contains(&cmd.as_str()) => match parse_patch(body) {
Ok(source) => MaybeApplyPatch::Body(source),
Err(e) => MaybeApplyPatch::PatchParseError(e),
},
// Shell heredoc form: (optional `cd <path> &&`) apply_patch <<'EOF' ...
_ => match parse_shell_script(argv) {
Some((shell, script)) => match extract_apply_patch_from_shell(shell, script) {
Ok((body, workdir)) => match parse_patch(&body) {
Ok(mut source) => {
source.workdir = workdir;
MaybeApplyPatch::Body(source)
}
Err(e) => MaybeApplyPatch::PatchParseError(e),
},
Err(ExtractHeredocError::CommandDidNotStartWithApplyPatch) => {
MaybeApplyPatch::NotApplyPatch
}
Err(e) => MaybeApplyPatch::ShellParseError(e),
},
None => MaybeApplyPatch::NotApplyPatch,
},
}
}
/// cwd must be an absolute path so that we can resolve relative paths in the
/// patch.
pub fn maybe_parse_apply_patch_verified(argv: &[String], cwd: &Path) -> MaybeApplyPatchVerified {
// Detect a raw patch body passed directly as the command or as the body of a shell
// script. In these cases, report an explicit error rather than applying the patch.
if let [body] = argv
&& parse_patch(body).is_ok()
{
return MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation);
}
if let Some((_, script)) = parse_shell_script(argv)
&& parse_patch(script).is_ok()
{
return MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation);
}
match maybe_parse_apply_patch(argv) {
MaybeApplyPatch::Body(ApplyPatchArgs {
patch,
hunks,
workdir,
}) => {
let effective_cwd = workdir
.as_ref()
.map(|dir| {
let path = Path::new(dir);
if path.is_absolute() {
path.to_path_buf()
} else {
cwd.join(path)
}
})
.unwrap_or_else(|| cwd.to_path_buf());
let mut changes = HashMap::new();
for hunk in hunks {
let path = hunk.resolve_path(&effective_cwd);
match hunk {
Hunk::AddFile { contents, .. } => {
changes.insert(path, ApplyPatchFileChange::Add { content: contents });
}
Hunk::DeleteFile { .. } => {
let content = match std::fs::read_to_string(&path) {
Ok(content) => content,
Err(e) => {
return MaybeApplyPatchVerified::CorrectnessError(
ApplyPatchError::IoError(IoError {
context: format!("Failed to read {}", path.display()),
source: e,
}),
);
}
};
changes.insert(path, ApplyPatchFileChange::Delete { content });
}
Hunk::UpdateFile {
move_path, chunks, ..
} => {
let ApplyPatchFileUpdate {
unified_diff,
content: contents,
} = match unified_diff_from_chunks(&path, &chunks) {
Ok(diff) => diff,
Err(e) => {
return MaybeApplyPatchVerified::CorrectnessError(e);
}
};
changes.insert(
path,
ApplyPatchFileChange::Update {
unified_diff,
move_path: move_path.map(|p| effective_cwd.join(p)),
new_content: contents,
},
);
}
}
}
MaybeApplyPatchVerified::Body(ApplyPatchAction {
changes,
patch,
cwd: effective_cwd,
})
}
MaybeApplyPatch::ShellParseError(e) => MaybeApplyPatchVerified::ShellParseError(e),
MaybeApplyPatch::PatchParseError(e) => MaybeApplyPatchVerified::CorrectnessError(e.into()),
MaybeApplyPatch::NotApplyPatch => MaybeApplyPatchVerified::NotApplyPatch,
}
}
/// Extract the heredoc body (and optional `cd` workdir) from a `bash -lc` script
/// that invokes the apply_patch tool using a heredoc.
///
/// Supported toplevel forms (must be the only toplevel statement):
/// - `apply_patch <<'EOF'\n...\nEOF`
/// - `cd <path> && apply_patch <<'EOF'\n...\nEOF`
///
/// Notes about matching:
/// - Parsed with Treesitter Bash and a strict query that uses anchors so the
/// heredocredirected statement is the only toplevel statement.
/// - The connector between `cd` and `apply_patch` must be `&&` (not `|` or `||`).
/// - Exactly one positional `word` argument is allowed for `cd` (no flags, no quoted
/// strings, no second argument).
/// - The apply command is validated inquery via `#any-of?` to allow `apply_patch`
/// or `applypatch`.
/// - Preceding or trailing commands (e.g., `echo ...;` or `... && echo done`) do not match.
///
/// Returns `(heredoc_body, Some(path))` when the `cd` variant matches, or
/// `(heredoc_body, None)` for the direct form. Errors are returned if the script
/// cannot be parsed or does not match the allowed patterns.
fn extract_apply_patch_from_bash(
src: &str,
) -> std::result::Result<(String, Option<String>), ExtractHeredocError> {
// This function uses a Tree-sitter query to recognize one of two
// whole-script forms, each expressed as a single top-level statement:
//
// 1. apply_patch <<'EOF'\n...\nEOF
// 2. cd <path> && apply_patch <<'EOF'\n...\nEOF
//
// Key ideas when reading the query:
// - dots (`.`) between named nodes enforces adjacency among named children and
// anchor to the start/end of the expression.
// - we match a single redirected_statement directly under program with leading
// and trailing anchors (`.`). This ensures it is the only top-level statement
// (so prefixes like `echo ...;` or suffixes like `... && echo done` do not match).
//
// Overall, we want to be conservative and only match the intended forms, as other
// forms are likely to be model errors, or incorrectly interpreted by later code.
//
// If you're editing this query, it's helpful to start by creating a debugging binary
// which will let you see the AST of an arbitrary bash script passed in, and optionally
// also run an arbitrary query against the AST. This is useful for understanding
// how tree-sitter parses the script and whether the query syntax is correct. Be sure
// to test both positive and negative cases.
static APPLY_PATCH_QUERY: LazyLock<Query> = LazyLock::new(|| {
let language = BASH.into();
#[expect(clippy::expect_used)]
Query::new(
&language,
r#"
(
program
. (redirected_statement
body: (command
name: (command_name (word) @apply_name) .)
(#any-of? @apply_name "apply_patch" "applypatch")
redirect: (heredoc_redirect
. (heredoc_start)
. (heredoc_body) @heredoc
. (heredoc_end)
.))
.)
(
program
. (redirected_statement
body: (list
. (command
name: (command_name (word) @cd_name) .
argument: [
(word) @cd_path
(string (string_content) @cd_path)
(raw_string) @cd_raw_string
] .)
"&&"
. (command
name: (command_name (word) @apply_name))
.)
(#eq? @cd_name "cd")
(#any-of? @apply_name "apply_patch" "applypatch")
redirect: (heredoc_redirect
. (heredoc_start)
. (heredoc_body) @heredoc
. (heredoc_end)
.))
.)
"#,
)
.expect("valid bash query")
});
let lang = BASH.into();
let mut parser = Parser::new();
parser
.set_language(&lang)
.map_err(ExtractHeredocError::FailedToLoadBashGrammar)?;
let tree = parser
.parse(src, None)
.ok_or(ExtractHeredocError::FailedToParsePatchIntoAst)?;
let bytes = src.as_bytes();
let root = tree.root_node();
let mut cursor = QueryCursor::new();
let mut matches = cursor.matches(&APPLY_PATCH_QUERY, root, bytes);
while let Some(m) = matches.next() {
let mut heredoc_text: Option<String> = None;
let mut cd_path: Option<String> = None;
for capture in m.captures.iter() {
let name = APPLY_PATCH_QUERY.capture_names()[capture.index as usize];
match name {
"heredoc" => {
let text = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?
.trim_end_matches('\n')
.to_string();
heredoc_text = Some(text);
}
"cd_path" => {
let text = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?
.to_string();
cd_path = Some(text);
}
"cd_raw_string" => {
let raw = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?;
let trimmed = raw
.strip_prefix('\'')
.and_then(|s| s.strip_suffix('\''))
.unwrap_or(raw);
cd_path = Some(trimmed.to_string());
}
_ => {}
}
}
if let Some(heredoc) = heredoc_text {
return Ok((heredoc, cd_path));
}
}
Err(ExtractHeredocError::CommandDidNotStartWithApplyPatch)
}
#[cfg(test)]
mod tests {
use super::*;
use assert_matches::assert_matches;
use pretty_assertions::assert_eq;
use std::fs;
use std::path::PathBuf;
use std::string::ToString;
use tempfile::tempdir;
/// Helper to construct a patch with the given body.
fn wrap_patch(body: &str) -> String {
format!("*** Begin Patch\n{body}\n*** End Patch")
}
fn strs_to_strings(strs: &[&str]) -> Vec<String> {
strs.iter().map(ToString::to_string).collect()
}
// Test helpers to reduce repetition when building bash -lc heredoc scripts
fn args_bash(script: &str) -> Vec<String> {
strs_to_strings(&["bash", "-lc", script])
}
fn args_powershell(script: &str) -> Vec<String> {
strs_to_strings(&["powershell.exe", "-Command", script])
}
fn args_powershell_no_profile(script: &str) -> Vec<String> {
strs_to_strings(&["powershell.exe", "-NoProfile", "-Command", script])
}
fn args_pwsh(script: &str) -> Vec<String> {
strs_to_strings(&["pwsh", "-NoProfile", "-Command", script])
}
fn args_cmd(script: &str) -> Vec<String> {
strs_to_strings(&["cmd.exe", "/c", script])
}
fn heredoc_script(prefix: &str) -> String {
format!(
"{prefix}apply_patch <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH"
)
}
fn heredoc_script_ps(prefix: &str, suffix: &str) -> String {
format!(
"{prefix}apply_patch <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH{suffix}"
)
}
fn expected_single_add() -> Vec<Hunk> {
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string(),
}]
}
fn assert_match_args(args: Vec<String>, expected_workdir: Option<&str>) {
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, workdir, .. }) => {
assert_eq!(workdir.as_deref(), expected_workdir);
assert_eq!(hunks, expected_single_add());
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
fn assert_match(script: &str, expected_workdir: Option<&str>) {
let args = args_bash(script);
assert_match_args(args, expected_workdir);
}
fn assert_not_match(script: &str) {
let args = args_bash(script);
assert_matches!(
maybe_parse_apply_patch(&args),
MaybeApplyPatch::NotApplyPatch
);
}
#[test]
fn test_implicit_patch_single_arg_is_error() {
let patch = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch".to_string();
let args = vec![patch];
let dir = tempdir().unwrap();
assert_matches!(
maybe_parse_apply_patch_verified(&args, dir.path()),
MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
);
}
#[test]
fn test_implicit_patch_bash_script_is_error() {
let script = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch";
let args = args_bash(script);
let dir = tempdir().unwrap();
assert_matches!(
maybe_parse_apply_patch_verified(&args, dir.path()),
MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
);
}
#[test]
fn test_literal() {
let args = strs_to_strings(&[
"apply_patch",
r#"*** Begin Patch
*** Add File: foo
+hi
*** End Patch
"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, .. }) => {
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_literal_applypatch() {
let args = strs_to_strings(&[
"applypatch",
r#"*** Begin Patch
*** Add File: foo
+hi
*** End Patch
"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, .. }) => {
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_heredoc() {
assert_match(&heredoc_script(""), None);
}
#[test]
fn test_heredoc_non_login_shell() {
let script = heredoc_script("");
let args = strs_to_strings(&["bash", "-c", &script]);
assert_match_args(args, None);
}
#[test]
fn test_heredoc_applypatch() {
let args = strs_to_strings(&[
"bash",
"-lc",
r#"applypatch <<'PATCH'
*** Begin Patch
*** Add File: foo
+hi
*** End Patch
PATCH"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, workdir, .. }) => {
assert_eq!(workdir, None);
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_powershell_heredoc() {
let script = heredoc_script("");
assert_match_args(args_powershell(&script), None);
}
#[test]
fn test_powershell_heredoc_no_profile() {
let script = heredoc_script("");
assert_match_args(args_powershell_no_profile(&script), None);
}
#[test]
fn test_pwsh_heredoc() {
let script = heredoc_script("");
assert_match_args(args_pwsh(&script), None);
}
#[test]
fn test_cmd_heredoc_with_cd() {
let script = heredoc_script("cd foo && ");
assert_match_args(args_cmd(&script), Some("foo"));
}
#[test]
fn test_heredoc_with_leading_cd() {
assert_match(&heredoc_script("cd foo && "), Some("foo"));
}
#[test]
fn test_cd_with_semicolon_is_ignored() {
assert_not_match(&heredoc_script("cd foo; "));
}
#[test]
fn test_cd_or_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd bar || "));
}
#[test]
fn test_cd_pipe_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd bar | "));
}
#[test]
fn test_cd_single_quoted_path_with_spaces() {
assert_match(&heredoc_script("cd 'foo bar' && "), Some("foo bar"));
}
#[test]
fn test_cd_double_quoted_path_with_spaces() {
assert_match(&heredoc_script("cd \"foo bar\" && "), Some("foo bar"));
}
#[test]
fn test_echo_and_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("echo foo && "));
}
#[test]
fn test_apply_patch_with_arg_is_ignored() {
let script = "apply_patch foo <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH";
assert_not_match(script);
}
#[test]
fn test_double_cd_then_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd foo && cd bar && "));
}
#[test]
fn test_cd_two_args_is_ignored() {
assert_not_match(&heredoc_script("cd foo bar && "));
}
#[test]
fn test_cd_then_apply_patch_then_extra_is_ignored() {
let script = heredoc_script_ps("cd bar && ", " && echo done");
assert_not_match(&script);
}
#[test]
fn test_echo_then_cd_and_apply_patch_is_ignored() {
// Ensure preceding commands before the `cd && apply_patch <<...` sequence do not match.
assert_not_match(&heredoc_script("echo foo; cd bar && "));
}
#[test]
fn test_unified_diff_last_line_replacement() {
// Replace the very last line of the file.
let dir = tempdir().unwrap();
let path = dir.path().join("last.txt");
fs::write(&path, "foo\nbar\nbaz\n").unwrap();
let patch = wrap_patch(&format!(
r#"*** Update File: {}
@@
foo
bar
-baz
+BAZ
"#,
path.display()
));
let patch = parse_patch(&patch).unwrap();
let chunks = match patch.hunks.as_slice() {
[Hunk::UpdateFile { chunks, .. }] => chunks,
_ => panic!("Expected a single UpdateFile hunk"),
};
let diff = unified_diff_from_chunks(&path, chunks).unwrap();
let expected_diff = r#"@@ -2,2 +2,2 @@
bar
-baz
+BAZ
"#;
let expected = ApplyPatchFileUpdate {
unified_diff: expected_diff.to_string(),
content: "foo\nbar\nBAZ\n".to_string(),
};
assert_eq!(expected, diff);
}
#[test]
fn test_unified_diff_insert_at_eof() {
// Insert a new line at endoffile.
let dir = tempdir().unwrap();
let path = dir.path().join("insert.txt");
fs::write(&path, "foo\nbar\nbaz\n").unwrap();
let patch = wrap_patch(&format!(
r#"*** Update File: {}
@@
+quux
*** End of File
"#,
path.display()
));
let patch = parse_patch(&patch).unwrap();
let chunks = match patch.hunks.as_slice() {
[Hunk::UpdateFile { chunks, .. }] => chunks,
_ => panic!("Expected a single UpdateFile hunk"),
};
let diff = unified_diff_from_chunks(&path, chunks).unwrap();
let expected_diff = r#"@@ -3 +3,2 @@
baz
+quux
"#;
let expected = ApplyPatchFileUpdate {
unified_diff: expected_diff.to_string(),
content: "foo\nbar\nbaz\nquux\n".to_string(),
};
assert_eq!(expected, diff);
}
#[test]
fn test_apply_patch_should_resolve_absolute_paths_in_cwd() {
let session_dir = tempdir().unwrap();
let relative_path = "source.txt";
// Note that we need this file to exist for the patch to be "verified"
// and parsed correctly.
let session_file_path = session_dir.path().join(relative_path);
fs::write(&session_file_path, "session directory content\n").unwrap();
let argv = vec![
"apply_patch".to_string(),
r#"*** Begin Patch
*** Update File: source.txt
@@
-session directory content
+updated session directory content
*** End Patch"#
.to_string(),
];
let result = maybe_parse_apply_patch_verified(&argv, session_dir.path());
// Verify the patch contents - as otherwise we may have pulled contents
// from the wrong file (as we're using relative paths)
assert_eq!(
result,
MaybeApplyPatchVerified::Body(ApplyPatchAction {
changes: HashMap::from([(
session_dir.path().join(relative_path),
ApplyPatchFileChange::Update {
unified_diff: r#"@@ -1 +1 @@
-session directory content
+updated session directory content
"#
.to_string(),
move_path: None,
new_content: "updated session directory content\n".to_string(),
},
)]),
patch: argv[1].clone(),
cwd: session_dir.path().to_path_buf(),
})
);
}
#[test]
fn test_apply_patch_resolves_move_path_with_effective_cwd() {
let session_dir = tempdir().unwrap();
let worktree_rel = "alt";
let worktree_dir = session_dir.path().join(worktree_rel);
fs::create_dir_all(&worktree_dir).unwrap();
let source_name = "old.txt";
let dest_name = "renamed.txt";
let source_path = worktree_dir.join(source_name);
fs::write(&source_path, "before\n").unwrap();
let patch = wrap_patch(&format!(
r#"*** Update File: {source_name}
*** Move to: {dest_name}
@@
-before
+after"#
));
let shell_script = format!("cd {worktree_rel} && apply_patch <<'PATCH'\n{patch}\nPATCH");
let argv = vec!["bash".into(), "-lc".into(), shell_script];
let result = maybe_parse_apply_patch_verified(&argv, session_dir.path());
let action = match result {
MaybeApplyPatchVerified::Body(action) => action,
other => panic!("expected verified body, got {other:?}"),
};
assert_eq!(action.cwd, worktree_dir);
let change = action
.changes()
.get(&worktree_dir.join(source_name))
.expect("source file change present");
match change {
ApplyPatchFileChange::Update { move_path, .. } => {
assert_eq!(
move_path.as_deref(),
Some(worktree_dir.join(dest_name).as_path())
);
}
other => panic!("expected update change, got {other:?}"),
}
}
}

View File

@@ -1,4 +1,3 @@
mod invocation;
mod parser;
mod seek_sequence;
mod standalone_executable;
@@ -6,6 +5,8 @@ mod standalone_executable;
use std::collections::HashMap;
use std::path::Path;
use std::path::PathBuf;
use std::str::Utf8Error;
use std::sync::LazyLock;
use anyhow::Context;
use anyhow::Result;
@@ -16,15 +17,27 @@ use parser::UpdateFileChunk;
pub use parser::parse_patch;
use similar::TextDiff;
use thiserror::Error;
use tree_sitter::LanguageError;
use tree_sitter::Parser;
use tree_sitter::Query;
use tree_sitter::QueryCursor;
use tree_sitter::StreamingIterator;
use tree_sitter_bash::LANGUAGE as BASH;
pub use invocation::maybe_parse_apply_patch_verified;
pub use standalone_executable::main;
use crate::invocation::ExtractHeredocError;
/// Detailed instructions for gpt-4.1 on how to use the `apply_patch` tool.
pub const APPLY_PATCH_TOOL_INSTRUCTIONS: &str = include_str!("../apply_patch_tool_instructions.md");
const APPLY_PATCH_COMMANDS: [&str; 2] = ["apply_patch", "applypatch"];
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum ApplyPatchShell {
Unix,
PowerShell,
Cmd,
}
#[derive(Debug, Error, PartialEq)]
pub enum ApplyPatchError {
#[error(transparent)]
@@ -73,6 +86,14 @@ impl PartialEq for IoError {
}
}
#[derive(Debug, PartialEq)]
pub enum MaybeApplyPatch {
Body(ApplyPatchArgs),
ShellParseError(ExtractHeredocError),
PatchParseError(ParseError),
NotApplyPatch,
}
/// Both the raw PATCH argument to `apply_patch` as well as the PATCH argument
/// parsed into hunks.
#[derive(Debug, PartialEq)]
@@ -82,6 +103,84 @@ pub struct ApplyPatchArgs {
pub workdir: Option<String>,
}
fn classify_shell_name(shell: &str) -> Option<String> {
std::path::Path::new(shell)
.file_stem()
.and_then(|name| name.to_str())
.map(str::to_ascii_lowercase)
}
fn classify_shell(shell: &str, flag: &str) -> Option<ApplyPatchShell> {
classify_shell_name(shell).and_then(|name| match name.as_str() {
"bash" | "zsh" | "sh" if flag == "-lc" => Some(ApplyPatchShell::Unix),
"pwsh" | "powershell" if flag.eq_ignore_ascii_case("-command") => {
Some(ApplyPatchShell::PowerShell)
}
"cmd" if flag.eq_ignore_ascii_case("/c") => Some(ApplyPatchShell::Cmd),
_ => None,
})
}
fn can_skip_flag(shell: &str, flag: &str) -> bool {
classify_shell_name(shell).is_some_and(|name| {
matches!(name.as_str(), "pwsh" | "powershell") && flag.eq_ignore_ascii_case("-noprofile")
})
}
fn parse_shell_script(argv: &[String]) -> Option<(ApplyPatchShell, &str)> {
match argv {
[shell, flag, script] => classify_shell(shell, flag).map(|shell_type| {
let script = script.as_str();
(shell_type, script)
}),
[shell, skip_flag, flag, script] if can_skip_flag(shell, skip_flag) => {
classify_shell(shell, flag).map(|shell_type| {
let script = script.as_str();
(shell_type, script)
})
}
_ => None,
}
}
fn extract_apply_patch_from_shell(
shell: ApplyPatchShell,
script: &str,
) -> std::result::Result<(String, Option<String>), ExtractHeredocError> {
match shell {
ApplyPatchShell::Unix | ApplyPatchShell::PowerShell | ApplyPatchShell::Cmd => {
extract_apply_patch_from_bash(script)
}
}
}
pub fn maybe_parse_apply_patch(argv: &[String]) -> MaybeApplyPatch {
match argv {
// Direct invocation: apply_patch <patch>
[cmd, body] if APPLY_PATCH_COMMANDS.contains(&cmd.as_str()) => match parse_patch(body) {
Ok(source) => MaybeApplyPatch::Body(source),
Err(e) => MaybeApplyPatch::PatchParseError(e),
},
// Shell heredoc form: (optional `cd <path> &&`) apply_patch <<'EOF' ...
_ => match parse_shell_script(argv) {
Some((shell, script)) => match extract_apply_patch_from_shell(shell, script) {
Ok((body, workdir)) => match parse_patch(&body) {
Ok(mut source) => {
source.workdir = workdir;
MaybeApplyPatch::Body(source)
}
Err(e) => MaybeApplyPatch::PatchParseError(e),
},
Err(ExtractHeredocError::CommandDidNotStartWithApplyPatch) => {
MaybeApplyPatch::NotApplyPatch
}
Err(e) => MaybeApplyPatch::ShellParseError(e),
},
None => MaybeApplyPatch::NotApplyPatch,
},
}
}
#[derive(Debug, PartialEq)]
pub enum ApplyPatchFileChange {
Add {
@@ -170,6 +269,256 @@ impl ApplyPatchAction {
}
}
/// cwd must be an absolute path so that we can resolve relative paths in the
/// patch.
pub fn maybe_parse_apply_patch_verified(argv: &[String], cwd: &Path) -> MaybeApplyPatchVerified {
// Detect a raw patch body passed directly as the command or as the body of a shell
// script. In these cases, report an explicit error rather than applying the patch.
if let [body] = argv
&& parse_patch(body).is_ok()
{
return MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation);
}
if let Some((_, script)) = parse_shell_script(argv)
&& parse_patch(script).is_ok()
{
return MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation);
}
match maybe_parse_apply_patch(argv) {
MaybeApplyPatch::Body(ApplyPatchArgs {
patch,
hunks,
workdir,
}) => {
let effective_cwd = workdir
.as_ref()
.map(|dir| {
let path = Path::new(dir);
if path.is_absolute() {
path.to_path_buf()
} else {
cwd.join(path)
}
})
.unwrap_or_else(|| cwd.to_path_buf());
let mut changes = HashMap::new();
for hunk in hunks {
let path = hunk.resolve_path(&effective_cwd);
match hunk {
Hunk::AddFile { contents, .. } => {
changes.insert(path, ApplyPatchFileChange::Add { content: contents });
}
Hunk::DeleteFile { .. } => {
let content = match std::fs::read_to_string(&path) {
Ok(content) => content,
Err(e) => {
return MaybeApplyPatchVerified::CorrectnessError(
ApplyPatchError::IoError(IoError {
context: format!("Failed to read {}", path.display()),
source: e,
}),
);
}
};
changes.insert(path, ApplyPatchFileChange::Delete { content });
}
Hunk::UpdateFile {
move_path, chunks, ..
} => {
let ApplyPatchFileUpdate {
unified_diff,
content: contents,
} = match unified_diff_from_chunks(&path, &chunks) {
Ok(diff) => diff,
Err(e) => {
return MaybeApplyPatchVerified::CorrectnessError(e);
}
};
changes.insert(
path,
ApplyPatchFileChange::Update {
unified_diff,
move_path: move_path.map(|p| effective_cwd.join(p)),
new_content: contents,
},
);
}
}
}
MaybeApplyPatchVerified::Body(ApplyPatchAction {
changes,
patch,
cwd: effective_cwd,
})
}
MaybeApplyPatch::ShellParseError(e) => MaybeApplyPatchVerified::ShellParseError(e),
MaybeApplyPatch::PatchParseError(e) => MaybeApplyPatchVerified::CorrectnessError(e.into()),
MaybeApplyPatch::NotApplyPatch => MaybeApplyPatchVerified::NotApplyPatch,
}
}
/// Extract the heredoc body (and optional `cd` workdir) from a `bash -lc` script
/// that invokes the apply_patch tool using a heredoc.
///
/// Supported toplevel forms (must be the only toplevel statement):
/// - `apply_patch <<'EOF'\n...\nEOF`
/// - `cd <path> && apply_patch <<'EOF'\n...\nEOF`
///
/// Notes about matching:
/// - Parsed with Treesitter Bash and a strict query that uses anchors so the
/// heredocredirected statement is the only toplevel statement.
/// - The connector between `cd` and `apply_patch` must be `&&` (not `|` or `||`).
/// - Exactly one positional `word` argument is allowed for `cd` (no flags, no quoted
/// strings, no second argument).
/// - The apply command is validated inquery via `#any-of?` to allow `apply_patch`
/// or `applypatch`.
/// - Preceding or trailing commands (e.g., `echo ...;` or `... && echo done`) do not match.
///
/// Returns `(heredoc_body, Some(path))` when the `cd` variant matches, or
/// `(heredoc_body, None)` for the direct form. Errors are returned if the script
/// cannot be parsed or does not match the allowed patterns.
fn extract_apply_patch_from_bash(
src: &str,
) -> std::result::Result<(String, Option<String>), ExtractHeredocError> {
// This function uses a Tree-sitter query to recognize one of two
// whole-script forms, each expressed as a single top-level statement:
//
// 1. apply_patch <<'EOF'\n...\nEOF
// 2. cd <path> && apply_patch <<'EOF'\n...\nEOF
//
// Key ideas when reading the query:
// - dots (`.`) between named nodes enforces adjacency among named children and
// anchor to the start/end of the expression.
// - we match a single redirected_statement directly under program with leading
// and trailing anchors (`.`). This ensures it is the only top-level statement
// (so prefixes like `echo ...;` or suffixes like `... && echo done` do not match).
//
// Overall, we want to be conservative and only match the intended forms, as other
// forms are likely to be model errors, or incorrectly interpreted by later code.
//
// If you're editing this query, it's helpful to start by creating a debugging binary
// which will let you see the AST of an arbitrary bash script passed in, and optionally
// also run an arbitrary query against the AST. This is useful for understanding
// how tree-sitter parses the script and whether the query syntax is correct. Be sure
// to test both positive and negative cases.
static APPLY_PATCH_QUERY: LazyLock<Query> = LazyLock::new(|| {
let language = BASH.into();
#[expect(clippy::expect_used)]
Query::new(
&language,
r#"
(
program
. (redirected_statement
body: (command
name: (command_name (word) @apply_name) .)
(#any-of? @apply_name "apply_patch" "applypatch")
redirect: (heredoc_redirect
. (heredoc_start)
. (heredoc_body) @heredoc
. (heredoc_end)
.))
.)
(
program
. (redirected_statement
body: (list
. (command
name: (command_name (word) @cd_name) .
argument: [
(word) @cd_path
(string (string_content) @cd_path)
(raw_string) @cd_raw_string
] .)
"&&"
. (command
name: (command_name (word) @apply_name))
.)
(#eq? @cd_name "cd")
(#any-of? @apply_name "apply_patch" "applypatch")
redirect: (heredoc_redirect
. (heredoc_start)
. (heredoc_body) @heredoc
. (heredoc_end)
.))
.)
"#,
)
.expect("valid bash query")
});
let lang = BASH.into();
let mut parser = Parser::new();
parser
.set_language(&lang)
.map_err(ExtractHeredocError::FailedToLoadBashGrammar)?;
let tree = parser
.parse(src, None)
.ok_or(ExtractHeredocError::FailedToParsePatchIntoAst)?;
let bytes = src.as_bytes();
let root = tree.root_node();
let mut cursor = QueryCursor::new();
let mut matches = cursor.matches(&APPLY_PATCH_QUERY, root, bytes);
while let Some(m) = matches.next() {
let mut heredoc_text: Option<String> = None;
let mut cd_path: Option<String> = None;
for capture in m.captures.iter() {
let name = APPLY_PATCH_QUERY.capture_names()[capture.index as usize];
match name {
"heredoc" => {
let text = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?
.trim_end_matches('\n')
.to_string();
heredoc_text = Some(text);
}
"cd_path" => {
let text = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?
.to_string();
cd_path = Some(text);
}
"cd_raw_string" => {
let raw = capture
.node
.utf8_text(bytes)
.map_err(ExtractHeredocError::HeredocNotUtf8)?;
let trimmed = raw
.strip_prefix('\'')
.and_then(|s| s.strip_suffix('\''))
.unwrap_or(raw);
cd_path = Some(trimmed.to_string());
}
_ => {}
}
}
if let Some(heredoc) = heredoc_text {
return Ok((heredoc, cd_path));
}
}
Err(ExtractHeredocError::CommandDidNotStartWithApplyPatch)
}
#[derive(Debug, PartialEq)]
pub enum ExtractHeredocError {
CommandDidNotStartWithApplyPatch,
FailedToLoadBashGrammar(LanguageError),
HeredocNotUtf8(Utf8Error),
FailedToParsePatchIntoAst,
FailedToFindHeredocBody,
}
/// Applies the patch and prints the result to stdout/stderr.
pub fn apply_patch(
patch: &str,
@@ -350,13 +699,7 @@ fn derive_new_contents_from_chunks(
}
};
let mut original_lines: Vec<String> = original_contents.split('\n').map(String::from).collect();
// Drop the trailing empty element that results from the final newline so
// that line counts match the behaviour of standard `diff`.
if original_lines.last().is_some_and(String::is_empty) {
original_lines.pop();
}
let original_lines: Vec<String> = build_lines_from_contents(&original_contents);
let replacements = compute_replacements(&original_lines, path, chunks)?;
let new_lines = apply_replacements(original_lines, &replacements);
@@ -364,13 +707,67 @@ fn derive_new_contents_from_chunks(
if !new_lines.last().is_some_and(String::is_empty) {
new_lines.push(String::new());
}
let new_contents = new_lines.join("\n");
let new_contents = build_contents_from_lines(&original_contents, &new_lines);
Ok(AppliedPatch {
original_contents,
new_contents,
})
}
// TODO(dylan-hurd-oai): I think we can migrate to just use `contents.lines()`
// across all platforms.
fn build_lines_from_contents(contents: &str) -> Vec<String> {
if cfg!(windows) {
contents.lines().map(String::from).collect()
} else {
let mut lines: Vec<String> = contents.split('\n').map(String::from).collect();
// Drop the trailing empty element that results from the final newline so
// that line counts match the behaviour of standard `diff`.
if lines.last().is_some_and(String::is_empty) {
lines.pop();
}
lines
}
}
fn build_contents_from_lines(original_contents: &str, lines: &[String]) -> String {
if cfg!(windows) {
// for now, only compute this if we're on Windows.
let uses_crlf = contents_uses_crlf(original_contents);
if uses_crlf {
lines.join("\r\n")
} else {
lines.join("\n")
}
} else {
lines.join("\n")
}
}
/// Detects whether the source file uses Windows CRLF line endings consistently.
/// We only consider a file CRLF-formatted if every newline is part of a
/// CRLF sequence. This avoids rewriting an LF-formatted file that merely
/// contains embedded sequences of "\r\n".
///
/// Returns `true` if the file uses CRLF line endings, `false` otherwise.
fn contents_uses_crlf(contents: &str) -> bool {
let bytes = contents.as_bytes();
let mut n_newlines = 0usize;
let mut n_crlf = 0usize;
for i in 0..bytes.len() {
if bytes[i] == b'\n' {
n_newlines += 1;
if i > 0 && bytes[i - 1] == b'\r' {
n_crlf += 1;
}
}
}
n_newlines > 0 && n_crlf == n_newlines
}
/// Compute a list of replacements needed to transform `original_lines` into the
/// new lines, given the patch `chunks`. Each replacement is returned as
/// `(start_index, old_len, new_lines)`.
@@ -545,6 +942,7 @@ pub fn print_summary(
#[cfg(test)]
mod tests {
use super::*;
use assert_matches::assert_matches;
use pretty_assertions::assert_eq;
use std::fs;
use std::string::ToString;
@@ -555,6 +953,263 @@ mod tests {
format!("*** Begin Patch\n{body}\n*** End Patch")
}
fn strs_to_strings(strs: &[&str]) -> Vec<String> {
strs.iter().map(ToString::to_string).collect()
}
// Test helpers to reduce repetition when building bash -lc heredoc scripts
fn args_bash(script: &str) -> Vec<String> {
strs_to_strings(&["bash", "-lc", script])
}
fn args_powershell(script: &str) -> Vec<String> {
strs_to_strings(&["powershell.exe", "-Command", script])
}
fn args_powershell_no_profile(script: &str) -> Vec<String> {
strs_to_strings(&["powershell.exe", "-NoProfile", "-Command", script])
}
fn args_pwsh(script: &str) -> Vec<String> {
strs_to_strings(&["pwsh", "-NoProfile", "-Command", script])
}
fn args_cmd(script: &str) -> Vec<String> {
strs_to_strings(&["cmd.exe", "/c", script])
}
fn heredoc_script(prefix: &str) -> String {
format!(
"{prefix}apply_patch <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH"
)
}
fn heredoc_script_ps(prefix: &str, suffix: &str) -> String {
format!(
"{prefix}apply_patch <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH{suffix}"
)
}
fn expected_single_add() -> Vec<Hunk> {
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string(),
}]
}
fn assert_match_args(args: Vec<String>, expected_workdir: Option<&str>) {
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, workdir, .. }) => {
assert_eq!(workdir.as_deref(), expected_workdir);
assert_eq!(hunks, expected_single_add());
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
fn assert_match(script: &str, expected_workdir: Option<&str>) {
let args = args_bash(script);
assert_match_args(args, expected_workdir);
}
fn assert_not_match(script: &str) {
let args = args_bash(script);
assert_matches!(
maybe_parse_apply_patch(&args),
MaybeApplyPatch::NotApplyPatch
);
}
#[test]
fn test_implicit_patch_single_arg_is_error() {
let patch = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch".to_string();
let args = vec![patch];
let dir = tempdir().unwrap();
assert_matches!(
maybe_parse_apply_patch_verified(&args, dir.path()),
MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
);
}
#[test]
fn test_implicit_patch_bash_script_is_error() {
let script = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch";
let args = args_bash(script);
let dir = tempdir().unwrap();
assert_matches!(
maybe_parse_apply_patch_verified(&args, dir.path()),
MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
);
}
#[test]
fn test_literal() {
let args = strs_to_strings(&[
"apply_patch",
r#"*** Begin Patch
*** Add File: foo
+hi
*** End Patch
"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, .. }) => {
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_literal_applypatch() {
let args = strs_to_strings(&[
"applypatch",
r#"*** Begin Patch
*** Add File: foo
+hi
*** End Patch
"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, .. }) => {
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_heredoc() {
assert_match(&heredoc_script(""), None);
}
#[test]
fn test_heredoc_applypatch() {
let args = strs_to_strings(&[
"bash",
"-lc",
r#"applypatch <<'PATCH'
*** Begin Patch
*** Add File: foo
+hi
*** End Patch
PATCH"#,
]);
match maybe_parse_apply_patch(&args) {
MaybeApplyPatch::Body(ApplyPatchArgs { hunks, workdir, .. }) => {
assert_eq!(workdir, None);
assert_eq!(
hunks,
vec![Hunk::AddFile {
path: PathBuf::from("foo"),
contents: "hi\n".to_string()
}]
);
}
result => panic!("expected MaybeApplyPatch::Body got {result:?}"),
}
}
#[test]
fn test_powershell_heredoc() {
let script = heredoc_script("");
assert_match_args(args_powershell(&script), None);
}
#[test]
fn test_powershell_heredoc_no_profile() {
let script = heredoc_script("");
assert_match_args(args_powershell_no_profile(&script), None);
}
#[test]
fn test_pwsh_heredoc() {
let script = heredoc_script("");
assert_match_args(args_pwsh(&script), None);
}
#[test]
fn test_cmd_heredoc_with_cd() {
let script = heredoc_script("cd foo && ");
assert_match_args(args_cmd(&script), Some("foo"));
}
#[test]
fn test_heredoc_with_leading_cd() {
assert_match(&heredoc_script("cd foo && "), Some("foo"));
}
#[test]
fn test_cd_with_semicolon_is_ignored() {
assert_not_match(&heredoc_script("cd foo; "));
}
#[test]
fn test_cd_or_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd bar || "));
}
#[test]
fn test_cd_pipe_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd bar | "));
}
#[test]
fn test_cd_single_quoted_path_with_spaces() {
assert_match(&heredoc_script("cd 'foo bar' && "), Some("foo bar"));
}
#[test]
fn test_cd_double_quoted_path_with_spaces() {
assert_match(&heredoc_script("cd \"foo bar\" && "), Some("foo bar"));
}
#[test]
fn test_echo_and_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("echo foo && "));
}
#[test]
fn test_apply_patch_with_arg_is_ignored() {
let script = "apply_patch foo <<'PATCH'\n*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch\nPATCH";
assert_not_match(script);
}
#[test]
fn test_double_cd_then_apply_patch_is_ignored() {
assert_not_match(&heredoc_script("cd foo && cd bar && "));
}
#[test]
fn test_cd_two_args_is_ignored() {
assert_not_match(&heredoc_script("cd foo bar && "));
}
#[test]
fn test_cd_then_apply_patch_then_extra_is_ignored() {
let script = heredoc_script_ps("cd bar && ", " && echo done");
assert_not_match(&script);
}
#[test]
fn test_echo_then_cd_and_apply_patch_is_ignored() {
// Ensure preceding commands before the `cd && apply_patch <<...` sequence do not match.
assert_not_match(&heredoc_script("echo foo; cd bar && "));
}
#[test]
fn test_add_file_hunk_creates_file_with_contents() {
let dir = tempdir().unwrap();
@@ -752,6 +1407,72 @@ mod tests {
assert_eq!(contents, "a\nB\nc\nd\nE\nf\ng\n");
}
/// Ensure CRLF line endings are preserved for updated files on Windowsstyle inputs.
#[cfg(windows)]
#[test]
fn test_preserve_crlf_line_endings_on_update() {
let dir = tempdir().unwrap();
let path = dir.path().join("crlf.txt");
// Original file uses CRLF (\r\n) endings.
std::fs::write(&path, b"a\r\nb\r\nc\r\n").unwrap();
// Replace `b` -> `B` and append `d`.
let patch = wrap_patch(&format!(
r#"*** Update File: {}
@@
a
-b
+B
@@
c
+d
*** End of File"#,
path.display()
));
let mut stdout = Vec::new();
let mut stderr = Vec::new();
apply_patch(&patch, &mut stdout, &mut stderr).unwrap();
let out = std::fs::read(&path).unwrap();
// Expect all CRLF endings; count occurrences of CRLF and ensure there are 4 lines.
let content = String::from_utf8_lossy(&out);
assert!(content.contains("\r\n"));
// No bare LF occurrences immediately preceding a non-CR: the text should not contain "a\nb".
assert!(!content.contains("a\nb"));
// Validate exact content sequence with CRLF delimiters.
assert_eq!(content, "a\r\nB\r\nc\r\nd\r\n");
}
/// Ensure CRLF inputs with embedded carriage returns in the content are preserved.
#[cfg(windows)]
#[test]
fn test_preserve_crlf_embedded_carriage_returns_on_append() {
let dir = tempdir().unwrap();
let path = dir.path().join("crlf_cr_content.txt");
// Original file: first line has a literal '\r' in the content before the CRLF terminator.
std::fs::write(&path, b"foo\r\r\nbar\r\n").unwrap();
// Append a new line without modifying existing ones.
let patch = wrap_patch(&format!(
r#"*** Update File: {}
@@
+BAZ
*** End of File"#,
path.display()
));
let mut stdout = Vec::new();
let mut stderr = Vec::new();
apply_patch(&patch, &mut stdout, &mut stderr).unwrap();
let out = std::fs::read(&path).unwrap();
// CRLF endings must be preserved and the extra CR in "foo\r\r" must not be collapsed.
assert_eq!(out.as_slice(), b"foo\r\r\nbar\r\nBAZ\r\n");
}
#[test]
fn test_pure_addition_chunk_followed_by_removal() {
let dir = tempdir().unwrap();
@@ -937,6 +1658,37 @@ mod tests {
assert_eq!(expected, diff);
}
/// For LF-only inputs with a trailing newline ensure that the helper used
/// on Windows-style builds drops the synthetic trailing empty element so
/// replacements behave like standard `diff` line numbering.
#[test]
fn test_derive_new_contents_lf_trailing_newline() {
let dir = tempdir().unwrap();
let path = dir.path().join("lf_trailing_newline.txt");
fs::write(&path, "foo\nbar\n").unwrap();
let patch = wrap_patch(&format!(
r#"*** Update File: {}
@@
foo
-bar
+BAR
"#,
path.display()
));
let patch = parse_patch(&patch).unwrap();
let chunks = match patch.hunks.as_slice() {
[Hunk::UpdateFile { chunks, .. }] => chunks,
_ => panic!("Expected a single UpdateFile hunk"),
};
let AppliedPatch { new_contents, .. } =
derive_new_contents_from_chunks(&path, chunks).unwrap();
assert_eq!(new_contents, "foo\nBAR\n");
}
#[test]
fn test_unified_diff_insert_at_eof() {
// Insert a new line at endoffile.
@@ -1043,6 +1795,99 @@ g
);
}
#[test]
fn test_apply_patch_should_resolve_absolute_paths_in_cwd() {
let session_dir = tempdir().unwrap();
let relative_path = "source.txt";
// Note that we need this file to exist for the patch to be "verified"
// and parsed correctly.
let session_file_path = session_dir.path().join(relative_path);
fs::write(&session_file_path, "session directory content\n").unwrap();
let argv = vec![
"apply_patch".to_string(),
r#"*** Begin Patch
*** Update File: source.txt
@@
-session directory content
+updated session directory content
*** End Patch"#
.to_string(),
];
let result = maybe_parse_apply_patch_verified(&argv, session_dir.path());
// Verify the patch contents - as otherwise we may have pulled contents
// from the wrong file (as we're using relative paths)
assert_eq!(
result,
MaybeApplyPatchVerified::Body(ApplyPatchAction {
changes: HashMap::from([(
session_dir.path().join(relative_path),
ApplyPatchFileChange::Update {
unified_diff: r#"@@ -1 +1 @@
-session directory content
+updated session directory content
"#
.to_string(),
move_path: None,
new_content: "updated session directory content\n".to_string(),
},
)]),
patch: argv[1].clone(),
cwd: session_dir.path().to_path_buf(),
})
);
}
#[test]
fn test_apply_patch_resolves_move_path_with_effective_cwd() {
let session_dir = tempdir().unwrap();
let worktree_rel = "alt";
let worktree_dir = session_dir.path().join(worktree_rel);
fs::create_dir_all(&worktree_dir).unwrap();
let source_name = "old.txt";
let dest_name = "renamed.txt";
let source_path = worktree_dir.join(source_name);
fs::write(&source_path, "before\n").unwrap();
let patch = wrap_patch(&format!(
r#"*** Update File: {source_name}
*** Move to: {dest_name}
@@
-before
+after"#
));
let shell_script = format!("cd {worktree_rel} && apply_patch <<'PATCH'\n{patch}\nPATCH");
let argv = vec!["bash".into(), "-lc".into(), shell_script];
let result = maybe_parse_apply_patch_verified(&argv, session_dir.path());
let action = match result {
MaybeApplyPatchVerified::Body(action) => action,
other => panic!("expected verified body, got {other:?}"),
};
assert_eq!(action.cwd, worktree_dir);
let change = action
.changes()
.get(&worktree_dir.join(source_name))
.expect("source file change present");
match change {
ApplyPatchFileChange::Update { move_path, .. } => {
assert_eq!(
move_path.as_deref(),
Some(worktree_dir.join(dest_name).as_path())
);
}
other => panic!("expected update change, got {other:?}"),
}
}
#[test]
fn test_apply_patch_fails_on_write_error() {
let dir = tempdir().unwrap();

View File

@@ -1,3 +0,0 @@
line1
naïve café ✅
line3

View File

@@ -1,3 +0,0 @@
line1
naïve café
line3

View File

@@ -1,7 +0,0 @@
*** Begin Patch
*** Update File: foo.txt
@@
line1
-naïve café
+naïve café ✅
*** End Patch

View File

@@ -36,7 +36,6 @@ codex-responses-api-proxy = { workspace = true }
codex-rmcp-client = { workspace = true }
codex-stdio-to-uds = { workspace = true }
codex-tui = { workspace = true }
codex-tui2 = { workspace = true }
ctor = { workspace = true }
libc = { workspace = true }
owo-colors = { workspace = true }

View File

@@ -136,9 +136,7 @@ async fn run_command_under_sandbox(
if let SandboxType::Windows = sandbox_type {
#[cfg(target_os = "windows")]
{
use codex_core::features::Feature;
use codex_windows_sandbox::run_windows_sandbox_capture;
use codex_windows_sandbox::run_windows_sandbox_capture_elevated;
let policy_str = serde_json::to_string(&config.sandbox_policy)?;
@@ -147,32 +145,18 @@ async fn run_command_under_sandbox(
let env_map = env.clone();
let command_vec = command.clone();
let base_dir = config.codex_home.clone();
let use_elevated = config.features.enabled(Feature::WindowsSandbox)
&& config.features.enabled(Feature::WindowsSandboxElevated);
// Preflight audit is invoked elsewhere at the appropriate times.
let res = tokio::task::spawn_blocking(move || {
if use_elevated {
run_windows_sandbox_capture_elevated(
policy_str.as_str(),
&sandbox_cwd,
base_dir.as_path(),
command_vec,
&cwd_clone,
env_map,
None,
)
} else {
run_windows_sandbox_capture(
policy_str.as_str(),
&sandbox_cwd,
base_dir.as_path(),
command_vec,
&cwd_clone,
env_map,
None,
)
}
run_windows_sandbox_capture(
policy_str.as_str(),
&sandbox_cwd,
base_dir.as_path(),
command_vec,
&cwd_clone,
env_map,
None,
)
})
.await;

View File

@@ -25,7 +25,6 @@ use codex_responses_api_proxy::Args as ResponsesApiProxyArgs;
use codex_tui::AppExitInfo;
use codex_tui::Cli as TuiCli;
use codex_tui::update_action::UpdateAction;
use codex_tui2 as tui2;
use owo_colors::OwoColorize;
use std::path::PathBuf;
use supports_color::Stream;
@@ -38,11 +37,6 @@ use crate::mcp_cmd::McpCli;
use codex_core::config::Config;
use codex_core::config::ConfigOverrides;
use codex_core::config::find_codex_home;
use codex_core::config::load_config_as_toml_with_cli_overrides;
use codex_core::features::Feature;
use codex_core::features::FeatureOverrides;
use codex_core::features::Features;
use codex_core::features::is_known_feature_key;
/// Codex CLI
@@ -450,7 +444,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
&mut interactive.config_overrides,
root_config_overrides.clone(),
);
let exit_info = run_interactive_tui(interactive, codex_linux_sandbox_exe).await?;
let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
handle_app_exit(exit_info)?;
}
Some(Subcommand::Exec(mut exec_cli)) => {
@@ -505,7 +499,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
all,
config_overrides,
);
let exit_info = run_interactive_tui(interactive, codex_linux_sandbox_exe).await?;
let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
handle_app_exit(exit_info)?;
}
Some(Subcommand::Login(mut login_cli)) => {
@@ -656,40 +650,6 @@ fn prepend_config_flags(
.splice(0..0, cli_config_overrides.raw_overrides);
}
/// Run the interactive Codex TUI, dispatching to either the legacy implementation or the
/// experimental TUI v2 shim based on feature flags resolved from config.
async fn run_interactive_tui(
interactive: TuiCli,
codex_linux_sandbox_exe: Option<PathBuf>,
) -> std::io::Result<AppExitInfo> {
if is_tui2_enabled(&interactive).await? {
let result = tui2::run_main(interactive.into(), codex_linux_sandbox_exe).await?;
Ok(result.into())
} else {
codex_tui::run_main(interactive, codex_linux_sandbox_exe).await
}
}
/// Returns `Ok(true)` when the resolved configuration enables the `tui2` feature flag.
///
/// This performs a lightweight config load (honoring the same precedence as the lower-level TUI
/// bootstrap: `$CODEX_HOME`, config.toml, profile, and CLI `-c` overrides) solely to decide which
/// TUI frontend to launch. The full configuration is still loaded later by the interactive TUI.
async fn is_tui2_enabled(cli: &TuiCli) -> std::io::Result<bool> {
let raw_overrides = cli.config_overrides.raw_overrides.clone();
let overrides_cli = codex_common::CliConfigOverrides { raw_overrides };
let cli_kv_overrides = overrides_cli
.parse_overrides()
.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidInput, e))?;
let codex_home = find_codex_home()?;
let config_toml = load_config_as_toml_with_cli_overrides(&codex_home, cli_kv_overrides).await?;
let config_profile = config_toml.get_config_profile(cli.config_profile.clone())?;
let overrides = FeatureOverrides::default();
let features = Features::from_config(&config_toml, &config_profile, overrides);
Ok(features.enabled(Feature::Tui2))
}
/// Build the final `TuiCli` for a `codex resume` invocation.
fn finalize_resume_interactive(
mut interactive: TuiCli,

View File

@@ -53,11 +53,11 @@ pub enum McpSubcommand {
Remove(RemoveArgs),
/// [experimental] Authenticate with a configured MCP server via OAuth.
/// Requires features.rmcp_client = true in config.toml.
/// Requires experimental_use_rmcp_client = true in config.toml.
Login(LoginArgs),
/// [experimental] Remove stored OAuth credentials for a server.
/// Requires features.rmcp_client = true in config.toml.
/// Requires experimental_use_rmcp_client = true in config.toml.
Logout(LogoutArgs),
}
@@ -285,7 +285,7 @@ async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Re
Ok(true) => {
if !config.features.enabled(Feature::RmcpClient) {
println!(
"MCP server supports login. Add `features.rmcp_client = true` \
"MCP server supports login. Add `experimental_use_rmcp_client = true` \
to your config.toml and run `codex mcp login {name}` to login."
);
} else {

View File

@@ -1,7 +1,24 @@
use std::ffi::OsStr;
/// Returns true if the current process is running under WSL.
pub use codex_core::env::is_wsl;
/// WSL-specific path helpers used by the updater logic.
///
/// See https://github.com/openai/codex/issues/6086.
pub fn is_wsl() -> bool {
#[cfg(target_os = "linux")]
{
if std::env::var_os("WSL_DISTRO_NAME").is_some() {
return true;
}
match std::fs::read_to_string("/proc/version") {
Ok(version) => version.to_lowercase().contains("microsoft"),
Err(_) => false,
}
}
#[cfg(not(target_os = "linux"))]
{
false
}
}
/// Convert a Windows absolute path (`C:\foo\bar` or `C:/foo/bar`) to a WSL mount path (`/mnt/c/foo/bar`).
/// Returns `None` if the input does not look like a Windows drive path.

View File

@@ -8,12 +8,7 @@ use tempfile::TempDir;
#[test]
fn execpolicy_check_matches_expected_json() -> 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"),
)?;
let policy_path = codex_home.path().join("policy.codexpolicy");
fs::write(
&policy_path,
r#"
@@ -29,7 +24,7 @@ prefix_rule(
.args([
"execpolicy",
"check",
"--rules",
"--policy",
policy_path
.to_str()
.expect("policy path should be valid UTF-8"),

View File

@@ -8,7 +8,6 @@ use codex_client::RequestTelemetry;
use codex_protocol::openai_models::ModelsResponse;
use http::HeaderMap;
use http::Method;
use http::header::ETAG;
use std::sync::Arc;
pub struct ModelsClient<T: HttpTransport, A: AuthProvider> {
@@ -60,23 +59,12 @@ impl<T: HttpTransport, A: AuthProvider> ModelsClient<T, A> {
)
.await?;
let header_etag = resp
.headers
.get(ETAG)
.and_then(|value| value.to_str().ok())
.map(ToString::to_string);
let ModelsResponse { models, etag } = serde_json::from_slice::<ModelsResponse>(&resp.body)
.map_err(|e| {
ApiError::Stream(format!(
"failed to decode models response: {e}; body: {}",
String::from_utf8_lossy(&resp.body)
))
})?;
let etag = header_etag.unwrap_or(etag);
Ok(ModelsResponse { models, etag })
serde_json::from_slice::<ModelsResponse>(&resp.body).map_err(|e| {
ApiError::Stream(format!(
"failed to decode models response: {e}; body: {}",
String::from_utf8_lossy(&resp.body)
))
})
}
}
@@ -98,36 +86,20 @@ mod tests {
use std::sync::Mutex;
use std::time::Duration;
#[derive(Clone)]
#[derive(Clone, Default)]
struct CapturingTransport {
last_request: Arc<Mutex<Option<Request>>>,
body: Arc<ModelsResponse>,
}
impl Default for CapturingTransport {
fn default() -> Self {
Self {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(ModelsResponse {
models: Vec::new(),
etag: String::new(),
}),
}
}
}
#[async_trait]
impl HttpTransport for CapturingTransport {
async fn execute(&self, req: Request) -> Result<Response, TransportError> {
*self.last_request.lock().unwrap() = Some(req);
let body = serde_json::to_vec(&*self.body).unwrap();
let mut headers = HeaderMap::new();
if !self.body.etag.is_empty() {
headers.insert(ETAG, self.body.etag.parse().unwrap());
}
Ok(Response {
status: StatusCode::OK,
headers,
headers: HeaderMap::new(),
body: body.into(),
})
}
@@ -166,10 +138,7 @@ mod tests {
#[tokio::test]
async fn appends_client_version_query() {
let response = ModelsResponse {
models: Vec::new(),
etag: String::new(),
};
let response = ModelsResponse { models: Vec::new() };
let transport = CapturingTransport {
last_request: Arc::new(Mutex::new(None)),
@@ -212,27 +181,15 @@ mod tests {
"display_name": "gpt-test",
"description": "desc",
"default_reasoning_level": "medium",
"supported_reasoning_levels": [{"effort": "low", "description": "low"}, {"effort": "medium", "description": "medium"}, {"effort": "high", "description": "high"}],
"supported_reasoning_levels": ["low", "medium", "high"],
"shell_type": "shell_command",
"visibility": "list",
"minimal_client_version": [0, 99, 0],
"supported_in_api": true,
"priority": 1,
"upgrade": null,
"base_instructions": null,
"supports_reasoning_summaries": false,
"support_verbosity": false,
"default_verbosity": null,
"apply_patch_tool_type": null,
"truncation_policy": {"mode": "bytes", "limit": 10_000},
"supports_parallel_tool_calls": false,
"context_window": null,
"reasoning_summary_format": "none",
"experimental_supported_tools": [],
"priority": 1
}))
.unwrap(),
],
etag: String::new(),
};
let transport = CapturingTransport {
@@ -256,31 +213,4 @@ mod tests {
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(),
etag: "\"abc\"".to_string(),
};
let transport = CapturingTransport {
last_request: Arc::new(Mutex::new(None)),
body: Arc::new(response),
};
let client = ModelsClient::new(
transport,
provider("https://example.com/api/codex"),
DummyAuth,
);
let result = client
.list_models("0.1.0", HeaderMap::new())
.await
.expect("request should succeed");
assert_eq!(result.models.len(), 0);
assert_eq!(result.etag, "\"abc\"");
}
}

View File

@@ -17,7 +17,6 @@ use codex_protocol::protocol::SessionSource;
use http::HeaderMap;
use serde_json::Value;
use std::sync::Arc;
use tracing::instrument;
pub struct ResponsesClient<T: HttpTransport, A: AuthProvider> {
streaming: StreamingClient<T, A>,
@@ -58,7 +57,6 @@ impl<T: HttpTransport, A: AuthProvider> ResponsesClient<T, A> {
self.stream(request.body, request.headers).await
}
#[instrument(level = "trace", skip_all, err)]
pub async fn stream_prompt(
&self,
model: &str,

View File

@@ -74,7 +74,7 @@ impl<'a> ChatRequestBuilder<'a> {
ResponseItem::CustomToolCallOutput { .. } => {}
ResponseItem::WebSearchCall { .. } => {}
ResponseItem::GhostSnapshot { .. } => {}
ResponseItem::Compaction { .. } => {}
ResponseItem::CompactionSummary { .. } => {}
}
}
@@ -303,7 +303,7 @@ impl<'a> ChatRequestBuilder<'a> {
ResponseItem::Reasoning { .. }
| ResponseItem::WebSearchCall { .. }
| ResponseItem::Other
| ResponseItem::Compaction { .. } => {
| ResponseItem::CompactionSummary { .. } => {
continue;
}
}

View File

@@ -10,9 +10,6 @@ use codex_protocol::openai_models::ModelInfo;
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;
use wiremock::Mock;
@@ -60,37 +57,16 @@ async fn models_client_hits_models_endpoint() {
description: Some("desc".to_string()),
default_reasoning_level: ReasoningEffort::Medium,
supported_reasoning_levels: vec![
ReasoningEffortPreset {
effort: ReasoningEffort::Low,
description: ReasoningEffort::Low.to_string(),
},
ReasoningEffortPreset {
effort: ReasoningEffort::Medium,
description: ReasoningEffort::Medium.to_string(),
},
ReasoningEffortPreset {
effort: ReasoningEffort::High,
description: ReasoningEffort::High.to_string(),
},
ReasoningEffort::Low,
ReasoningEffort::Medium,
ReasoningEffort::High,
],
shell_type: ConfigShellToolType::ShellCommand,
visibility: ModelVisibility::List,
minimal_client_version: ClientVersion(0, 1, 0),
supported_in_api: true,
priority: 1,
upgrade: None,
base_instructions: None,
supports_reasoning_summaries: false,
support_verbosity: false,
default_verbosity: None,
apply_patch_tool_type: None,
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"))

View File

@@ -10,7 +10,6 @@ bytes = { workspace = true }
eventsource-stream = { workspace = true }
futures = { workspace = true }
http = { workspace = true }
opentelemetry = { workspace = true }
rand = { workspace = true }
reqwest = { workspace = true, features = ["json", "stream"] }
serde = { workspace = true, features = ["derive"] }
@@ -18,11 +17,6 @@ serde_json = { workspace = true }
thiserror = { workspace = true }
tokio = { workspace = true, features = ["macros", "rt", "time", "sync"] }
tracing = { workspace = true }
tracing-opentelemetry = { workspace = true }
[lints]
workspace = true
[dev-dependencies]
opentelemetry_sdk = { workspace = true }
tracing-subscriber = { workspace = true }

View File

@@ -1,225 +0,0 @@
use http::Error as HttpError;
use opentelemetry::global;
use opentelemetry::propagation::Injector;
use reqwest::IntoUrl;
use reqwest::Method;
use reqwest::Response;
use reqwest::header::HeaderMap;
use reqwest::header::HeaderName;
use reqwest::header::HeaderValue;
use serde::Serialize;
use std::collections::HashMap;
use std::fmt::Display;
use std::time::Duration;
use tracing::Span;
use tracing_opentelemetry::OpenTelemetrySpanExt;
#[derive(Clone, Debug)]
pub struct CodexHttpClient {
inner: reqwest::Client,
}
impl CodexHttpClient {
pub fn new(inner: reqwest::Client) -> Self {
Self { inner }
}
pub fn get<U>(&self, url: U) -> CodexRequestBuilder
where
U: IntoUrl,
{
self.request(Method::GET, url)
}
pub fn post<U>(&self, url: U) -> CodexRequestBuilder
where
U: IntoUrl,
{
self.request(Method::POST, url)
}
pub fn request<U>(&self, method: Method, url: U) -> CodexRequestBuilder
where
U: IntoUrl,
{
let url_str = url.as_str().to_string();
CodexRequestBuilder::new(self.inner.request(method.clone(), url), method, url_str)
}
}
#[must_use = "requests are not sent unless `send` is awaited"]
#[derive(Debug)]
pub struct CodexRequestBuilder {
builder: reqwest::RequestBuilder,
method: Method,
url: String,
}
impl CodexRequestBuilder {
fn new(builder: reqwest::RequestBuilder, method: Method, url: String) -> Self {
Self {
builder,
method,
url,
}
}
fn map(self, f: impl FnOnce(reqwest::RequestBuilder) -> reqwest::RequestBuilder) -> Self {
Self {
builder: f(self.builder),
method: self.method,
url: self.url,
}
}
pub fn headers(self, headers: HeaderMap) -> Self {
self.map(|builder| builder.headers(headers))
}
pub fn header<K, V>(self, key: K, value: V) -> Self
where
HeaderName: TryFrom<K>,
<HeaderName as TryFrom<K>>::Error: Into<HttpError>,
HeaderValue: TryFrom<V>,
<HeaderValue as TryFrom<V>>::Error: Into<HttpError>,
{
self.map(|builder| builder.header(key, value))
}
pub fn bearer_auth<T>(self, token: T) -> Self
where
T: Display,
{
self.map(|builder| builder.bearer_auth(token))
}
pub fn timeout(self, timeout: Duration) -> Self {
self.map(|builder| builder.timeout(timeout))
}
pub fn json<T>(self, value: &T) -> Self
where
T: ?Sized + Serialize,
{
self.map(|builder| builder.json(value))
}
pub async fn send(self) -> Result<Response, reqwest::Error> {
let headers = trace_headers();
match self.builder.headers(headers).send().await {
Ok(response) => {
let request_ids = Self::extract_request_ids(&response);
tracing::debug!(
method = %self.method,
url = %self.url,
status = %response.status(),
request_ids = ?request_ids,
version = ?response.version(),
"Request completed"
);
Ok(response)
}
Err(error) => {
let status = error.status();
tracing::debug!(
method = %self.method,
url = %self.url,
status = status.map(|s| s.as_u16()),
error = %error,
"Request failed"
);
Err(error)
}
}
}
fn extract_request_ids(response: &Response) -> HashMap<String, String> {
["cf-ray", "x-request-id", "x-oai-request-id"]
.iter()
.filter_map(|&name| {
let header_name = HeaderName::from_static(name);
let value = response.headers().get(header_name)?;
let value = value.to_str().ok()?.to_owned();
Some((name.to_owned(), value))
})
.collect()
}
}
struct HeaderMapInjector<'a>(&'a mut HeaderMap);
impl<'a> Injector for HeaderMapInjector<'a> {
fn set(&mut self, key: &str, value: String) {
if let (Ok(name), Ok(val)) = (
HeaderName::from_bytes(key.as_bytes()),
HeaderValue::from_str(&value),
) {
self.0.insert(name, val);
}
}
}
fn trace_headers() -> HeaderMap {
let mut headers = HeaderMap::new();
global::get_text_map_propagator(|prop| {
prop.inject_context(
&Span::current().context(),
&mut HeaderMapInjector(&mut headers),
);
});
headers
}
#[cfg(test)]
mod tests {
use super::*;
use opentelemetry::propagation::Extractor;
use opentelemetry::propagation::TextMapPropagator;
use opentelemetry::trace::TraceContextExt;
use opentelemetry::trace::TracerProvider;
use opentelemetry_sdk::propagation::TraceContextPropagator;
use opentelemetry_sdk::trace::SdkTracerProvider;
use tracing::trace_span;
use tracing_subscriber::layer::SubscriberExt;
use tracing_subscriber::util::SubscriberInitExt;
#[test]
fn inject_trace_headers_uses_current_span_context() {
global::set_text_map_propagator(TraceContextPropagator::new());
let provider = SdkTracerProvider::builder().build();
let tracer = provider.tracer("test-tracer");
let subscriber =
tracing_subscriber::registry().with(tracing_opentelemetry::layer().with_tracer(tracer));
let _guard = subscriber.set_default();
let span = trace_span!("client_request");
let _entered = span.enter();
let span_context = span.context().span().span_context().clone();
let headers = trace_headers();
let extractor = HeaderMapExtractor(&headers);
let extracted = TraceContextPropagator::new().extract(&extractor);
let extracted_span = extracted.span();
let extracted_context = extracted_span.span_context();
assert!(extracted_context.is_valid());
assert_eq!(extracted_context.trace_id(), span_context.trace_id());
assert_eq!(extracted_context.span_id(), span_context.span_id());
}
struct HeaderMapExtractor<'a>(&'a HeaderMap);
impl<'a> Extractor for HeaderMapExtractor<'a> {
fn get(&self, key: &str) -> Option<&str> {
self.0.get(key).and_then(|value| value.to_str().ok())
}
fn keys(&self) -> Vec<&str> {
self.0.keys().map(HeaderName::as_str).collect()
}
}
}

View File

@@ -1,4 +1,3 @@
mod default_client;
mod error;
mod request;
mod retry;
@@ -6,8 +5,6 @@ mod sse;
mod telemetry;
mod transport;
pub use crate::default_client::CodexHttpClient;
pub use crate::default_client::CodexRequestBuilder;
pub use crate::error::StreamError;
pub use crate::error::TransportError;
pub use crate::request::Request;

View File

@@ -1,5 +1,3 @@
use crate::default_client::CodexHttpClient;
use crate::default_client::CodexRequestBuilder;
use crate::error::TransportError;
use crate::request::Request;
use crate::request::Response;
@@ -30,17 +28,15 @@ pub trait HttpTransport: Send + Sync {
#[derive(Clone, Debug)]
pub struct ReqwestTransport {
client: CodexHttpClient,
client: reqwest::Client,
}
impl ReqwestTransport {
pub fn new(client: reqwest::Client) -> Self {
Self {
client: CodexHttpClient::new(client),
}
Self { client }
}
fn build(&self, req: Request) -> Result<CodexRequestBuilder, TransportError> {
fn build(&self, req: Request) -> Result<reqwest::RequestBuilder, TransportError> {
let mut builder = self
.client
.request(

View File

@@ -4,10 +4,10 @@ use codex_core::config::Config;
use crate::sandbox_summary::summarize_sandbox_policy;
/// Build a list of key/value pairs summarizing the effective configuration.
pub fn create_config_summary_entries(config: &Config, model: &str) -> Vec<(&'static str, String)> {
pub fn create_config_summary_entries(config: &Config) -> Vec<(&'static str, String)> {
let mut entries = vec![
("workdir", config.cwd.display().to_string()),
("model", model.to_string()),
("model", config.model.clone()),
("provider", config.model_provider_id.clone()),
("approval", config.approval_policy.to_string()),
("sandbox", summarize_sandbox_policy(&config.sandbox_policy)),

View File

@@ -1,8 +1,8 @@
[package]
edition.workspace = true
license.workspace = true
name = "codex-core"
version.workspace = true
edition.workspace = true
license.workspace = true
[lib]
doctest = false
@@ -14,32 +14,31 @@ workspace = true
[dependencies]
anyhow = { workspace = true }
askama = { workspace = true }
async-channel = { workspace = true }
async-trait = { workspace = true }
base64 = { workspace = true }
chardetng = { workspace = true }
chrono = { workspace = true, features = ["serde"] }
codex-api = { workspace = true }
chardetng = { workspace = true }
codex-app-server-protocol = { workspace = true }
codex-apply-patch = { workspace = true }
codex-async-utils = { workspace = true }
codex-client = { workspace = true }
codex-api = { workspace = true }
codex-execpolicy = { workspace = true }
codex-file-search = { workspace = true }
codex-git = { workspace = true }
codex-keyring-store = { workspace = true }
codex-otel = { workspace = true }
codex-otel = { workspace = true, features = ["otel"] }
codex-protocol = { workspace = true }
codex-rmcp-client = { workspace = true }
codex-utils-absolute-path = { workspace = true }
codex-utils-pty = { workspace = true }
codex-utils-readiness = { workspace = true }
codex-utils-string = { workspace = true }
codex-windows-sandbox = { package = "codex-windows-sandbox", path = "../windows-sandbox-rs" }
dirs = { workspace = true }
dunce = { workspace = true }
encoding_rs = { workspace = true }
env-flags = { workspace = true }
encoding_rs = { workspace = true }
eventsource-stream = { workspace = true }
futures = { workspace = true }
http = { workspace = true }
@@ -47,10 +46,8 @@ indexmap = { workspace = true }
keyring = { workspace = true, features = ["crypto-rust"] }
libc = { workspace = true }
mcp-types = { workspace = true }
once_cell = { workspace = true }
os_info = { workspace = true }
rand = { workspace = true }
regex = { workspace = true }
regex-lite = { workspace = true }
reqwest = { workspace = true, features = ["json", "stream"] }
serde = { workspace = true, features = ["derive"] }
@@ -61,6 +58,9 @@ sha2 = { workspace = true }
shlex = { workspace = true }
similar = { workspace = true }
strum_macros = { workspace = true }
url = { workspace = true }
once_cell = { workspace = true }
regex = { workspace = true }
tempfile = { workspace = true }
test-case = "3.3.1"
test-log = { workspace = true }
@@ -84,7 +84,6 @@ toml_edit = { workspace = true }
tracing = { workspace = true, features = ["log"] }
tree-sitter = { workspace = true }
tree-sitter-bash = { workspace = true }
url = { workspace = true }
uuid = { workspace = true, features = ["serde", "v4", "v5"] }
which = { workspace = true }
wildmatch = { workspace = true }
@@ -95,9 +94,9 @@ test-support = []
[target.'cfg(target_os = "linux")'.dependencies]
keyring = { workspace = true, features = ["linux-native-async-persistent"] }
landlock = { workspace = true }
seccompiler = { workspace = true }
keyring = { workspace = true, features = ["linux-native-async-persistent"] }
[target.'cfg(target_os = "macos")'.dependencies]
core-foundation = "0.9"
@@ -132,7 +131,6 @@ pretty_assertions = { workspace = true }
serial_test = { workspace = true }
tempfile = { workspace = true }
tokio-test = { workspace = true }
tracing-subscriber = { workspace = true }
tracing-test = { workspace = true, features = ["no-env-filter"] }
walkdir = { workspace = true }
wiremock = { workspace = true }

View File

@@ -48,7 +48,7 @@ When you are running with `approval_policy == on-request`, and sandboxing enable
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `sandbox_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
- (for all of these, you should weigh alternative paths that do not require approval)
@@ -59,8 +59,8 @@ You will be told what filesystem sandboxing, network sandboxing, and approval mo
Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
When requesting approval to execute a command that will require escalated privileges:
- Provide the `sandbox_permissions` parameter with the value `"require_escalated"`
- Include a short, 1 sentence explanation for why you need escalated permissions in the justification parameter
- Provide the `with_escalated_permissions` parameter with the boolean value true
- Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
## Special user requests

View File

@@ -182,7 +182,7 @@ When you are running with `approval_policy == on-request`, and sandboxing enable
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `sandbox_permissions` and `justification` parameters. Within this harness, prefer requesting approval via the tool over asking in natural language.
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters. Within this harness, prefer requesting approval via the tool over asking in natural language.
- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
- (for all of these, you should weigh alternative paths that do not require approval)
@@ -193,8 +193,8 @@ You will be told what filesystem sandboxing, network sandboxing, and approval mo
Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
When requesting approval to execute a command that will require escalated privileges:
- Provide the `sandbox_permissions` parameter with the value `"require_escalated"`
- Include a short, 1 sentence explanation for why you need escalated permissions in the justification parameter
- Provide the `with_escalated_permissions` parameter with the boolean value true
- Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
## Validating your work
@@ -319,7 +319,7 @@ For casual greetings, acknowledgements, or other one-off conversational messages
When using the shell, you must adhere to the following guidelines:
- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
- Do not use python scripts to attempt to output larger chunks of a file.
- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
## apply_patch

View File

@@ -1,335 +0,0 @@
You are GPT-5.2 running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
Your capabilities:
- Receive user prompts and other context provided by the harness, such as files in the workspace.
- Communicate with the user by streaming thinking & responses, and by making & updating plans.
- Emit function calls to run terminal commands and apply patches. Depending on how this specific run is configured, you can request that these function calls be escalated to the user for approval before running. More on this in the "Sandbox and approvals" section.
Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
# How you work
## Personality
Your default personality and tone is concise, direct, and friendly. You communicate efficiently, always keeping the user clearly informed about ongoing actions without unnecessary detail. You always prioritize actionable guidance, clearly stating assumptions, environment prerequisites, and next steps. Unless explicitly asked, you avoid excessively verbose explanations about your work.
## AGENTS.md spec
- Repos often contain AGENTS.md files. These files can appear anywhere within the repository.
- These files are a way for humans to give you (the agent) instructions or tips for working within the container.
- Some examples might be: coding conventions, info about how code is organized, or instructions for how to run or test code.
- Instructions in AGENTS.md files:
- The scope of an AGENTS.md file is the entire directory tree rooted at the folder that contains it.
- For every file you touch in the final patch, you must obey instructions in any AGENTS.md file whose scope includes that file.
- Instructions about code style, structure, naming, etc. apply only to code within the AGENTS.md file's scope, unless the file states otherwise.
- More-deeply-nested AGENTS.md files take precedence in the case of conflicting instructions.
- Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
## Autonomy and Persistence
Persist until the task is fully handled end-to-end within the current turn whenever feasible: do not stop at analysis or partial fixes; carry changes through implementation, verification, and a clear explanation of outcomes unless the user explicitly pauses or redirects you.
Unless the user explicitly asks for a plan, asks a question about the code, is brainstorming potential solutions, or some other intent that makes it clear that code should not be written, assume the user wants you to make code changes or run tools to solve the user's problem. In these cases, it's bad to output your proposed solution in a message, you should go ahead and actually implement the change. If you encounter challenges or blockers, you should attempt to resolve them yourself.
## Responsiveness
## Planning
You have access to an `update_plan` tool which tracks steps and progress and renders them to the user. Using the tool helps demonstrate that you've understood the task and convey how you're approaching it. Plans can help to make complex, ambiguous, or multi-phase work clearer and more collaborative for the user. A good plan should break the task into meaningful, logically ordered steps that are easy to verify as you go.
Note that plans are not for padding out simple work with filler steps or stating the obvious. The content of your plan should not involve doing anything that you aren't capable of doing (i.e. don't try to test things that you can't test). Do not use plans for simple or single-step queries that you can just do or answer immediately.
Do not repeat the full contents of the plan after an `update_plan` call — the harness already displays it. Instead, summarize the change made and highlight any important context or next step.
Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
Maintain statuses in the tool: exactly one item in_progress at a time; mark items complete when done; post timely status transitions. Do not jump an item from pending to completed: always set it to in_progress first. Do not batch-complete multiple items after the fact. Finish with all items completed or explicitly canceled/deferred before ending the turn. Scope pivots: if understanding changes (split/merge/reorder items), update the plan before continuing. Do not let the plan go stale while coding.
Use a plan when:
- The task is non-trivial and will require multiple actions over a long time horizon.
- There are logical phases or dependencies where sequencing matters.
- The work has ambiguity that benefits from outlining high-level goals.
- You want intermediate checkpoints for feedback and validation.
- When the user asked you to do more than one thing in a single prompt
- The user has asked you to use the plan tool (aka "TODOs")
- You generate additional steps while working, and plan to do them before yielding to the user
### Examples
**High-quality plans**
Example 1:
1. Add CLI entry with file args
2. Parse Markdown via CommonMark library
3. Apply semantic HTML template
4. Handle code blocks, images, links
5. Add error handling for invalid files
Example 2:
1. Define CSS variables for colors
2. Add toggle with localStorage state
3. Refactor components to use variables
4. Verify all views for readability
5. Add smooth theme-change transition
Example 3:
1. Set up Node.js + WebSocket server
2. Add join/leave broadcast events
3. Implement messaging with timestamps
4. Add usernames + mention highlighting
5. Persist messages in lightweight DB
6. Add typing indicators + unread count
**Low-quality plans**
Example 1:
1. Create CLI tool
2. Add Markdown parser
3. Convert to HTML
Example 2:
1. Add dark mode toggle
2. Save preference
3. Make styles look good
Example 3:
1. Create single-file HTML game
2. Run quick sanity check
3. Summarize usage instructions
If you need to write a plan, only write high quality plans, not low quality ones.
## Task execution
You are a coding agent. You must keep going until the query or task is completely resolved, before ending your turn and yielding back to the user. Persist until the task is fully handled end-to-end within the current turn whenever feasible and persevere even when function calls fail. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
You MUST adhere to the following criteria when solving queries:
- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
- Analyzing code for vulnerabilities is allowed.
- Showing user code and tool call details is allowed.
- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`). This is a FREEFORM tool, so do not wrap the patch in JSON.
If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
- Fix the problem at the root cause rather than applying surface-level patches, when possible.
- Avoid unneeded complexity in your solution.
- Do not attempt to fix unrelated bugs or broken tests. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
- Update documentation as necessary.
- Keep changes consistent with the style of the existing codebase. Changes should be minimal and focused on the task.
- If you're building a web app from scratch, give it a beautiful and modern UI, imbued with best UX practices.
- Use `git log` and `git blame` to search the history of the codebase if additional context is required.
- NEVER add copyright or license headers unless specifically requested.
- Do not waste tokens by re-reading files after calling `apply_patch` on them. The tool call will fail if it didn't work. The same goes for making folders, deleting folders, etc.
- Do not `git commit` your changes or create new git branches unless explicitly requested.
- Do not add inline comments within code unless explicitly requested.
- Do not use one-letter variable names unless explicitly requested.
- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
## Codex CLI harness, sandboxing, and approvals
The Codex CLI harness supports several different configurations for sandboxing and escalation approvals that the user can choose from.
Filesystem sandboxing defines which files can be read or written. The options for `sandbox_mode` are:
- **read-only**: The sandbox only permits reading files.
- **workspace-write**: The sandbox permits reading files, and editing files in `cwd` and `writable_roots`. Editing files in other directories requires approval.
- **danger-full-access**: No filesystem sandboxing - all commands are permitted.
Network sandboxing defines whether network can be accessed without approval. Options for `network_access` are:
- **restricted**: Requires approval
- **enabled**: No approval needed
Approvals are your mechanism to get user consent to run shell commands without the sandbox. Possible configuration options for `approval_policy` are
- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for escalating in the tool definition.)
- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is paired with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
When you are running with `approval_policy == on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `sandbox_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
- (for all of these, you should weigh alternative paths that do not require approval)
When `sandbox_mode` is set to read-only, you'll need to request approval for any command that isn't a read.
You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing enabled, and approval on-failure.
Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
When requesting approval to execute a command that will require escalated privileges:
- Provide the `sandbox_permissions` parameter with the value `"require_escalated"`
- Include a short, 1 sentence explanation for why you need escalated permissions in the justification parameter
## Validating your work
If the codebase has tests, or the ability to build or run tests, consider using them to verify changes once your work is complete.
When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
Similarly, once you're confident in correctness, you can suggest or use formatting commands to ensure that your code is well formatted. If there are issues you can iterate up to 3 times to get formatting right, but if you still can't manage it's better to save the user time and present them a correct solution where you call out the formatting in your final message. If the codebase does not have a formatter configured, do not add one.
For all of testing, running, building, and formatting, do not attempt to fix unrelated bugs. It is not your responsibility to fix them. (You may mention them to the user in your final message though.)
Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
- When running in non-interactive approval modes like **never** or **on-failure**, you can proactively run tests, lint and do whatever you need to ensure you've completed the task. If you are unable to run tests, you must still do your utmost best to complete the task.
- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
## Ambition vs. precision
For tasks that have no prior context (i.e. the user is starting something brand new), you should feel free to be ambitious and demonstrate creativity with your implementation.
If you're operating in an existing codebase, you should make sure you do exactly what the user asks with surgical precision. Treat the surrounding codebase with respect, and don't overstep (i.e. changing filenames or variables unnecessarily). You should balance being sufficiently ambitious and proactive when completing tasks of this nature.
You should use judicious initiative to decide on the right level of detail and complexity to deliver based on the user's needs. This means showing good judgment that you're capable of doing the right extras without gold-plating. This might be demonstrated by high-value, creative touches when scope of the task is vague; while being surgical and targeted when scope is tightly specified.
## Presenting your work
Your final message should read naturally, like an update from a concise teammate. For casual conversation, brainstorming tasks, or quick questions from the user, respond in a friendly, conversational tone. You should ask questions, suggest ideas, and adapt to the users style. If you've finished a large amount of work, when describing what you've done to the user, you should follow the final answer formatting guidelines to communicate substantive changes. You don't need to add structured formatting for one-word answers, greetings, or purely conversational exchanges.
You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
The user is working on the same computer as you, and has access to your work. As such there's no need to show the contents of files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If theres something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
Brevity is very important as a default. You should be very concise (i.e. no more than 10 lines), but can relax this requirement for tasks where additional detail and comprehensiveness is important for the user's understanding.
### Final answer structure and style guidelines
You are producing plain text that will later be styled by the CLI. Follow these rules exactly. Formatting should make results easy to scan, but not feel mechanical. Use judgment to decide how much structure adds value.
**Section Headers**
- Use only when they improve clarity — they are not mandatory for every answer.
- Choose descriptive names that fit the content
- Keep headers short (13 words) and in `**Title Case**`. Always start headers with `**` and end with `**`
- Leave no blank line before the first bullet under a header.
- Section headers should only be used where they genuinely improve scanability; avoid fragmenting the answer.
**Bullets**
- Use `-` followed by a space for every bullet.
- Merge related points when possible; avoid a bullet for every trivial detail.
- Keep bullets to one line unless breaking for clarity is unavoidable.
- Group into short lists (46 bullets) ordered by importance.
- Use consistent keyword phrasing and formatting across sections.
**Monospace**
- Wrap all commands, file paths, env vars, code identifiers, and code samples in backticks (`` `...` ``).
- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
- Never mix monospace and bold markers; choose one based on whether its a keyword (`**`) or inline code/path (`` ` ``).
**File References**
When referencing files in your response, make sure to include the relevant start line and always follow the below rules:
* Use inline code to make file paths clickable.
* Each reference should have a stand alone path. Even if it's the same file.
* Accepted: absolute, workspacerelative, a/ or b/ diff prefixes, or bare filename/suffix.
* Line/column (1based, optional): :line[:column] or #Lline[Ccolumn] (column defaults to 1).
* Do not use URIs like file://, vscode://, or https://.
* Do not provide range of lines
* Examples: src/app.ts, src/app.ts:42, b/server/index.js#L10, C:\repo\project\main.rs:12:5
**Structure**
- Place related bullets together; dont mix unrelated concepts in the same section.
- Order sections from general → specific → supporting info.
- For subsections (e.g., “Binaries” under “Rust Workspace”), introduce with a bolded keyword bullet, then list items under it.
- Match structure to complexity:
- Multi-part or detailed results → use clear headers and grouped bullets.
- Simple results → minimal headers, possibly just a short list or paragraph.
**Tone**
- Keep the voice collaborative and natural, like a coding partner handing off work.
- Be concise and factual — no filler or conversational commentary and avoid unnecessary repetition
- Use present tense and active voice (e.g., “Runs tests” not “This will run tests”).
- Keep descriptions self-contained; dont refer to “above” or “below”.
- Use parallel structure in lists for consistency.
**Verbosity**
- Final answer compactness rules (enforced):
- Tiny/small single-file change (≤ ~10 lines): 25 sentences or ≤3 bullets. No headings. 01 short snippet (≤3 lines) only if essential.
- Medium change (single area or a few files): ≤6 bullets or 610 sentences. At most 12 short snippets total (≤8 lines each).
- Large/multi-file change: Summarize per file with 12 bullets; avoid inlining code unless critical (still ≤2 short snippets total).
- Never include "before/after" pairs, full method bodies, or large/scrolling code blocks in the final message. Prefer referencing file/symbol names instead.
**Dont**
- Dont use literal words “bold” or “monospace” in the content.
- Dont nest bullets or create deep hierarchies.
- Dont output ANSI escape codes directly — the CLI renderer applies them.
- Dont cram unrelated keywords into a single bullet; split for clarity.
- Dont let keyword lists run long — wrap or reformat for scanability.
Generally, ensure your final answers adapt their shape and depth to the request. For example, answers to code explanations should have a precise, structured explanation with code references that answer the question directly. For tasks with a simple implementation, lead with the outcome and supplement only with whats needed for clarity. Larger changes can be presented as a logical walkthrough of your approach, grouping related steps, explaining rationale where it adds value, and highlighting next actions to accelerate the user. Your answers should provide the right level of detail while being easily scannable.
For casual greetings, acknowledgements, or other one-off conversational messages that are not delivering substantive information or structured results, respond naturally without section headers or bullet formatting.
# Tool Guidelines
## Shell commands
When using the shell, you must adhere to the following guidelines:
- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
- Do not use python scripts to attempt to output larger chunks of a file.
- Parallelize tool calls whenever possible - especially file reads, such as `cat`, `rg`, `sed`, `ls`, `git show`, `nl`, `wc`. Use `multi_tool_use.parallel` to parallelize tool calls and only this.
## apply_patch
Use the `apply_patch` tool to edit files. Your patch language is a strippeddown, fileoriented diff format designed to be easy to parse and safe to apply. You can think of it as a highlevel envelope:
*** Begin Patch
[ one or more file sections ]
*** End Patch
Within that envelope, you get a sequence of file operations.
You MUST include a header to specify the action you are taking.
Each operation starts with one of three headers:
*** Add File: <path> - create a new file. Every following line is a + line (the initial contents).
*** Delete File: <path> - remove an existing file. Nothing follows.
*** Update File: <path> - patch an existing file in place (optionally with a rename).
Example patch:
```
*** Begin Patch
*** Add File: hello.txt
+Hello world
*** Update File: src/app.py
*** Move to: src/main.py
@@ def greet():
-print("Hi")
+print("Hello, world!")
*** Delete File: obsolete.txt
*** End Patch
```
It is important to remember:
- You must include a header with your intended action (Add/Delete/Update)
- You must prefix new lines with `+` even when creating a new file
## `update_plan`
A tool named `update_plan` is available to you. You can use it to keep an uptodate, stepbystep plan for the task.
To create a new plan, call `update_plan` with a short list of 1sentence steps (no more than 5-7 words each) with a `status` for each step (`pending`, `in_progress`, or `completed`).
When steps have been completed, use `update_plan` to mark each finished step as `completed` and the next step you are working on as `in_progress`. There should always be exactly one `in_progress` step until everything is done. You can mark multiple items as complete in a single `update_plan` call.
If all steps are complete, ensure you call `update_plan` to mark all steps as `completed`.

View File

@@ -48,7 +48,7 @@ When you are running with `approval_policy == on-request`, and sandboxing enable
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `sandbox_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters - do not message the user before requesting approval for the command.
- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
- (for all of these, you should weigh alternative paths that do not require approval)
@@ -59,8 +59,8 @@ You will be told what filesystem sandboxing, network sandboxing, and approval mo
Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
When requesting approval to execute a command that will require escalated privileges:
- Provide the `sandbox_permissions` parameter with the value `"require_escalated"`
- Include a short, 1 sentence explanation for why you need escalated permissions in the justification parameter
- Provide the `with_escalated_permissions` parameter with the boolean value true
- Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
## Special user requests

View File

@@ -297,7 +297,7 @@ For casual greetings, acknowledgements, or other one-off conversational messages
When using the shell, you must adhere to the following guidelines:
- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
- Do not use python scripts to attempt to output larger chunks of a file.
- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
## `update_plan`

View File

@@ -23,6 +23,7 @@ pub use crate::auth::storage::AuthDotJson;
use crate::auth::storage::AuthStorageBackend;
use crate::auth::storage::create_auth_storage;
use crate::config::Config;
use crate::default_client::CodexHttpClient;
use crate::error::RefreshTokenFailedError;
use crate::error::RefreshTokenFailedReason;
use crate::token_data::KnownPlan as InternalKnownPlan;
@@ -30,13 +31,8 @@ use crate::token_data::PlanType as InternalPlanType;
use crate::token_data::TokenData;
use crate::token_data::parse_id_token;
use crate::util::try_parse_error_message;
use codex_client::CodexHttpClient;
use codex_protocol::account::PlanType as AccountPlanType;
#[cfg(any(test, feature = "test-support"))]
use once_cell::sync::Lazy;
use serde_json::Value;
#[cfg(any(test, feature = "test-support"))]
use tempfile::TempDir;
use thiserror::Error;
#[derive(Debug, Clone)]
@@ -66,9 +62,6 @@ const REFRESH_TOKEN_UNKNOWN_MESSAGE: &str =
const REFRESH_TOKEN_URL: &str = "https://auth.openai.com/oauth/token";
pub const REFRESH_TOKEN_URL_OVERRIDE_ENV_VAR: &str = "CODEX_REFRESH_TOKEN_URL_OVERRIDE";
#[cfg(any(test, feature = "test-support"))]
static TEST_AUTH_TEMP_DIRS: Lazy<Mutex<Vec<TempDir>>> = Lazy::new(|| Mutex::new(Vec::new()));
#[derive(Debug, Error)]
pub enum RefreshTokenError {
#[error("{0}")]
@@ -1095,31 +1088,11 @@ impl AuthManager {
}
}
#[cfg(any(test, feature = "test-support"))]
#[expect(clippy::expect_used)]
/// Create an AuthManager with a specific CodexAuth, for testing only.
pub fn from_auth_for_testing(auth: CodexAuth) -> Arc<Self> {
let cached = CachedAuth { auth: Some(auth) };
let temp_dir = tempfile::tempdir().expect("temp codex home");
let codex_home = temp_dir.path().to_path_buf();
TEST_AUTH_TEMP_DIRS
.lock()
.expect("lock test codex homes")
.push(temp_dir);
Arc::new(Self {
codex_home,
inner: RwLock::new(cached),
enable_codex_api_key_env: false,
auth_credentials_store_mode: AuthCredentialsStoreMode::File,
})
}
#[cfg(any(test, feature = "test-support"))]
/// Create an AuthManager with a specific CodexAuth and codex home, for testing only.
pub fn from_auth_for_testing_with_home(auth: CodexAuth, codex_home: PathBuf) -> Arc<Self> {
let cached = CachedAuth { auth: Some(auth) };
Arc::new(Self {
codex_home,
codex_home: PathBuf::new(),
inner: RwLock::new(cached),
enable_codex_api_key_env: false,
auth_credentials_store_mode: AuthCredentialsStoreMode::File,
@@ -1131,10 +1104,6 @@ impl AuthManager {
self.inner.read().ok().and_then(|c| c.auth.clone())
}
pub fn codex_home(&self) -> &Path {
&self.codex_home
}
/// Force a reload of the auth information from auth.json. Returns
/// whether the auth value changed.
pub fn reload(&self) -> bool {

View File

@@ -18,7 +18,7 @@ use codex_api::common::Reasoning;
use codex_api::create_text_param_for_request;
use codex_api::error::ApiError;
use codex_app_server_protocol::AuthMode;
use codex_otel::otel_manager::OtelManager;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
use codex_protocol::models::ResponseItem;
@@ -48,6 +48,7 @@ use crate::error::Result;
use crate::flags::CODEX_RS_SSE_FIXTURE;
use crate::model_provider_info::ModelProviderInfo;
use crate::model_provider_info::WireApi;
use crate::openai_model_info::get_model_info;
use crate::openai_models::model_family::ModelFamily;
use crate::tools::spec::create_tools_json_for_chat_completions_api;
use crate::tools::spec::create_tools_json_for_responses_api;
@@ -57,7 +58,7 @@ pub struct ModelClient {
config: Arc<Config>,
auth_manager: Option<Arc<AuthManager>>,
model_family: ModelFamily,
otel_manager: OtelManager,
otel_event_manager: OtelEventManager,
provider: ModelProviderInfo,
conversation_id: ConversationId,
effort: Option<ReasoningEffortConfig>,
@@ -71,7 +72,7 @@ impl ModelClient {
config: Arc<Config>,
auth_manager: Option<Arc<AuthManager>>,
model_family: ModelFamily,
otel_manager: OtelManager,
otel_event_manager: OtelEventManager,
provider: ModelProviderInfo,
effort: Option<ReasoningEffortConfig>,
summary: ReasoningSummaryConfig,
@@ -82,7 +83,7 @@ impl ModelClient {
config,
auth_manager,
model_family,
otel_manager,
otel_event_manager,
provider,
conversation_id,
effort,
@@ -94,11 +95,19 @@ impl ModelClient {
pub fn get_model_context_window(&self) -> Option<i64> {
let model_family = self.get_model_family();
let effective_context_window_percent = model_family.effective_context_window_percent;
model_family
.context_window
self.config
.model_context_window
.or_else(|| get_model_info(&model_family).map(|info| info.context_window))
.map(|w| w.saturating_mul(effective_context_window_percent) / 100)
}
pub fn get_auto_compact_token_limit(&self) -> Option<i64> {
let model_family = self.get_model_family();
self.config.model_auto_compact_token_limit.or_else(|| {
get_model_info(&model_family).and_then(|info| info.auto_compact_token_limit)
})
}
pub fn config(&self) -> Arc<Config> {
Arc::clone(&self.config)
}
@@ -121,12 +130,12 @@ impl ModelClient {
if self.config.show_raw_agent_reasoning {
Ok(map_response_stream(
api_stream.streaming_mode(),
self.otel_manager.clone(),
self.otel_event_manager.clone(),
))
} else {
Ok(map_response_stream(
api_stream.aggregate(),
self.otel_manager.clone(),
self.otel_event_manager.clone(),
))
}
}
@@ -166,7 +175,7 @@ impl ModelClient {
let stream_result = client
.stream_prompt(
&self.get_model(),
&self.config.model,
&api_prompt,
Some(conversation_id.clone()),
Some(session_source.clone()),
@@ -195,7 +204,7 @@ impl ModelClient {
warn!(path, "Streaming from fixture");
let stream = codex_api::stream_from_fixture(path, self.provider.stream_idle_timeout())
.map_err(map_api_error)?;
return Ok(map_response_stream(stream, self.otel_manager.clone()));
return Ok(map_response_stream(stream, self.otel_event_manager.clone()));
}
let auth_manager = self.auth_manager.clone();
@@ -206,11 +215,7 @@ impl ModelClient {
let reasoning = if model_family.supports_reasoning_summaries {
Some(Reasoning {
effort: self.effort.or(model_family.default_reasoning_effort),
summary: if self.summary == ReasoningSummaryConfig::None {
None
} else {
Some(self.summary)
},
summary: Some(self.summary),
})
} else {
None
@@ -264,12 +269,12 @@ impl ModelClient {
};
let stream_result = client
.stream_prompt(&self.get_model(), &api_prompt, options)
.stream_prompt(&self.config.model, &api_prompt, options)
.await;
match stream_result {
Ok(stream) => {
return Ok(map_response_stream(stream, self.otel_manager.clone()));
return Ok(map_response_stream(stream, self.otel_event_manager.clone()));
}
Err(ApiError::Transport(TransportError::Http { status, .. }))
if status == StatusCode::UNAUTHORIZED =>
@@ -286,8 +291,8 @@ impl ModelClient {
self.provider.clone()
}
pub fn get_otel_manager(&self) -> OtelManager {
self.otel_manager.clone()
pub fn get_otel_event_manager(&self) -> OtelEventManager {
self.otel_event_manager.clone()
}
pub fn get_session_source(&self) -> SessionSource {
@@ -296,7 +301,7 @@ impl ModelClient {
/// Returns the currently configured model slug.
pub fn get_model(&self) -> String {
self.get_model_family().get_model_slug().to_string()
self.config.model.clone()
}
/// Returns the currently configured model family.
@@ -341,7 +346,7 @@ impl ModelClient {
.get_full_instructions(&self.get_model_family())
.into_owned();
let payload = ApiCompactionInput {
model: &self.get_model(),
model: &self.config.model,
input: &prompt.input,
instructions: &instructions,
};
@@ -371,7 +376,7 @@ impl ModelClient {
impl ModelClient {
/// Builds request and SSE telemetry for streaming API calls (Chat/Responses).
fn build_streaming_telemetry(&self) -> (Arc<dyn RequestTelemetry>, Arc<dyn SseTelemetry>) {
let telemetry = Arc::new(ApiTelemetry::new(self.otel_manager.clone()));
let telemetry = Arc::new(ApiTelemetry::new(self.otel_event_manager.clone()));
let request_telemetry: Arc<dyn RequestTelemetry> = telemetry.clone();
let sse_telemetry: Arc<dyn SseTelemetry> = telemetry;
(request_telemetry, sse_telemetry)
@@ -379,7 +384,7 @@ impl ModelClient {
/// Builds request telemetry for unary API calls (e.g., Compact endpoint).
fn build_request_telemetry(&self) -> Arc<dyn RequestTelemetry> {
let telemetry = Arc::new(ApiTelemetry::new(self.otel_manager.clone()));
let telemetry = Arc::new(ApiTelemetry::new(self.otel_event_manager.clone()));
let request_telemetry: Arc<dyn RequestTelemetry> = telemetry;
request_telemetry
}
@@ -396,7 +401,7 @@ fn build_api_prompt(prompt: &Prompt, instructions: String, tools_json: Vec<Value
}
}
fn map_response_stream<S>(api_stream: S, otel_manager: OtelManager) -> ResponseStream
fn map_response_stream<S>(api_stream: S, otel_event_manager: OtelEventManager) -> ResponseStream
where
S: futures::Stream<Item = std::result::Result<ResponseEvent, ApiError>>
+ Unpin
@@ -404,6 +409,7 @@ where
+ 'static,
{
let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
let manager = otel_event_manager;
tokio::spawn(async move {
let mut logged_error = false;
@@ -415,7 +421,7 @@ where
token_usage,
}) => {
if let Some(usage) = &token_usage {
otel_manager.sse_event_completed(
manager.sse_event_completed(
usage.input_tokens,
usage.output_tokens,
Some(usage.cached_input_tokens),
@@ -442,7 +448,7 @@ where
Err(err) => {
let mapped = map_api_error(err);
if !logged_error {
otel_manager.see_event_completed_failed(&mapped);
manager.see_event_completed_failed(&mapped);
logged_error = true;
}
if tx_event.send(Err(mapped)).await.is_err() {
@@ -496,12 +502,12 @@ fn map_unauthorized_status(status: StatusCode) -> CodexErr {
}
struct ApiTelemetry {
otel_manager: OtelManager,
otel_event_manager: OtelEventManager,
}
impl ApiTelemetry {
fn new(otel_manager: OtelManager) -> Self {
Self { otel_manager }
fn new(otel_event_manager: OtelEventManager) -> Self {
Self { otel_event_manager }
}
}
@@ -514,7 +520,7 @@ impl RequestTelemetry for ApiTelemetry {
duration: Duration,
) {
let error_message = error.map(std::string::ToString::to_string);
self.otel_manager.record_api_request(
self.otel_event_manager.record_api_request(
attempt,
status.map(|s| s.as_u16()),
error_message.as_deref(),
@@ -532,6 +538,6 @@ impl SseTelemetry for ApiTelemetry {
>,
duration: Duration,
) {
self.otel_manager.log_sse_event(result, duration);
self.otel_event_manager.log_sse_event(result, duration);
}
}

View File

@@ -252,15 +252,13 @@ impl Stream for ResponseStream {
#[cfg(test)]
mod tests {
use crate::openai_models::model_family::find_family_for_model;
use codex_api::ResponsesApiRequest;
use codex_api::common::OpenAiVerbosity;
use codex_api::common::TextControls;
use codex_api::create_text_param_for_request;
use pretty_assertions::assert_eq;
use crate::config::test_config;
use crate::openai_models::models_manager::ModelsManager;
use super::*;
struct InstructionsTestCase {
@@ -311,9 +309,7 @@ mod tests {
},
];
for test_case in test_cases {
let config = test_config();
let model_family =
ModelsManager::construct_model_family_offline(test_case.slug, &config);
let model_family = find_family_for_model(test_case.slug);
let expected = if test_case.expects_apply_patch_instructions {
format!(
"{}\n{}",

File diff suppressed because it is too large Load Diff

View File

@@ -49,7 +49,6 @@ pub(crate) async fn run_codex_conversation_interactive(
config,
auth_manager,
models_manager,
Arc::clone(&parent_session.services.skills_manager),
initial_history.unwrap_or(InitialHistory::New),
SessionSource::SubAgent(SubAgentSource::Review),
)
@@ -281,6 +280,7 @@ async fn handle_exec_approval(
event.command,
event.cwd,
event.reason,
event.risk,
event.proposed_execpolicy_amendment,
);
let decision = await_approval_with_cancel(

View File

@@ -47,47 +47,24 @@ fn is_safe_to_call_with_exec(command: &[String]) -> bool {
.file_name()
.and_then(|osstr| osstr.to_str())
{
Some(cmd) if cfg!(target_os = "linux") && matches!(cmd, "numfmt" | "tac") => true,
#[rustfmt::skip]
Some(
"cat" |
"cd" |
"cut" |
"echo" |
"expr" |
"false" |
"grep" |
"head" |
"id" |
"ls" |
"nl" |
"paste" |
"pwd" |
"rev" |
"seq" |
"stat" |
"tail" |
"tr" |
"true" |
"uname" |
"uniq" |
"wc" |
"which" |
"whoami") => {
"which") => {
true
},
Some("base64") => {
const UNSAFE_BASE64_OPTIONS: &[&str] = &["-o", "--output"];
!command.iter().skip(1).any(|arg| {
UNSAFE_BASE64_OPTIONS.contains(&arg.as_str())
|| arg.starts_with("--output=")
|| (arg.starts_with("-o") && arg != "-o")
})
}
Some("find") => {
// Certain options to `find` can delete files, write to files, or
// execute arbitrary commands, so we cannot auto-approve the
@@ -207,7 +184,6 @@ mod tests {
fn known_safe_examples() {
assert!(is_safe_to_call_with_exec(&vec_str(&["ls"])));
assert!(is_safe_to_call_with_exec(&vec_str(&["git", "status"])));
assert!(is_safe_to_call_with_exec(&vec_str(&["base64"])));
assert!(is_safe_to_call_with_exec(&vec_str(&[
"sed", "-n", "1,5p", "file.txt"
])));
@@ -221,14 +197,6 @@ mod tests {
assert!(is_safe_to_call_with_exec(&vec_str(&[
"find", ".", "-name", "file.txt"
])));
if cfg!(target_os = "linux") {
assert!(is_safe_to_call_with_exec(&vec_str(&["numfmt", "1000"])));
assert!(is_safe_to_call_with_exec(&vec_str(&["tac", "Cargo.toml"])));
} else {
assert!(!is_safe_to_call_with_exec(&vec_str(&["numfmt", "1000"])));
assert!(!is_safe_to_call_with_exec(&vec_str(&["tac", "Cargo.toml"])));
}
}
#[test]
@@ -265,21 +233,6 @@ mod tests {
}
}
#[test]
fn base64_output_options_are_unsafe() {
for args in [
vec_str(&["base64", "-o", "out.bin"]),
vec_str(&["base64", "--output", "out.bin"]),
vec_str(&["base64", "--output=out.bin"]),
vec_str(&["base64", "-ob64.txt"]),
] {
assert!(
!is_safe_to_call_with_exec(&args),
"expected {args:?} to be considered unsafe due to output option"
);
}
}
#[test]
fn ripgrep_rules() {
// Safe ripgrep invocations none of the unsafe flags are present.

View File

@@ -1,201 +0,0 @@
$ErrorActionPreference = 'Stop'
$payload = $env:CODEX_POWERSHELL_PAYLOAD
if ([string]::IsNullOrEmpty($payload)) {
Write-Output '{"status":"parse_failed"}'
exit 0
}
try {
$source =
[System.Text.Encoding]::Unicode.GetString(
[System.Convert]::FromBase64String($payload)
)
} catch {
Write-Output '{"status":"parse_failed"}'
exit 0
}
$tokens = $null
$errors = $null
$ast = $null
try {
$ast = [System.Management.Automation.Language.Parser]::ParseInput(
$source,
[ref]$tokens,
[ref]$errors
)
} catch {
Write-Output '{"status":"parse_failed"}'
exit 0
}
if ($errors.Count -gt 0) {
Write-Output '{"status":"parse_errors"}'
exit 0
}
function Convert-CommandElement {
param($element)
if ($element -is [System.Management.Automation.Language.StringConstantExpressionAst]) {
return @($element.Value)
}
if ($element -is [System.Management.Automation.Language.ExpandableStringExpressionAst]) {
if ($element.NestedExpressions.Count -gt 0) {
return $null
}
return @($element.Value)
}
if ($element -is [System.Management.Automation.Language.ConstantExpressionAst]) {
return @($element.Value.ToString())
}
if ($element -is [System.Management.Automation.Language.CommandParameterAst]) {
if ($element.Argument -eq $null) {
return @('-' + $element.ParameterName)
}
if ($element.Argument -is [System.Management.Automation.Language.StringConstantExpressionAst]) {
return @('-' + $element.ParameterName, $element.Argument.Value)
}
if ($element.Argument -is [System.Management.Automation.Language.ConstantExpressionAst]) {
return @('-' + $element.ParameterName, $element.Argument.Value.ToString())
}
return $null
}
return $null
}
function Convert-PipelineElement {
param($element)
if ($element -is [System.Management.Automation.Language.CommandAst]) {
if ($element.Redirections.Count -gt 0) {
return $null
}
if (
$element.InvocationOperator -ne $null -and
$element.InvocationOperator -ne [System.Management.Automation.Language.TokenKind]::Unknown
) {
return $null
}
$parts = @()
foreach ($commandElement in $element.CommandElements) {
$converted = Convert-CommandElement $commandElement
if ($converted -eq $null) {
return $null
}
$parts += $converted
}
return $parts
}
if ($element -is [System.Management.Automation.Language.CommandExpressionAst]) {
if ($element.Redirections.Count -gt 0) {
return $null
}
if ($element.Expression -is [System.Management.Automation.Language.ParenExpressionAst]) {
$innerPipeline = $element.Expression.Pipeline
if ($innerPipeline -and $innerPipeline.PipelineElements.Count -eq 1) {
return Convert-PipelineElement $innerPipeline.PipelineElements[0]
}
}
return $null
}
return $null
}
function Add-CommandsFromPipelineAst {
param($pipeline, $commands)
if ($pipeline.PipelineElements.Count -eq 0) {
return $false
}
foreach ($element in $pipeline.PipelineElements) {
$words = Convert-PipelineElement $element
if ($words -eq $null -or $words.Count -eq 0) {
return $false
}
$null = $commands.Add($words)
}
return $true
}
function Add-CommandsFromPipelineChain {
param($chain, $commands)
if (-not (Add-CommandsFromPipelineBase $chain.LhsPipelineChain $commands)) {
return $false
}
if (-not (Add-CommandsFromPipelineAst $chain.RhsPipeline $commands)) {
return $false
}
return $true
}
function Add-CommandsFromPipelineBase {
param($pipeline, $commands)
if ($pipeline -is [System.Management.Automation.Language.PipelineAst]) {
return Add-CommandsFromPipelineAst $pipeline $commands
}
if ($pipeline -is [System.Management.Automation.Language.PipelineChainAst]) {
return Add-CommandsFromPipelineChain $pipeline $commands
}
return $false
}
$commands = [System.Collections.ArrayList]::new()
foreach ($statement in $ast.EndBlock.Statements) {
if (-not (Add-CommandsFromPipelineBase $statement $commands)) {
$commands = $null
break
}
}
if ($commands -ne $null) {
$normalized = [System.Collections.ArrayList]::new()
foreach ($cmd in $commands) {
if ($cmd -is [string]) {
$null = $normalized.Add(@($cmd))
continue
}
if ($cmd -is [System.Array] -or $cmd -is [System.Collections.IEnumerable]) {
$null = $normalized.Add(@($cmd))
continue
}
$normalized = $null
break
}
$commands = $normalized
}
$result = if ($commands -eq $null) {
@{ status = 'unsupported' }
} else {
@{ status = 'ok'; commands = $commands }
}
,$result | ConvertTo-Json -Depth 3

View File

@@ -1,38 +1,30 @@
use base64::Engine;
use base64::engine::general_purpose::STANDARD as BASE64_STANDARD;
use serde::Deserialize;
use shlex::split as shlex_split;
use std::path::Path;
use std::process::Command;
use std::sync::LazyLock;
const POWERSHELL_PARSER_SCRIPT: &str = include_str!("powershell_parser.ps1");
/// On Windows, we conservatively allow only clearly read-only PowerShell invocations
/// that match a small safelist. Anything else (including direct CMD commands) is unsafe.
pub fn is_safe_command_windows(command: &[String]) -> bool {
if let Some(commands) = try_parse_powershell_command_sequence(command) {
commands
return commands
.iter()
.all(|cmd| is_safe_powershell_command(cmd.as_slice()))
} else {
// Only PowerShell invocations are allowed on Windows for now; anything else is unsafe.
false
.all(|cmd| is_safe_powershell_command(cmd.as_slice()));
}
// Only PowerShell invocations are allowed on Windows for now; anything else is unsafe.
false
}
/// Returns each command sequence if the invocation starts with a PowerShell binary.
/// For example, the tokens from `pwsh Get-ChildItem | Measure-Object` become two sequences.
fn try_parse_powershell_command_sequence(command: &[String]) -> Option<Vec<Vec<String>>> {
let (exe, rest) = command.split_first()?;
if is_powershell_executable(exe) {
parse_powershell_invocation(exe, rest)
} else {
None
if !is_powershell_executable(exe) {
return None;
}
parse_powershell_invocation(rest)
}
/// Parses a PowerShell invocation into discrete command vectors, rejecting unsafe patterns.
fn parse_powershell_invocation(executable: &str, args: &[String]) -> Option<Vec<Vec<String>>> {
fn parse_powershell_invocation(args: &[String]) -> Option<Vec<Vec<String>>> {
if args.is_empty() {
// Examples rejected here: "pwsh" and "powershell.exe" with no additional arguments.
return None;
@@ -50,7 +42,7 @@ fn parse_powershell_invocation(executable: &str, args: &[String]) -> Option<Vec<
// Examples rejected here: "pwsh -Command foo bar" and "powershell -c ls extra".
return None;
}
return parse_powershell_script(executable, script);
return parse_powershell_script(script);
}
_ if lower.starts_with("-command:") || lower.starts_with("/command:") => {
if idx + 1 != args.len() {
@@ -59,7 +51,7 @@ fn parse_powershell_invocation(executable: &str, args: &[String]) -> Option<Vec<
return None;
}
let script = arg.split_once(':')?.1;
return parse_powershell_script(executable, script);
return parse_powershell_script(script);
}
// Benign, no-arg flags we tolerate.
@@ -85,8 +77,7 @@ fn parse_powershell_invocation(executable: &str, args: &[String]) -> Option<Vec<
// This happens if powershell is invoked without -Command, e.g.
// ["pwsh", "-NoLogo", "git", "-c", "core.pager=cat", "status"]
_ => {
let script = join_arguments_as_script(&args[idx..]);
return parse_powershell_script(executable, &script);
return split_into_commands(args[idx..].to_vec());
}
}
}
@@ -97,14 +88,46 @@ fn parse_powershell_invocation(executable: &str, args: &[String]) -> Option<Vec<
/// Tokenizes an inline PowerShell script and delegates to the command splitter.
/// Examples of when this is called: pwsh.exe -Command '<script>' or pwsh.exe -Command:<script>
fn parse_powershell_script(executable: &str, script: &str) -> Option<Vec<Vec<String>>> {
if let PowershellParseOutcome::Commands(commands) =
parse_with_powershell_ast(executable, script)
{
Some(commands)
} else {
None
fn parse_powershell_script(script: &str) -> Option<Vec<Vec<String>>> {
let tokens = shlex_split(script)?;
split_into_commands(tokens)
}
/// Splits tokens into pipeline segments while ensuring no unsafe separators slip through.
/// e.g. Get-ChildItem | Measure-Object -> [['Get-ChildItem'], ['Measure-Object']]
fn split_into_commands(tokens: Vec<String>) -> Option<Vec<Vec<String>>> {
if tokens.is_empty() {
// Examples rejected here: "pwsh -Command ''" and "powershell -Command \"\"".
return None;
}
let mut commands = Vec::new();
let mut current = Vec::new();
for token in tokens.into_iter() {
match token.as_str() {
"|" | "||" | "&&" | ";" => {
if current.is_empty() {
// Examples rejected here: "pwsh -Command '| Get-ChildItem'" and "pwsh -Command '; dir'".
return None;
}
commands.push(current);
current = Vec::new();
}
// Reject if any token embeds separators, redirection, or call operator characters.
_ if token.contains(['|', ';', '>', '<', '&']) || token.contains("$(") => {
// Examples rejected here: "pwsh -Command 'dir|select'" and "pwsh -Command 'echo hi > out.txt'".
return None;
}
_ => current.push(token),
}
}
if current.is_empty() {
// Examples rejected here: "pwsh -Command 'dir |'" and "pwsh -Command 'Get-ChildItem ;'".
return None;
}
commands.push(current);
Some(commands)
}
/// Returns true when the executable name is one of the supported PowerShell binaries.
@@ -121,105 +144,6 @@ fn is_powershell_executable(exe: &str) -> bool {
)
}
/// Attempts to parse PowerShell using the real PowerShell parser, returning every pipeline element
/// as a flat argv vector when possible. If parsing fails or the AST includes unsupported constructs,
/// we conservatively reject the command instead of trying to split it manually.
fn parse_with_powershell_ast(executable: &str, script: &str) -> PowershellParseOutcome {
let encoded_script = encode_powershell_base64(script);
let encoded_parser_script = encoded_parser_script();
match Command::new(executable)
.args([
"-NoLogo",
"-NoProfile",
"-NonInteractive",
"-EncodedCommand",
encoded_parser_script,
])
.env("CODEX_POWERSHELL_PAYLOAD", &encoded_script)
.output()
{
Ok(output) if output.status.success() => {
if let Ok(result) =
serde_json::from_slice::<PowershellParserOutput>(output.stdout.as_slice())
{
result.into_outcome()
} else {
PowershellParseOutcome::Failed
}
}
_ => PowershellParseOutcome::Failed,
}
}
fn encode_powershell_base64(script: &str) -> String {
let mut utf16 = Vec::with_capacity(script.len() * 2);
for unit in script.encode_utf16() {
utf16.extend_from_slice(&unit.to_le_bytes());
}
BASE64_STANDARD.encode(utf16)
}
fn encoded_parser_script() -> &'static str {
static ENCODED: LazyLock<String> =
LazyLock::new(|| encode_powershell_base64(POWERSHELL_PARSER_SCRIPT));
&ENCODED
}
#[derive(Deserialize)]
#[serde(deny_unknown_fields)]
struct PowershellParserOutput {
status: String,
commands: Option<Vec<Vec<String>>>,
}
impl PowershellParserOutput {
fn into_outcome(self) -> PowershellParseOutcome {
match self.status.as_str() {
"ok" => self
.commands
.filter(|commands| {
!commands.is_empty()
&& commands
.iter()
.all(|cmd| !cmd.is_empty() && cmd.iter().all(|word| !word.is_empty()))
})
.map(PowershellParseOutcome::Commands)
.unwrap_or(PowershellParseOutcome::Unsupported),
"unsupported" => PowershellParseOutcome::Unsupported,
_ => PowershellParseOutcome::Failed,
}
}
}
enum PowershellParseOutcome {
Commands(Vec<Vec<String>>),
Unsupported,
Failed,
}
fn join_arguments_as_script(args: &[String]) -> String {
let mut words = Vec::with_capacity(args.len());
if let Some((first, rest)) = args.split_first() {
words.push(first.clone());
for arg in rest {
words.push(quote_argument(arg));
}
}
words.join(" ")
}
fn quote_argument(arg: &str) -> String {
if arg.is_empty() {
return "''".to_string();
}
if arg.chars().all(|ch| !ch.is_whitespace()) {
return arg.to_string();
}
format!("'{}'", arg.replace('\'', "''"))
}
/// Validates that a parsed PowerShell command stays within our read-only safelist.
/// Everything before this is parsing, and rejecting things that make us feel uncomfortable.
fn is_safe_powershell_command(words: &[String]) -> bool {
@@ -252,6 +176,17 @@ fn is_safe_powershell_command(words: &[String]) -> bool {
}
}
// Block PowerShell call operator or any redirection explicitly.
if words.iter().any(|w| {
matches!(
w.as_str(),
"&" | ">" | ">>" | "1>" | "2>" | "2>&1" | "*>" | "<" | "<<"
)
}) {
// Examples rejected here: "pwsh -Command '& Remove-Item foo'" and "pwsh -Command 'Get-Content foo > bar'".
return false;
}
let command = words[0]
.trim_matches(|c| c == '(' || c == ')')
.trim_start_matches('-')
@@ -344,10 +279,9 @@ fn is_safe_git_command(words: &[String]) -> bool {
false
}
#[cfg(all(test, windows))]
#[cfg(test)]
mod tests {
use super::*;
use crate::powershell::try_find_pwsh_executable_blocking;
use super::is_safe_command_windows;
use std::string::ToString;
/// Converts a slice of string literals into owned `String`s for the tests.
@@ -378,14 +312,12 @@ mod tests {
])));
// pwsh parity
if let Some(pwsh) = try_find_pwsh_executable_blocking() {
assert!(is_safe_command_windows(&[
pwsh.as_path().to_str().unwrap().into(),
"-NoProfile".to_string(),
"-Command".to_string(),
"Get-ChildItem".to_string(),
]));
}
assert!(is_safe_command_windows(&vec_str(&[
"pwsh.exe",
"-NoProfile",
"-Command",
"Get-ChildItem",
])));
}
#[test]
@@ -395,14 +327,12 @@ mod tests {
return;
}
if let Some(pwsh) = try_find_pwsh_executable_blocking() {
assert!(is_safe_command_windows(&[
pwsh.as_path().to_str().unwrap().into(),
"-NoProfile".to_string(),
"-Command".to_string(),
"Get-ChildItem -Path .".to_string(),
]));
}
assert!(is_safe_command_windows(&vec_str(&[
r"C:\Program Files\PowerShell\7\pwsh.exe",
"-NoProfile",
"-Command",
"Get-ChildItem -Path .",
])));
assert!(is_safe_command_windows(&vec_str(&[
r"C:\Windows\System32\WindowsPowerShell\v1.0\powershell.exe",
@@ -413,53 +343,47 @@ mod tests {
#[test]
fn allows_read_only_pipelines_and_git_usage() {
let Some(pwsh) = try_find_pwsh_executable_blocking() else {
return;
};
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-NoLogo",
"-NoProfile",
"-Command",
"rg --files-with-matches foo | Measure-Object | Select-Object -ExpandProperty Count",
])));
let pwsh: String = pwsh.as_path().to_str().unwrap().into();
assert!(is_safe_command_windows(&[
pwsh.clone(),
"-NoLogo".to_string(),
"-NoProfile".to_string(),
"-Command".to_string(),
"rg --files-with-matches foo | Measure-Object | Select-Object -ExpandProperty Count"
.to_string()
]));
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-NoLogo",
"-NoProfile",
"-Command",
"Get-Content foo.rs | Select-Object -Skip 200",
])));
assert!(is_safe_command_windows(&[
pwsh.clone(),
"-NoLogo".to_string(),
"-NoProfile".to_string(),
"-Command".to_string(),
"Get-Content foo.rs | Select-Object -Skip 200".to_string()
]));
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-NoLogo",
"-NoProfile",
"-Command",
"git -c core.pager=cat show HEAD:foo.rs",
])));
assert!(is_safe_command_windows(&[
pwsh.clone(),
"-NoLogo".to_string(),
"-NoProfile".to_string(),
"-Command".to_string(),
"git -c core.pager=cat show HEAD:foo.rs".to_string()
]));
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-Command",
"-git cat-file -p HEAD:foo.rs",
])));
assert!(is_safe_command_windows(&[
pwsh.clone(),
"-Command".to_string(),
"-git cat-file -p HEAD:foo.rs".to_string()
]));
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-Command",
"(Get-Content foo.rs -Raw)",
])));
assert!(is_safe_command_windows(&[
pwsh.clone(),
"-Command".to_string(),
"(Get-Content foo.rs -Raw)".to_string()
]));
assert!(is_safe_command_windows(&[
pwsh,
"-Command".to_string(),
"Get-Item foo.rs | Select-Object Length".to_string()
]));
assert!(is_safe_command_windows(&vec_str(&[
"pwsh",
"-Command",
"Get-Item foo.rs | Select-Object Length",
])));
}
#[test]
@@ -531,93 +455,5 @@ mod tests {
"-Command",
"Get-Content (New-Item bar.txt)",
])));
// Unsafe @ expansion.
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"ls @(calc.exe)"
])));
// Unsupported constructs that the AST parser refuses (no fallback to manual splitting).
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"ls && pwd"
])));
// Sub-expressions are rejected even if they contain otherwise safe commands.
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"Write-Output $(Get-Content foo)"
])));
// Empty words from the parser (e.g. '') are rejected.
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"''"
])));
}
#[test]
fn accepts_constant_expression_arguments() {
assert!(is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"Get-Content 'foo bar'"
])));
assert!(is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"Get-Content \"foo bar\""
])));
}
#[test]
fn rejects_dynamic_arguments() {
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"Get-Content $foo"
])));
assert!(!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-Command",
"Write-Output \"foo $bar\""
])));
}
#[test]
fn uses_invoked_powershell_variant_for_parsing() {
if !cfg!(windows) {
return;
}
let chain = "pwd && ls";
assert!(
!is_safe_command_windows(&vec_str(&[
"powershell.exe",
"-NoProfile",
"-Command",
chain,
])),
"`{chain}` is not recognized by powershell.exe"
);
if let Some(pwsh) = try_find_pwsh_executable_blocking() {
assert!(
is_safe_command_windows(&[
pwsh.as_path().to_str().unwrap().into(),
"-NoProfile".to_string(),
"-Command".to_string(),
chain.to_string(),
]),
"`{chain}` should be considered safe to pwsh.exe"
);
}
}
}

View File

@@ -1,6 +1,5 @@
use std::sync::Arc;
use crate::ModelProviderInfo;
use crate::Prompt;
use crate::client_common::ResponseEvent;
use crate::codex::Session;
@@ -19,6 +18,7 @@ use crate::truncate::TruncationPolicy;
use crate::truncate::approx_token_count;
use crate::truncate::truncate_text;
use crate::util::backoff;
use codex_app_server_protocol::AuthMode;
use codex_protocol::items::TurnItem;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseInputItem;
@@ -32,11 +32,13 @@ pub const SUMMARIZATION_PROMPT: &str = include_str!("../templates/compact/prompt
pub const SUMMARY_PREFIX: &str = include_str!("../templates/compact/summary_prefix.md");
const COMPACT_USER_MESSAGE_MAX_TOKENS: usize = 20_000;
pub(crate) fn should_use_remote_compact_task(
session: &Session,
provider: &ModelProviderInfo,
) -> bool {
provider.is_openai() && session.enabled(Feature::RemoteCompaction)
pub(crate) fn should_use_remote_compact_task(session: &Session) -> bool {
session
.services
.auth_manager
.auth()
.is_some_and(|auth| auth.mode == AuthMode::ChatGPT)
&& session.enabled(Feature::RemoteCompaction)
}
pub(crate) async fn run_inline_auto_compact_task(

View File

@@ -32,8 +32,6 @@ pub enum ConfigEdit {
SetWindowsWslSetupAcknowledged(bool),
/// Toggle the model migration prompt acknowledgement flag.
SetNoticeHideModelMigrationPrompt(String, bool),
/// Record that a migration prompt was shown for an old->new model mapping.
RecordModelMigrationSeen { from: String, to: String },
/// Replace the entire `[mcp_servers]` table.
ReplaceMcpServers(BTreeMap<String, McpServerConfig>),
/// Set trust_level under `[projects."<path>"]`,
@@ -90,7 +88,7 @@ mod document_helpers {
}
}
fn serialize_mcp_server_table(config: &McpServerConfig) -> TomlTable {
pub(super) fn serialize_mcp_server(config: &McpServerConfig) -> TomlItem {
let mut entry = TomlTable::new();
entry.set_implicit(false);
@@ -161,29 +159,7 @@ mod document_helpers {
entry["disabled_tools"] = array_from_iter(disabled_tools.iter().cloned());
}
entry
}
pub(super) fn serialize_mcp_server(config: &McpServerConfig) -> TomlItem {
TomlItem::Table(serialize_mcp_server_table(config))
}
pub(super) fn serialize_mcp_server_inline(config: &McpServerConfig) -> InlineTable {
serialize_mcp_server_table(config).into_inline_table()
}
pub(super) fn merge_inline_table(existing: &mut InlineTable, replacement: InlineTable) {
existing.retain(|key, _| replacement.get(key).is_some());
for (key, value) in replacement.iter() {
if let Some(existing_value) = existing.get_mut(key) {
let mut updated_value = value.clone();
*updated_value.decor_mut() = existing_value.decor().clone();
*existing_value = updated_value;
} else {
existing.insert(key.to_string(), value.clone());
}
}
TomlItem::Table(entry)
}
fn table_from_inline(inline: &InlineTable) -> TomlTable {
@@ -287,11 +263,6 @@ impl ConfigDocument {
value(*acknowledged),
))
}
ConfigEdit::RecordModelMigrationSeen { from, to } => Ok(self.write_value(
Scope::Global,
&[Notice::TABLE_KEY, "model_migrations", from.as_str()],
value(to.clone()),
)),
ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged) => Ok(self.write_value(
Scope::Global,
&["windows_wsl_setup_acknowledged"],
@@ -339,52 +310,15 @@ impl ConfigDocument {
return self.clear(Scope::Global, &["mcp_servers"]);
}
let root = self.doc.as_table_mut();
if !root.contains_key("mcp_servers") {
root.insert(
"mcp_servers",
TomlItem::Table(document_helpers::new_implicit_table()),
);
}
let Some(item) = root.get_mut("mcp_servers") else {
return false;
};
if document_helpers::ensure_table_for_write(item).is_none() {
*item = TomlItem::Table(document_helpers::new_implicit_table());
}
let Some(table) = item.as_table_mut() else {
return false;
};
let keys_to_remove: Vec<String> = table
.iter()
.map(|(key, _)| key.to_string())
.filter(|key| !servers.contains_key(key.as_str()))
.collect();
for key in keys_to_remove {
table.remove(&key);
}
let mut table = TomlTable::new();
table.set_implicit(true);
for (name, config) in servers {
if let Some(existing) = table.get_mut(name.as_str()) {
if let TomlItem::Value(value) = existing
&& let Some(inline) = value.as_inline_table_mut()
{
let replacement = document_helpers::serialize_mcp_server_inline(config);
document_helpers::merge_inline_table(inline, replacement);
} else {
*existing = document_helpers::serialize_mcp_server(config);
}
} else {
table.insert(name, document_helpers::serialize_mcp_server(config));
}
table.insert(name, document_helpers::serialize_mcp_server(config));
}
true
let item = TomlItem::Table(table);
self.write_value(Scope::Global, &["mcp_servers"], item)
}
fn scoped_segments(&self, scope: Scope, segments: &[&str]) -> Vec<String> {
@@ -416,10 +350,6 @@ impl ConfigDocument {
return false;
};
let mut value = value;
if let Some(existing) = parent.get(last) {
Self::preserve_decor(existing, &mut value);
}
parent[last] = value;
true
}
@@ -461,37 +391,6 @@ impl ConfigDocument {
Some(current)
}
fn preserve_decor(existing: &TomlItem, replacement: &mut TomlItem) {
match (existing, replacement) {
(TomlItem::Table(existing_table), TomlItem::Table(replacement_table)) => {
replacement_table
.decor_mut()
.clone_from(existing_table.decor());
for (key, existing_item) in existing_table.iter() {
if let (Some(existing_key), Some(mut replacement_key)) =
(existing_table.key(key), replacement_table.key_mut(key))
{
replacement_key
.leaf_decor_mut()
.clone_from(existing_key.leaf_decor());
replacement_key
.dotted_decor_mut()
.clone_from(existing_key.dotted_decor());
}
if let Some(replacement_item) = replacement_table.get_mut(key) {
Self::preserve_decor(existing_item, replacement_item);
}
}
}
(TomlItem::Value(existing_value), TomlItem::Value(replacement_value)) => {
replacement_value
.decor_mut()
.clone_from(existing_value.decor());
}
_ => {}
}
}
}
/// Persist edits using a blocking strategy.
@@ -623,14 +522,6 @@ impl ConfigEditsBuilder {
self
}
pub fn record_model_migration_seen(mut self, from: &str, to: &str) -> Self {
self.edits.push(ConfigEdit::RecordModelMigrationSeen {
from: from.to_string(),
to: to.to_string(),
});
self
}
pub fn set_windows_wsl_setup_acknowledged(mut self, acknowledged: bool) -> Self {
self.edits
.push(ConfigEdit::SetWindowsWslSetupAcknowledged(acknowledged));
@@ -664,14 +555,6 @@ impl ConfigEditsBuilder {
self
}
pub fn with_edits<I>(mut self, edits: I) -> Self
where
I: IntoIterator<Item = ConfigEdit>,
{
self.edits.extend(edits);
self
}
/// Apply edits on a blocking thread.
pub fn apply_blocking(self) -> anyhow::Result<()> {
apply_blocking(&self.codex_home, self.profile.as_deref(), &self.edits)
@@ -720,24 +603,6 @@ model_reasoning_effort = "high"
assert_eq!(contents, expected);
}
#[test]
fn builder_with_edits_applies_custom_paths() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
ConfigEditsBuilder::new(codex_home)
.with_edits(vec![ConfigEdit::SetPath {
segments: vec!["enabled".to_string()],
value: value(true),
}])
.apply_blocking()
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert_eq!(contents, "enabled = true\n");
}
#[test]
fn blocking_set_model_preserves_inline_table_contents() {
let tmp = tempdir().expect("tmpdir");
@@ -785,68 +650,6 @@ profiles = { fast = { model = "gpt-4o", sandbox_mode = "strict" } }
);
}
#[test]
fn batch_write_table_upsert_preserves_inline_comments() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
let original = r#"approval_policy = "never"
[mcp_servers.linear]
name = "linear"
# ok
url = "https://linear.example"
[mcp_servers.linear.http_headers]
foo = "bar"
[sandbox_workspace_write]
# ok 3
network_access = false
"#;
std::fs::write(codex_home.join(CONFIG_TOML_FILE), original).expect("seed config");
apply_blocking(
codex_home,
None,
&[
ConfigEdit::SetPath {
segments: vec![
"mcp_servers".to_string(),
"linear".to_string(),
"url".to_string(),
],
value: value("https://linear.example/v2"),
},
ConfigEdit::SetPath {
segments: vec![
"sandbox_workspace_write".to_string(),
"network_access".to_string(),
],
value: value(true),
},
],
)
.expect("apply");
let updated =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"approval_policy = "never"
[mcp_servers.linear]
name = "linear"
# ok
url = "https://linear.example/v2"
[mcp_servers.linear.http_headers]
foo = "bar"
[sandbox_workspace_write]
# ok 3
network_access = true
"#;
assert_eq!(updated, expected);
}
#[test]
fn blocking_clear_model_removes_inline_table_entry() {
let tmp = tempdir().expect("tmpdir");
@@ -1068,38 +871,6 @@ existing = "value"
assert_eq!(contents, expected);
}
#[test]
fn blocking_record_model_migration_seen_preserves_table() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[notice]
existing = "value"
"#,
)
.expect("seed");
apply_blocking(
codex_home,
None,
&[ConfigEdit::RecordModelMigrationSeen {
from: "gpt-5".to_string(),
to: "gpt-5.1".to_string(),
}],
)
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[notice]
existing = "value"
[notice.model_migrations]
gpt-5 = "gpt-5.1"
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_replace_mcp_servers_round_trips() {
let tmp = tempdir().expect("tmpdir");
@@ -1184,178 +955,6 @@ B = \"2\"
assert_eq!(raw, expected);
}
#[test]
fn blocking_replace_mcp_servers_preserves_inline_comments() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[mcp_servers]
# keep me
foo = { command = "cmd" }
"#,
)
.expect("seed");
let mut servers = BTreeMap::new();
servers.insert(
"foo".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::Stdio {
command: "cmd".to_string(),
args: Vec::new(),
env: None,
env_vars: Vec::new(),
cwd: None,
},
enabled: true,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
);
apply_blocking(codex_home, None, &[ConfigEdit::ReplaceMcpServers(servers)])
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[mcp_servers]
# keep me
foo = { command = "cmd" }
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_replace_mcp_servers_preserves_inline_comment_suffix() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[mcp_servers]
foo = { command = "cmd" } # keep me
"#,
)
.expect("seed");
let mut servers = BTreeMap::new();
servers.insert(
"foo".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::Stdio {
command: "cmd".to_string(),
args: Vec::new(),
env: None,
env_vars: Vec::new(),
cwd: None,
},
enabled: false,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
);
apply_blocking(codex_home, None, &[ConfigEdit::ReplaceMcpServers(servers)])
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[mcp_servers]
foo = { command = "cmd" , enabled = false } # keep me
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_replace_mcp_servers_preserves_inline_comment_after_removing_keys() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[mcp_servers]
foo = { command = "cmd", args = ["--flag"] } # keep me
"#,
)
.expect("seed");
let mut servers = BTreeMap::new();
servers.insert(
"foo".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::Stdio {
command: "cmd".to_string(),
args: Vec::new(),
env: None,
env_vars: Vec::new(),
cwd: None,
},
enabled: true,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
);
apply_blocking(codex_home, None, &[ConfigEdit::ReplaceMcpServers(servers)])
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[mcp_servers]
foo = { command = "cmd"} # keep me
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_replace_mcp_servers_preserves_inline_comment_prefix_on_update() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[mcp_servers]
# keep me
foo = { command = "cmd" }
"#,
)
.expect("seed");
let mut servers = BTreeMap::new();
servers.insert(
"foo".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::Stdio {
command: "cmd".to_string(),
args: Vec::new(),
env: None,
env_vars: Vec::new(),
cwd: None,
},
enabled: false,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
);
apply_blocking(codex_home, None, &[ConfigEdit::ReplaceMcpServers(servers)])
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"[mcp_servers]
# keep me
foo = { command = "cmd" , enabled = false }
"#;
assert_eq!(contents, expected);
}
#[test]
fn blocking_clear_path_noop_when_missing() {
let tmp = tempdir().expect("tmpdir");

View File

@@ -7,12 +7,16 @@ use crate::config::types::Notifications;
use crate::config::types::OtelConfig;
use crate::config::types::OtelConfigToml;
use crate::config::types::OtelExporterKind;
use crate::config::types::ReasoningSummaryFormat;
use crate::config::types::SandboxWorkspaceWrite;
use crate::config::types::ShellEnvironmentPolicy;
use crate::config::types::ShellEnvironmentPolicyToml;
use crate::config::types::Tui;
use crate::config::types::UriBasedFileOpener;
use crate::config_loader::load_config_layers_state;
use crate::config_loader::LoadedConfigLayers;
use crate::config_loader::load_config_as_toml;
use crate::config_loader::load_config_layers_with_overrides;
use crate::config_loader::merge_toml_values;
use crate::features::Feature;
use crate::features::FeatureOverrides;
use crate::features::Features;
@@ -22,6 +26,8 @@ use crate::model_provider_info::LMSTUDIO_OSS_PROVIDER_ID;
use crate::model_provider_info::ModelProviderInfo;
use crate::model_provider_info::OLLAMA_OSS_PROVIDER_ID;
use crate::model_provider_info::built_in_model_providers;
use crate::openai_model_info::get_model_info;
use crate::openai_models::model_family::find_family_for_model;
use crate::project_doc::DEFAULT_PROJECT_DOC_FILENAME;
use crate::project_doc::LOCAL_PROJECT_DOC_FILENAME;
use crate::protocol::AskForApproval;
@@ -35,11 +41,9 @@ 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;
use dirs::home_dir;
use dunce::canonicalize;
use serde::Deserialize;
use similar::DiffableStr;
use std::collections::BTreeMap;
@@ -47,8 +51,6 @@ use std::collections::HashMap;
use std::io::ErrorKind;
use std::path::Path;
use std::path::PathBuf;
#[cfg(test)]
use tempfile::tempdir;
use crate::config::profile::ConfigProfile;
use toml::Value as TomlValue;
@@ -56,16 +58,11 @@ use toml_edit::DocumentMut;
pub mod edit;
pub mod profile;
pub mod service;
pub mod types;
pub use service::ConfigService;
pub use service::ConfigServiceError;
pub const OPENAI_DEFAULT_MODEL: &str = "gpt-5.1-codex-max";
const OPENAI_DEFAULT_REVIEW_MODEL: &str = "gpt-5.1-codex-max";
pub use codex_git::GhostSnapshotConfig;
/// Maximum number of bytes of the documentation that will be embedded. Larger
/// files are *silently truncated* to this size so we do not take up too much of
/// the context window.
@@ -73,22 +70,11 @@ pub(crate) const PROJECT_DOC_MAX_BYTES: usize = 32 * 1024; // 32 KiB
pub const CONFIG_TOML_FILE: &str = "config.toml";
#[cfg(test)]
pub(crate) fn test_config() -> Config {
let codex_home = tempdir().expect("create temp dir");
Config::load_from_base_config_with_overrides(
ConfigToml::default(),
ConfigOverrides::default(),
codex_home.path().to_path_buf(),
)
.expect("load default test config")
}
/// Application configuration loaded from disk and merged with overrides.
#[derive(Debug, Clone, PartialEq)]
pub struct Config {
/// Optional override of model selection.
pub model: Option<String>,
pub model: String,
/// Model used specifically for review sessions. Defaults to "gpt-5.1-codex-max".
pub review_model: String,
@@ -261,6 +247,9 @@ pub struct Config {
pub tools_web_search_request: bool,
/// When `true`, run a model-based assessment for commands denied by the sandbox.
pub experimental_sandbox_command_assessment: bool,
/// If set to `true`, used only the experimental unified exec tool.
pub use_experimental_unified_exec_tool: bool,
@@ -268,9 +257,6 @@ pub struct Config {
/// https://github.com/modelcontextprotocol/rust-sdk
pub use_experimental_use_rmcp_client: bool,
/// Settings for ghost snapshots (used for undo).
pub ghost_snapshot: GhostSnapshotConfig,
/// Centralized feature flags; source of truth for feature gating.
pub features: Features,
@@ -315,9 +301,9 @@ impl Config {
)
.await?;
let cfg = deserialize_config_toml_with_base(root_value, &codex_home).map_err(|e| {
let cfg: ConfigToml = root_value.try_into().map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
e
std::io::Error::new(std::io::ErrorKind::InvalidData, e)
})?;
Self::load_from_base_config_with_overrides(cfg, overrides, codex_home)
@@ -335,9 +321,9 @@ pub async fn load_config_as_toml_with_cli_overrides(
)
.await?;
let cfg = deserialize_config_toml_with_base(root_value, codex_home).map_err(|e| {
let cfg: ConfigToml = root_value.try_into().map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
e
std::io::Error::new(std::io::ErrorKind::InvalidData, e)
})?;
Ok(cfg)
@@ -348,31 +334,35 @@ async fn load_resolved_config(
cli_overrides: Vec<(String, TomlValue)>,
overrides: crate::config_loader::LoaderOverrides,
) -> std::io::Result<TomlValue> {
let layers = load_config_layers_state(codex_home, &cli_overrides, overrides).await?;
Ok(layers.effective_config())
let layers = load_config_layers_with_overrides(codex_home, overrides).await?;
Ok(apply_overlays(layers, cli_overrides))
}
fn deserialize_config_toml_with_base(
root_value: TomlValue,
config_base_dir: &Path,
) -> std::io::Result<ConfigToml> {
// This guard ensures that any relative paths that is deserialized into an
// [AbsolutePathBuf] is resolved against `config_base_dir`.
let _guard = AbsolutePathBufGuard::new(config_base_dir);
root_value
.try_into()
.map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e))
fn apply_overlays(
layers: LoadedConfigLayers,
cli_overrides: Vec<(String, TomlValue)>,
) -> TomlValue {
let LoadedConfigLayers {
mut base,
managed_config,
managed_preferences,
} = layers;
for (path, value) in cli_overrides.into_iter() {
apply_toml_override(&mut base, &path, value);
}
for overlay in [managed_config, managed_preferences].into_iter().flatten() {
merge_toml_values(&mut base, &overlay);
}
base
}
pub async fn load_global_mcp_servers(
codex_home: &Path,
) -> std::io::Result<BTreeMap<String, McpServerConfig>> {
let root_value = load_resolved_config(
codex_home,
Vec::new(),
crate::config_loader::LoaderOverrides::default(),
)
.await?;
let root_value = load_config_as_toml(codex_home).await?;
let Some(servers_value) = root_value.get("mcp_servers") else {
return Ok(BTreeMap::new());
};
@@ -531,6 +521,49 @@ pub fn set_default_oss_provider(codex_home: &Path, provider: &str) -> std::io::R
Ok(())
}
/// Apply a single dotted-path override onto a TOML value.
fn apply_toml_override(root: &mut TomlValue, path: &str, value: TomlValue) {
use toml::value::Table;
let segments: Vec<&str> = path.split('.').collect();
let mut current = root;
for (idx, segment) in segments.iter().enumerate() {
let is_last = idx == segments.len() - 1;
if is_last {
match current {
TomlValue::Table(table) => {
table.insert(segment.to_string(), value);
}
_ => {
let mut table = Table::new();
table.insert(segment.to_string(), value);
*current = TomlValue::Table(table);
}
}
return;
}
// Traverse or create intermediate object.
match current {
TomlValue::Table(table) => {
current = table
.entry(segment.to_string())
.or_insert_with(|| TomlValue::Table(Table::new()));
}
_ => {
*current = TomlValue::Table(Table::new());
if let TomlValue::Table(tbl) = current {
current = tbl
.entry(segment.to_string())
.or_insert_with(|| TomlValue::Table(Table::new()));
}
}
}
}
}
/// Base config deserialized from ~/.codex/config.toml.
#[derive(Deserialize, Debug, Clone, Default, PartialEq)]
pub struct ConfigToml {
@@ -663,10 +696,6 @@ pub struct ConfigToml {
#[serde(default)]
pub features: Option<FeaturesToml>,
/// Settings for ghost snapshots (used for undo).
#[serde(default)]
pub ghost_snapshot: Option<GhostSnapshotToml>,
/// When `true`, checks for Codex updates on startup and surfaces update prompts.
/// Set to `false` only if your Codex updates are centrally managed.
/// Defaults to `true`.
@@ -693,6 +722,7 @@ pub struct ConfigToml {
pub experimental_use_unified_exec_tool: Option<bool>,
pub experimental_use_rmcp_client: Option<bool>,
pub experimental_use_freeform_apply_patch: Option<bool>,
pub experimental_sandbox_command_assessment: Option<bool>,
/// Preferred OSS provider for local models, e.g. "lmstudio" or "ollama".
pub oss_provider: Option<String>,
}
@@ -756,17 +786,6 @@ impl From<ToolsToml> for Tools {
}
}
#[derive(Deserialize, Debug, Clone, Default, PartialEq, Eq)]
pub struct GhostSnapshotToml {
/// Exclude untracked files larger than this many bytes from ghost snapshots.
#[serde(alias = "ignore_untracked_files_over_bytes")]
pub ignore_large_untracked_files: Option<i64>,
/// Ignore untracked directories that contain this many files or more.
/// (Still emits a warning.)
#[serde(alias = "large_untracked_dir_warning_threshold")]
pub ignore_large_untracked_dirs: Option<i64>,
}
#[derive(Debug, PartialEq, Eq)]
pub struct SandboxPolicyResolution {
pub policy: SandboxPolicy,
@@ -889,6 +908,7 @@ pub struct ConfigOverrides {
pub include_apply_patch_tool: Option<bool>,
pub show_raw_agent_reasoning: Option<bool>,
pub tools_web_search_request: Option<bool>,
pub experimental_sandbox_command_assessment: Option<bool>,
/// Additional directories that should be treated as writable roots for this session.
pub additional_writable_roots: Vec<PathBuf>,
}
@@ -947,6 +967,7 @@ impl Config {
include_apply_patch_tool: include_apply_patch_tool_override,
show_raw_agent_reasoning,
tools_web_search_request: override_tools_web_search_request,
experimental_sandbox_command_assessment: sandbox_command_assessment_override,
additional_writable_roots,
} = overrides;
@@ -971,17 +992,13 @@ impl Config {
let feature_overrides = FeatureOverrides {
include_apply_patch_tool: include_apply_patch_tool_override,
web_search_request: override_tools_web_search_request,
experimental_sandbox_command_assessment: sandbox_command_assessment_override,
};
let features = Features::from_config(&cfg, &config_profile, feature_overrides);
#[cfg(target_os = "windows")]
{
// Base flag controls sandbox on/off; elevated only applies when base is enabled.
let sandbox_enabled = features.enabled(Feature::WindowsSandbox);
crate::safety::set_windows_sandbox_enabled(sandbox_enabled);
let elevated_enabled =
sandbox_enabled && features.enabled(Feature::WindowsSandboxElevated);
crate::safety::set_windows_elevated_sandbox_enabled(elevated_enabled);
crate::safety::set_windows_sandbox_enabled(features.enabled(Feature::WindowsSandbox));
}
let resolved_cwd = {
@@ -1002,10 +1019,13 @@ impl Config {
}
}
};
let additional_writable_roots: Vec<AbsolutePathBuf> = additional_writable_roots
let additional_writable_roots: Vec<PathBuf> = additional_writable_roots
.into_iter()
.map(|path| AbsolutePathBuf::resolve_path_against_base(path, &resolved_cwd))
.collect::<Result<Vec<_>, _>>()?;
.map(|path| {
let absolute = resolve_path(&resolved_cwd, &path);
canonicalize(&absolute).unwrap_or(absolute)
})
.collect();
let active_project = cfg
.get_active_project(&resolved_cwd)
.unwrap_or(ProjectConfig { trust_level: None });
@@ -1067,30 +1087,12 @@ impl Config {
let history = cfg.history.unwrap_or_default();
let ghost_snapshot = {
let mut config = GhostSnapshotConfig::default();
if let Some(ghost_snapshot) = cfg.ghost_snapshot.as_ref()
&& let Some(ignore_over_bytes) = ghost_snapshot.ignore_large_untracked_files
{
config.ignore_large_untracked_files = if ignore_over_bytes > 0 {
Some(ignore_over_bytes)
} else {
None
};
}
if let Some(ghost_snapshot) = cfg.ghost_snapshot.as_ref()
&& let Some(threshold) = ghost_snapshot.ignore_large_untracked_dirs
{
config.ignore_large_untracked_dirs =
if threshold > 0 { Some(threshold) } else { None };
}
config
};
let include_apply_patch_tool_flag = features.enabled(Feature::ApplyPatchFreeform);
let tools_web_search_request = features.enabled(Feature::WebSearchRequest);
let use_experimental_unified_exec_tool = features.enabled(Feature::UnifiedExec);
let use_experimental_use_rmcp_client = features.enabled(Feature::RmcpClient);
let experimental_sandbox_command_assessment =
features.enabled(Feature::SandboxCommandAssessment);
let forced_chatgpt_workspace_id =
cfg.forced_chatgpt_workspace_id.as_ref().and_then(|value| {
@@ -1104,7 +1106,22 @@ impl Config {
let forced_login_method = cfg.forced_login_method;
let model = model.or(config_profile.model).or(cfg.model);
let model = model
.or(config_profile.model)
.or(cfg.model)
.unwrap_or_else(default_model);
let model_family = find_family_for_model(&model);
let openai_model_info = get_model_info(&model_family);
let model_context_window = cfg
.model_context_window
.or_else(|| openai_model_info.as_ref().map(|info| info.context_window));
let model_auto_compact_token_limit = cfg.model_auto_compact_token_limit.or_else(|| {
openai_model_info
.as_ref()
.and_then(|info| info.auto_compact_token_limit)
});
let compact_prompt = compact_prompt.or(cfg.compact_prompt).and_then(|value| {
let trimmed = value.trim();
@@ -1151,8 +1168,8 @@ impl Config {
let config = Self {
model,
review_model,
model_context_window: cfg.model_context_window,
model_auto_compact_token_limit: cfg.model_auto_compact_token_limit,
model_context_window,
model_auto_compact_token_limit,
model_provider_id,
model_provider,
cwd: resolved_cwd,
@@ -1217,9 +1234,9 @@ impl Config {
forced_login_method,
include_apply_patch_tool: include_apply_patch_tool_flag,
tools_web_search_request,
experimental_sandbox_command_assessment,
use_experimental_unified_exec_tool,
use_experimental_use_rmcp_client,
ghost_snapshot,
features,
active_profile: active_profile_name,
active_project,
@@ -1241,12 +1258,10 @@ impl Config {
.environment
.unwrap_or(DEFAULT_OTEL_ENVIRONMENT.to_string());
let exporter = t.exporter.unwrap_or(OtelExporterKind::None);
let trace_exporter = t.trace_exporter.unwrap_or_else(|| exporter.clone());
OtelConfig {
log_user_prompt,
environment,
exporter,
trace_exporter,
}
},
};
@@ -1308,6 +1323,10 @@ impl Config {
}
}
fn default_model() -> String {
OPENAI_DEFAULT_MODEL.to_string()
}
fn default_review_model() -> String {
OPENAI_DEFAULT_REVIEW_MODEL.to_string()
}
@@ -1358,7 +1377,6 @@ mod tests {
use crate::features::Feature;
use super::*;
use core_test_support::test_absolute_path;
use pretty_assertions::assert_eq;
use std::time::Duration;
@@ -1457,22 +1475,18 @@ network_access = true # This should be ignored.
}
);
let writable_root = test_absolute_path("/my/workspace");
let sandbox_workspace_write = format!(
r#"
let sandbox_workspace_write = r#"
sandbox_mode = "workspace-write"
[sandbox_workspace_write]
writable_roots = [
{},
"/my/workspace",
]
exclude_tmpdir_env_var = true
exclude_slash_tmp = true
"#,
serde_json::json!(writable_root)
);
"#;
let sandbox_workspace_write_cfg = toml::from_str::<ConfigToml>(&sandbox_workspace_write)
let sandbox_workspace_write_cfg = toml::from_str::<ConfigToml>(sandbox_workspace_write)
.expect("TOML deserialization should succeed");
let sandbox_mode_override = None;
let resolution = sandbox_workspace_write_cfg.derive_sandbox_policy(
@@ -1493,7 +1507,7 @@ exclude_slash_tmp = true
resolution,
SandboxPolicyResolution {
policy: SandboxPolicy::WorkspaceWrite {
writable_roots: vec![writable_root.clone()],
writable_roots: vec![PathBuf::from("/my/workspace")],
network_access: false,
exclude_tmpdir_env_var: true,
exclude_slash_tmp: true,
@@ -1503,24 +1517,21 @@ exclude_slash_tmp = true
);
}
let sandbox_workspace_write = format!(
r#"
let sandbox_workspace_write = r#"
sandbox_mode = "workspace-write"
[sandbox_workspace_write]
writable_roots = [
{},
"/my/workspace",
]
exclude_tmpdir_env_var = true
exclude_slash_tmp = true
[projects."/tmp/test"]
trust_level = "trusted"
"#,
serde_json::json!(writable_root)
);
"#;
let sandbox_workspace_write_cfg = toml::from_str::<ConfigToml>(&sandbox_workspace_write)
let sandbox_workspace_write_cfg = toml::from_str::<ConfigToml>(sandbox_workspace_write)
.expect("TOML deserialization should succeed");
let sandbox_mode_override = None;
let resolution = sandbox_workspace_write_cfg.derive_sandbox_policy(
@@ -1541,7 +1552,7 @@ trust_level = "trusted"
resolution,
SandboxPolicyResolution {
policy: SandboxPolicy::WorkspaceWrite {
writable_roots: vec![writable_root],
writable_roots: vec![PathBuf::from("/my/workspace")],
network_access: false,
exclude_tmpdir_env_var: true,
exclude_slash_tmp: true,
@@ -1573,7 +1584,7 @@ trust_level = "trusted"
temp_dir.path().to_path_buf(),
)?;
let expected_backend = AbsolutePathBuf::try_from(backend).unwrap();
let expected_backend = canonicalize(&backend).expect("canonicalize backend directory");
if cfg!(target_os = "windows") {
assert!(
config.forced_auto_mode_downgraded_on_windows,
@@ -1854,11 +1865,10 @@ trust_level = "trusted"
};
let root_value = load_resolved_config(codex_home.path(), Vec::new(), overrides).await?;
let cfg =
deserialize_config_toml_with_base(root_value, codex_home.path()).map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
e
})?;
let cfg: ConfigToml = root_value.try_into().map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
std::io::Error::new(std::io::ErrorKind::InvalidData, e)
})?;
assert_eq!(
cfg.mcp_oauth_credentials_store,
Some(OAuthCredentialsStoreMode::Keyring),
@@ -1975,11 +1985,10 @@ trust_level = "trusted"
)
.await?;
let cfg =
deserialize_config_toml_with_base(root_value, codex_home.path()).map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
e
})?;
let cfg: ConfigToml = root_value.try_into().map_err(|e| {
tracing::error!("Failed to deserialize overridden config: {e}");
std::io::Error::new(std::io::ErrorKind::InvalidData, e)
})?;
assert_eq!(cfg.model.as_deref(), Some("managed_config"));
Ok(())
@@ -2939,10 +2948,10 @@ model_verbosity = "high"
)?;
assert_eq!(
Config {
model: Some("o3".to_string()),
model: "o3".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: None,
model_auto_compact_token_limit: None,
model_context_window: Some(200_000),
model_auto_compact_token_limit: Some(180_000),
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::Never,
@@ -2979,9 +2988,9 @@ model_verbosity = "high"
forced_login_method: None,
include_apply_patch_tool: false,
tools_web_search_request: false,
experimental_sandbox_command_assessment: false,
use_experimental_unified_exec_tool: false,
use_experimental_use_rmcp_client: false,
ghost_snapshot: GhostSnapshotConfig::default(),
features: Features::with_defaults(),
active_profile: Some("o3".to_string()),
active_project: ProjectConfig { trust_level: None },
@@ -3014,10 +3023,10 @@ model_verbosity = "high"
fixture.codex_home(),
)?;
let expected_gpt3_profile_config = Config {
model: Some("gpt-3.5-turbo".to_string()),
model: "gpt-3.5-turbo".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: None,
model_auto_compact_token_limit: None,
model_context_window: Some(16_385),
model_auto_compact_token_limit: Some(14_746),
model_provider_id: "openai-chat-completions".to_string(),
model_provider: fixture.openai_chat_completions_provider.clone(),
approval_policy: AskForApproval::UnlessTrusted,
@@ -3054,9 +3063,9 @@ model_verbosity = "high"
forced_login_method: None,
include_apply_patch_tool: false,
tools_web_search_request: false,
experimental_sandbox_command_assessment: false,
use_experimental_unified_exec_tool: false,
use_experimental_use_rmcp_client: false,
ghost_snapshot: GhostSnapshotConfig::default(),
features: Features::with_defaults(),
active_profile: Some("gpt3".to_string()),
active_project: ProjectConfig { trust_level: None },
@@ -3104,10 +3113,10 @@ model_verbosity = "high"
fixture.codex_home(),
)?;
let expected_zdr_profile_config = Config {
model: Some("o3".to_string()),
model: "o3".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: None,
model_auto_compact_token_limit: None,
model_context_window: Some(200_000),
model_auto_compact_token_limit: Some(180_000),
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::OnFailure,
@@ -3144,9 +3153,9 @@ model_verbosity = "high"
forced_login_method: None,
include_apply_patch_tool: false,
tools_web_search_request: false,
experimental_sandbox_command_assessment: false,
use_experimental_unified_exec_tool: false,
use_experimental_use_rmcp_client: false,
ghost_snapshot: GhostSnapshotConfig::default(),
features: Features::with_defaults(),
active_profile: Some("zdr".to_string()),
active_project: ProjectConfig { trust_level: None },
@@ -3180,10 +3189,10 @@ model_verbosity = "high"
fixture.codex_home(),
)?;
let expected_gpt5_profile_config = Config {
model: Some("gpt-5.1".to_string()),
model: "gpt-5.1".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_context_window: None,
model_auto_compact_token_limit: None,
model_context_window: Some(272_000),
model_auto_compact_token_limit: Some(244_800),
model_provider_id: "openai".to_string(),
model_provider: fixture.openai_provider.clone(),
approval_policy: AskForApproval::OnFailure,
@@ -3220,9 +3229,9 @@ model_verbosity = "high"
forced_login_method: None,
include_apply_patch_tool: false,
tools_web_search_request: false,
experimental_sandbox_command_assessment: false,
use_experimental_unified_exec_tool: false,
use_experimental_use_rmcp_client: false,
ghost_snapshot: GhostSnapshotConfig::default(),
features: Features::with_defaults(),
active_profile: Some("gpt5".to_string()),
active_project: ProjectConfig { trust_level: None },

View File

@@ -27,6 +27,7 @@ pub struct ConfigProfile {
pub experimental_use_unified_exec_tool: Option<bool>,
pub experimental_use_rmcp_client: Option<bool>,
pub experimental_use_freeform_apply_patch: Option<bool>,
pub experimental_sandbox_command_assessment: Option<bool>,
pub tools_web_search: Option<bool>,
pub tools_view_image: Option<bool>,
/// Optional feature toggles scoped to this profile.

File diff suppressed because it is too large Load Diff

View File

@@ -3,15 +3,13 @@
// Note this file should generally be restricted to simple struct/enum
// definitions that do not contain business logic.
use codex_utils_absolute_path::AbsolutePathBuf;
use std::collections::BTreeMap;
use serde::Deserializer;
use std::collections::HashMap;
use std::path::PathBuf;
use std::time::Duration;
use wildmatch::WildMatchPattern;
use serde::Deserialize;
use serde::Deserializer;
use serde::Serialize;
use serde::de::Error as SerdeError;
@@ -287,9 +285,9 @@ pub enum OtelHttpProtocol {
#[derive(Deserialize, Debug, Clone, PartialEq, Default)]
#[serde(rename_all = "kebab-case")]
pub struct OtelTlsConfig {
pub ca_certificate: Option<AbsolutePathBuf>,
pub client_certificate: Option<AbsolutePathBuf>,
pub client_private_key: Option<AbsolutePathBuf>,
pub ca_certificate: Option<PathBuf>,
pub client_certificate: Option<PathBuf>,
pub client_private_key: Option<PathBuf>,
}
/// Which OTEL exporter to use.
@@ -323,11 +321,8 @@ pub struct OtelConfigToml {
/// Mark traces with environment (dev, staging, prod, test). Defaults to dev.
pub environment: Option<String>,
/// Optional log exporter
/// Exporter to use. Defaults to `otlp-file`.
pub exporter: Option<OtelExporterKind>,
/// Optional trace exporter
pub trace_exporter: Option<OtelExporterKind>,
}
/// Effective OTEL settings after defaults are applied.
@@ -336,7 +331,6 @@ pub struct OtelConfig {
pub log_user_prompt: bool,
pub environment: String,
pub exporter: OtelExporterKind,
pub trace_exporter: OtelExporterKind,
}
impl Default for OtelConfig {
@@ -345,7 +339,6 @@ impl Default for OtelConfig {
log_user_prompt: false,
environment: DEFAULT_OTEL_ENVIRONMENT.to_owned(),
exporter: OtelExporterKind::None,
trace_exporter: OtelExporterKind::None,
}
}
}
@@ -402,9 +395,6 @@ pub struct Notice {
/// Tracks whether the user has seen the gpt-5.1-codex-max migration prompt
#[serde(rename = "hide_gpt-5.1-codex-max_migration_prompt")]
pub hide_gpt_5_1_codex_max_migration_prompt: Option<bool>,
/// Tracks acknowledged model migrations as old->new model slug mappings.
#[serde(default)]
pub model_migrations: BTreeMap<String, String>,
}
impl Notice {
@@ -415,7 +405,7 @@ impl Notice {
#[derive(Deserialize, Debug, Clone, PartialEq, Default)]
pub struct SandboxWorkspaceWrite {
#[serde(default)]
pub writable_roots: Vec<AbsolutePathBuf>,
pub writable_roots: Vec<PathBuf>,
#[serde(default)]
pub network_access: bool,
#[serde(default)]
@@ -531,6 +521,14 @@ impl From<ShellEnvironmentPolicyToml> for ShellEnvironmentPolicy {
}
}
#[derive(Deserialize, Debug, Clone, PartialEq, Eq, Default, Hash)]
#[serde(rename_all = "kebab-case")]
pub enum ReasoningSummaryFormat {
#[default]
None,
Experimental,
}
#[cfg(test)]
mod tests {
use super::*;

View File

@@ -1,64 +0,0 @@
# `codex-core` config loader
This module is the canonical place to **load and describe Codex configuration layers** (user config, CLI/session overrides, managed config, and MDM-managed preferences) and to produce:
- An **effective merged** TOML config.
- **Per-key origins** metadata (which layer “wins” for a given key).
- **Per-layer versions** (stable fingerprints) used for optimistic concurrency / conflict detection.
## Public surface
Exported from `codex_core::config_loader`:
- `load_config_layers_state(codex_home, cli_overrides, overrides) -> ConfigLayerStack`
- `ConfigLayerStack`
- `effective_config() -> toml::Value`
- `origins() -> HashMap<String, ConfigLayerMetadata>`
- `layers_high_to_low() -> Vec<ConfigLayer>`
- `with_user_config(user_config) -> ConfigLayerStack`
- `ConfigLayerEntry` (one layers `{name, source, config, version}`)
- `LoaderOverrides` (test/override hooks for managed config sources)
- `merge_toml_values(base, overlay)` (public helper used elsewhere)
## Layering model
Precedence is **top overrides bottom**:
1. **MDM** managed preferences (macOS only)
2. **System** managed config (e.g. `managed_config.toml`)
3. **Session flags** (CLI overrides, applied as dotted-path TOML writes)
4. **User** config (`config.toml`)
This is what `ConfigLayerStack::effective_config()` implements.
## Typical usage
Most callers want the effective config plus metadata:
```rust
use codex_core::config_loader::{load_config_layers_state, LoaderOverrides};
use toml::Value as TomlValue;
let cli_overrides: Vec<(String, TomlValue)> = Vec::new();
let layers = load_config_layers_state(
&codex_home,
&cli_overrides,
LoaderOverrides::default(),
).await?;
let effective = layers.effective_config();
let origins = layers.origins();
let layers_for_ui = layers.layers_high_to_low();
```
## Internal layout
Implementation is split by concern:
- `state.rs`: public types (`ConfigLayerEntry`, `ConfigLayerStack`) + merge/origins convenience methods.
- `layer_io.rs`: reading `config.toml`, managed config, and managed preferences inputs.
- `overrides.rs`: CLI dotted-path overrides → TOML “session flags” layer.
- `merge.rs`: recursive TOML merge.
- `fingerprint.rs`: stable per-layer hashing and per-key origins traversal.
- `macos.rs`: managed preferences integration (macOS only).

View File

@@ -1,67 +0,0 @@
use codex_app_server_protocol::ConfigLayerMetadata;
use serde_json::Value as JsonValue;
use sha2::Digest;
use sha2::Sha256;
use std::collections::HashMap;
use toml::Value as TomlValue;
pub(super) fn record_origins(
value: &TomlValue,
meta: &ConfigLayerMetadata,
path: &mut Vec<String>,
origins: &mut HashMap<String, ConfigLayerMetadata>,
) {
match value {
TomlValue::Table(table) => {
for (key, val) in table {
path.push(key.clone());
record_origins(val, meta, path, origins);
path.pop();
}
}
TomlValue::Array(items) => {
for (idx, item) in (0_i32..).zip(items.iter()) {
path.push(idx.to_string());
record_origins(item, meta, path, origins);
path.pop();
}
}
_ => {
if !path.is_empty() {
origins.insert(path.join("."), meta.clone());
}
}
}
}
pub(super) fn version_for_toml(value: &TomlValue) -> String {
let json = serde_json::to_value(value).unwrap_or(JsonValue::Null);
let canonical = canonical_json(&json);
let serialized = serde_json::to_vec(&canonical).unwrap_or_default();
let mut hasher = Sha256::new();
hasher.update(serialized);
let hash = hasher.finalize();
let hex = hash
.iter()
.map(|byte| format!("{byte:02x}"))
.collect::<String>();
format!("sha256:{hex}")
}
fn canonical_json(value: &JsonValue) -> JsonValue {
match value {
JsonValue::Object(map) => {
let mut sorted = serde_json::Map::new();
let mut keys = map.keys().cloned().collect::<Vec<_>>();
keys.sort();
for key in keys {
if let Some(val) = map.get(&key) {
sorted.insert(key, canonical_json(val));
}
}
JsonValue::Object(sorted)
}
JsonValue::Array(items) => JsonValue::Array(items.iter().map(canonical_json).collect()),
other => other.clone(),
}
}

View File

@@ -1,100 +0,0 @@
use super::LoaderOverrides;
use super::macos::load_managed_admin_config_layer;
use super::overrides::default_empty_table;
use crate::config::CONFIG_TOML_FILE;
use std::io;
use std::path::Path;
use std::path::PathBuf;
use tokio::fs;
use toml::Value as TomlValue;
#[cfg(unix)]
const CODEX_MANAGED_CONFIG_SYSTEM_PATH: &str = "/etc/codex/managed_config.toml";
#[derive(Debug, Clone)]
pub(super) struct LoadedConfigLayers {
pub base: TomlValue,
pub managed_config: Option<TomlValue>,
pub managed_preferences: Option<TomlValue>,
}
pub(super) async fn load_config_layers_internal(
codex_home: &Path,
overrides: LoaderOverrides,
) -> io::Result<LoadedConfigLayers> {
#[cfg(target_os = "macos")]
let LoaderOverrides {
managed_config_path,
managed_preferences_base64,
} = overrides;
#[cfg(not(target_os = "macos"))]
let LoaderOverrides {
managed_config_path,
} = overrides;
let managed_config_path =
managed_config_path.unwrap_or_else(|| managed_config_default_path(codex_home));
let user_config_path = codex_home.join(CONFIG_TOML_FILE);
let user_config = read_config_from_path(&user_config_path, true).await?;
let managed_config = read_config_from_path(&managed_config_path, false).await?;
#[cfg(target_os = "macos")]
let managed_preferences =
load_managed_admin_config_layer(managed_preferences_base64.as_deref()).await?;
#[cfg(not(target_os = "macos"))]
let managed_preferences = load_managed_admin_config_layer(None).await?;
Ok(LoadedConfigLayers {
base: user_config.unwrap_or_else(default_empty_table),
managed_config,
managed_preferences,
})
}
pub(super) async fn read_config_from_path(
path: &Path,
log_missing_as_info: bool,
) -> io::Result<Option<TomlValue>> {
match fs::read_to_string(path).await {
Ok(contents) => match toml::from_str::<TomlValue>(&contents) {
Ok(value) => Ok(Some(value)),
Err(err) => {
tracing::error!("Failed to parse {}: {err}", path.display());
Err(io::Error::new(io::ErrorKind::InvalidData, err))
}
},
Err(err) if err.kind() == io::ErrorKind::NotFound => {
if log_missing_as_info {
tracing::info!("{} not found, using defaults", path.display());
} else {
tracing::debug!("{} not found", path.display());
}
Ok(None)
}
Err(err) => {
tracing::error!("Failed to read {}: {err}", path.display());
Err(err)
}
}
}
/// 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;
PathBuf::from(CODEX_MANAGED_CONFIG_SYSTEM_PATH)
}
#[cfg(not(unix))]
{
codex_home.join("managed_config.toml")
}
}

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