Compare commits

..

59 Commits

Author SHA1 Message Date
Vincent Koc
6c375c7e90 fix(ci): prefer Anthropic OAuth in live validation 2026-06-16 19:12:19 +08:00
Vincent Koc
b1e925f6c5 fix(ci): support Anthropic OAuth release validation 2026-06-16 18:11:21 +08:00
Vincent Koc
cf9dea72a5 fix(release): satisfy retry delay lint 2026-06-16 16:12:06 +08:00
Vincent Koc
780373fd96 chore(release): prepare 2026.6.8 stable 2026-06-16 16:06:33 +08:00
Vincent Koc
11af11ad05 fix(release): tolerate npm propagation after publish 2026-06-16 15:49:04 +08:00
Vincent Koc
0c236ee246 fix(e2e): avoid Linux snapshot apt races 2026-06-16 15:18:55 +08:00
Vincent Koc
0c5478e58a fix(e2e): wait for Ubuntu package maintenance 2026-06-16 14:45:34 +08:00
Vincent Koc
e655247d9e fix(e2e): repair omitted Codex platform package 2026-06-16 14:32:05 +08:00
Vincent Koc
772bdeb18e fix(plugins): repair missing required platform packages 2026-06-16 14:01:13 +08:00
Vincent Koc
a24b0e3ce3 fix(e2e): retry macOS Parallels npm install 2026-06-16 13:19:13 +08:00
Vincent Koc
e48860e005 fix(e2e): budget macOS Parallels onboarding 2026-06-16 13:15:51 +08:00
Vincent Koc
abb3566ad2 fix(e2e): clear restored macOS npm cache 2026-06-16 12:41:15 +08:00
Vincent Koc
b1736723d7 fix(telegram): remove stale rich markdown argument 2026-06-16 07:20:07 +08:00
Vincent Koc
d13b000e65 chore(release): prepare 2026.6.8 beta.2 2026-06-16 07:13:11 +08:00
Ayaan Zaidi
955ba1f3a6 fix(telegram): preserve rich markdown line breaks 2026-06-16 07:03:25 +08:00
Vincent Koc
237ca5f832 fix(status): preserve rich message line breaks 2026-06-16 07:03:25 +08:00
Vincent Koc
7400417ab3 test(agents): mark runtime contract delivery as sent 2026-06-16 06:30:06 +08:00
Vincent Koc
fa2192b7e5 chore(release): refresh plugin SDK API baseline 2026-06-16 06:23:42 +08:00
Vincent Koc
d4feeb0039 fix(ci): update Vitest past browser advisory 2026-06-16 06:16:34 +08:00
Vincent Koc
4931a81ec2 test(agents): require delivery evidence in suppression fixtures 2026-06-16 06:16:34 +08:00
Shakker
3f24ebf7a9 fix: update dependency pins 2026-06-16 06:16:33 +08:00
Vincent Koc
09da6acc6f fix(release): align delivery evidence port 2026-06-16 05:59:19 +08:00
Vincent Koc
b3a0d62b09 fix(agents): distinguish delivery routes from references 2026-06-16 05:58:22 +08:00
Vincent Koc
e2b8e475f0 fix(agents): validate visible delivery receipts 2026-06-16 05:58:03 +08:00
Vincent Koc
2032a8a427 fix(agents): record visible delivery target evidence 2026-06-16 05:57:26 +08:00
Vincent Koc
96a27247f3 fix(cron): require explicit poll delivery targets 2026-06-16 03:12:04 +08:00
Vincent Koc
f20e1a921d style(gateway): format usage regression test 2026-06-16 02:56:09 +08:00
Vincent Koc
c78c9de2d8 fix(release): align delivery evidence with branch contracts 2026-06-16 02:53:19 +08:00
Vincent Koc
4fafd89926 fix(agents): preserve release CLI target evidence 2026-06-16 02:50:57 +08:00
Jason (Json)
63a58d9417 fix(tui): keep parent stdin paused after exit (#93159)
Keep the setup TUI parent stdin paused after its inherited-stdio child exits so Docker and PTY setup parents terminate cleanly. Align pre/post setup terminal cleanup with the cleanup-then-exit contract and add lifecycle regression coverage.

Thanks @fuller-stack-dev.

(cherry picked from commit 773ffd87a1)
2026-06-16 02:47:05 +08:00
Vincent Koc
c396935965 fix(agents): recognize bare-ok broadcast sends
(cherry picked from commit e3ca10438a)
2026-06-16 02:47:01 +08:00
Vincent Koc
086f43946f test(agents): align fallback delivery mock
(cherry picked from commit 68cbd1ae63)
2026-06-16 02:46:59 +08:00
Vincent Koc
e428117a37 fix(agents): preserve CLI message delivery evidence
(cherry picked from commit 0ea08076c3)
2026-06-16 02:46:54 +08:00
Vincent Koc
742f61ec5f fix(cron): expose safe explicit delivery context
(cherry picked from commit e573b751bf)
2026-06-15 16:46:33 +08:00
Vincent Koc
997ff477fd fix(agents): track normalized message target evidence
(cherry picked from commit 928b5932a3)
2026-06-15 15:10:52 +08:00
Vincent Koc
fc452cd034 fix(cron): preserve explicit delivery and timeout semantics
(cherry picked from commit efbefceb0e)
2026-06-15 15:00:10 +08:00
openclaw-clownfish[bot]
ea71a0f2b1 fix(gateway): repair usage cost aggregation across agents (#93022)
* fix(gateway): aggregate usage cost across agents

* fix(clownfish): address review for ghcrawl-157040-autonomous-smoke (1)

Co-authored-by: luke-skywalker-open-claw <262978557+luke-skywalker-open-claw@users.noreply.github.com>

Co-authored-by: Stable Genius <259448942+stablegenius49@users.noreply.github.com>

---------

Co-authored-by: openclaw-clownfish[bot] <280122609+openclaw-clownfish[bot]@users.noreply.github.com>
Co-authored-by: Stable Genius <259448942+stablegenius49@users.noreply.github.com>
(cherry picked from commit 4c2fef4a3b)
2026-06-15 14:33:10 +08:00
Mason Huang
2c056eeb65 fix(cron): require explicit message target proof (#92318)
Summary:
- Merged fix(cron): require explicit message target proof after ClawSweeper review.

Automerge notes:
- No ClawSweeper repair was needed after automerge opt-in.

Validation:
- ClawSweeper review passed for head 2aff537f9f.
- Required merge gates passed before the squash merge.

Prepared head SHA: 2aff537f9f
Review: https://github.com/openclaw/openclaw/pull/92318#issuecomment-4704342205

Co-authored-by: Mason Huang <masonxhuang@tencent.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
Approved-by: hxy91819
Co-authored-by: hxy91819 <8814856+hxy91819@users.noreply.github.com>
(cherry picked from commit 4fc805320f)
2026-06-15 14:33:10 +08:00
zengLingbiao
49535d525f fix(agents): prevent duplicate before-tool-call hooks (#93009)
Prevent duplicate `before_tool_call` execution when an already wrapped tool passes through schema normalization and coding-tool assembly. Preserve the normalized schema while replacing stale wrapper context with the current agent/session/run context.

Fixes #92973.

Co-authored-by: zengLingbiao <zeng.lingbiao@xydigit.com>
(cherry picked from commit 07dfdd4bd0)
2026-06-15 14:33:09 +08:00
Chunyue Wang
316ebc6eba fix(agents): do not misclassify client-disconnect abort as run timeout (#90936)
Summary:
- The PR adds an abort-signal-specific timeout classifier, switches two embedded attempt abort handlers to it, and adds focused failover tests.
- PR surface: Source +5, Tests +32. Total +37 across 3 files.
- Reproducibility: yes. from source inspection and a focused Node abort-reason check, but not from a live 180- ... ault AbortController abort reason through the broad timeout classifier used by the embedded abort handlers.

Automerge notes:
- PR branch already contained follow-up commit before automerge: fix(agents): do not misclassify client-disconnect abort as run timeout

Validation:
- ClawSweeper review passed for head 2708b0a37d.
- Required merge gates passed before the squash merge.

Prepared head SHA: 2708b0a37d
Review: https://github.com/openclaw/openclaw/pull/90936#issuecomment-4638919394

Co-authored-by: openperf <16864032@qq.com>
Co-authored-by: clawsweeper <274271284+clawsweeper[bot]@users.noreply.github.com>
Co-authored-by: clawsweeper[bot] <274271284+clawsweeper[bot]@users.noreply.github.com>
(cherry picked from commit fd80e0dd6b)
2026-06-15 14:33:09 +08:00
Vincent Koc
f3714e3a9d fix(release): support ClawHub-only runtime builds
(cherry picked from commit 94833b2c90)
2026-06-15 14:33:08 +08:00
Vincent Koc
98acc4125e fix(release): harden plugin package preflight
(cherry picked from commit 5b18b7560e)
2026-06-15 14:32:59 +08:00
Vincent Koc
a231ab8acf fix(e2e): resolve macOS Parallels VM 2026-06-15 08:27:48 +08:00
Vincent Koc
a7e0822a1a fix(e2e): resume restored Parallels snapshots 2026-06-15 08:05:39 +08:00
Vincent Koc
47ec5be9ef fix(docker): seed prune store from lockfile 2026-06-15 07:35:19 +08:00
Vincent Koc
43d00c7724 test(gateway): expect trajectory session branch bundle 2026-06-14 23:33:59 +08:00
Vincent Koc
7e6379377b test(gateway): allow optional trajectory metadata bundle 2026-06-14 23:24:33 +08:00
Vincent Koc
72d45dea92 fix(gateway): run export helpers through cli entry 2026-06-14 23:07:24 +08:00
Vincent Koc
3c43b59270 fix(gateway): async trajectory export approvals 2026-06-14 22:25:05 +08:00
Vincent Koc
714d894fba docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:22:27 +08:00
Vincent Koc
5482d94419 fix(webchat): route trajectory slash export before agent dispatch 2026-06-14 21:22:16 +08:00
Vincent Koc
52cefa06ae docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:10:21 +08:00
Vincent Koc
4dfc55e01d docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:09:19 +08:00
Vincent Koc
2fe9734208 docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:09:19 +08:00
Vincent Koc
f6f0e42af2 docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:09:19 +08:00
Vincent Koc
878a531b60 chore(release): refresh generated baseline hashes 2026-06-14 21:09:18 +08:00
Vincent Koc
ae1dcac871 docs(changelog): refresh 2026.6.8 notes 2026-06-14 21:09:18 +08:00
Vincent Koc
f157a4280c chore(release): refresh generated baseline hashes 2026-06-14 21:09:18 +08:00
Vincent Koc
d257d17261 chore(release): prepare 2026.6.8 beta 2026-06-14 21:09:18 +08:00
708 changed files with 23607 additions and 42456 deletions

View File

@@ -16,6 +16,15 @@ Use this with `$release-openclaw-maintainer` and `$openclaw-testing` when a rele
- Watch one parent run plus compact child summaries. Avoid broad `gh run view` polling loops; REST quota is easy to burn.
- Fetch logs only for failed or currently-blocking jobs. If quota is low, stop polling and wait for reset.
- Treat live-provider flakes separately from code failures: prove key validity, provider HTTP status, retry evidence, and exact failing lane before editing code.
- Anthropic release lanes support both API keys and OAuth. When API keys are
exhausted but a maintainer-owned OAuth token passes a live Anthropic probe,
set `ANTHROPIC_OAUTH_TOKEN` for provider/runtime lanes and
refreshable `OPENCLAW_CLAUDE_CREDENTIALS_JSON` or
`CLAUDE_CODE_OAUTH_TOKEN` for Claude CLI subscription lanes before rerunning
the matrix. Revalidate short-lived OAuth immediately before dispatch. Never
keep retrying a known exhausted API key. Live-cache validation must prefer
the proven OAuth token instead of leaving an exhausted API key first in the
runtime key pool.
- Full Release Validation parent monitors fail fast: once a required child job
fails, the parent cancels the remaining child matrix and prints the failed
job summary. Inspect that first red job instead of waiting for unrelated
@@ -36,6 +45,8 @@ git rev-parse HEAD
preflight. Inject those exact targeted keys first, then run the verifier; use
ambient env only when it was already intentionally injected for this release.
The script prints only provider status and HTTP class, never tokens.
For Anthropic it prefers `ANTHROPIC_OAUTH_TOKEN` and validates it with bearer
OAuth headers when present; otherwise it checks API-key-shaped credentials.
## Dispatch
@@ -107,6 +118,10 @@ Stop watchers before ending the turn or switching strategy.
```
3. Fetch one failed job log. If rate-limited, note reset time and avoid more REST calls.
4. For secret-looking failures, validate the provider endpoint from the same secret source before editing code.
For Docker CLI-backend failures, also validate
`OPENCLAW_CLAUDE_CREDENTIALS_JSON` or `CLAUDE_CODE_OAUTH_TOKEN` in a
clean-home Claude CLI probe; that lane should use subscription mode when
either credential exists.
5. For live-cache failures, inspect whether it is missing/invalid key, empty text, provider refusal, timeout, or baseline miss. Do not weaken release gates without clear provider evidence.
6. Fix narrowly, run local/changed proof, commit, push, rerun the smallest matching group.

View File

@@ -42,7 +42,7 @@ async function checkProvider(id, config) {
const controller = new AbortController();
const timer = setTimeout(() => controller.abort(), timeoutMs);
try {
const headers = config.headers(secret.value);
const headers = config.headers(secret);
const response = await fetch(config.url, {
headers,
signal: controller.signal,
@@ -69,25 +69,32 @@ const providers = {
openai: {
env: ["OPENAI_API_KEY"],
url: "https://api.openai.com/v1/models",
headers: (token) => ({ authorization: `Bearer ${token}` }),
headers: ({ value }) => ({ authorization: `Bearer ${value}` }),
},
anthropic: {
env: ["ANTHROPIC_API_KEY", "ANTHROPIC_API_TOKEN"],
env: ["ANTHROPIC_OAUTH_TOKEN", "ANTHROPIC_API_KEY", "ANTHROPIC_API_TOKEN"],
url: "https://api.anthropic.com/v1/models",
headers: (token) => ({
"anthropic-version": "2023-06-01",
"x-api-key": token,
}),
headers: ({ name, value }) =>
name === "ANTHROPIC_OAUTH_TOKEN"
? {
"anthropic-beta": "oauth-2025-04-20",
"anthropic-version": "2023-06-01",
authorization: `Bearer ${value}`,
}
: {
"anthropic-version": "2023-06-01",
"x-api-key": value,
},
},
fireworks: {
env: ["FIREWORKS_API_KEY"],
url: "https://api.fireworks.ai/inference/v1/models",
headers: (token) => ({ authorization: `Bearer ${token}` }),
headers: ({ value }) => ({ authorization: `Bearer ${value}` }),
},
openrouter: {
env: ["OPENROUTER_API_KEY"],
url: "https://openrouter.ai/api/v1/models",
headers: (token) => ({ authorization: `Bearer ${token}` }),
headers: ({ value }) => ({ authorization: `Bearer ${value}` }),
},
};

View File

@@ -321,6 +321,7 @@ Upgrade with the beta channel.
Before tagging or publishing, run:
```bash
pnpm release:fast-pretag-check
pnpm check:architecture
pnpm build
pnpm ui:build
@@ -329,6 +330,21 @@ pnpm release:check
pnpm test:install:smoke
```
- Treat `pnpm release:fast-pretag-check` as a hard packaging gate. Every
publishable plugin must have a non-empty package-root `README.md`, build its
package-local runtime, and pass the npm and ClawHub release metadata checks
before a tag or publish workflow can start. Do not defer README, entrypoint,
or packed-artifact failures to postpublish verification.
- Before tagging, require green CI for the exact release-candidate SHA, not an
earlier branch SHA. Heal every related red CI, release-check, packaging, or
root-Dockerfile lane on the release branch, forward-port the fix to `main`,
and rerun the affected exact-SHA gates. Never waive a red Docker lane because
npm preflight passed.
- Root Dockerfile proof is mandatory before every beta and stable tag. Run the
release `install-smoke` group or equivalent root Dockerfile build for the
exact candidate SHA and require it to pass. The tag-triggered Docker Release
workflow is post-tag publishing, not the first valid proof that the root
Dockerfile can build.
- Before tagging, diff publishable plugin package manifests against the last
reachable stable/beta release tag. For every newly publishable package
(`openclaw.release.publishToNpm: true` or `publishToClawHub: true`) whose
@@ -536,6 +552,16 @@ node --import tsx scripts/openclaw-npm-postpublish-verify.ts <published-version>
- `preflight_only=true` on the npm workflow is also the right way to validate an
existing tag after publish; it should keep running the build checks even when
the npm version is already published.
- npm registry metadata is eventually consistent immediately after trusted
publishing. Keep postpublish `npm view` checks on bounded `--prefer-online`
retries, and carry that verified tarball/integrity metadata into later proof
steps instead of reading the registry again. If the OpenClaw npm child
succeeded but the parent publish workflow failed on an immediate exact-version
`E404`, verify the exact version with a cache-bypassed registry read, run the
standalone postpublish verifier and the full beta verifier with the original
successful child run IDs, then finalize the draft, dependency evidence asset,
and release proof manually. Never rerun the publish workflow for that
already-published version.
- npm validation-only preflight may still be dispatched from ordinary branches
when testing workflow changes before merge. Release checks and real publish
use only `main` or `release/YYYY.M.PATCH`.
@@ -644,9 +670,10 @@ node --import tsx scripts/openclaw-npm-postpublish-verify.ts <published-version>
off, live OpenAI off, and regression failure off. Let it run in parallel
with preflight and validation work.
10. Run the fast local beta preflight from the release branch before any npm
preflight or publish. Keep expensive Docker, Parallels, and published-package
install/update lanes for after the beta is live unless the operator asks to
run them before beta publication.
preflight or publish. Require exact-SHA CI and root Dockerfile install-smoke
to be green before tagging. Keep the remaining expensive Docker, Parallels,
and published-package install/update lanes for after the beta is live unless
the operator asks to run them before beta publication.
11. For beta releases, skip mac app build/sign/notarize unless beta scope or a
release blocker specifically requires it. For stable releases, include the
mac app, signing, notarization, and appcast path.
@@ -703,8 +730,13 @@ node --import tsx scripts/openclaw-npm-postpublish-verify.ts <published-version>
waited plugin publish or Windows Hub promotion fails after OpenClaw npm
succeeds, the workflow keeps the release draft with OpenClaw npm evidence
and exits red; do not undraft until the gap is repaired. The standalone
verifier command remains the recovery probe:
verifier command remains the first recovery probe:
`node --import tsx scripts/openclaw-npm-postpublish-verify.ts <published-version>`.
For a failed postpublish parent after successful publish children, also run
`pnpm release:verify-beta -- <published-version> ... --skip-github-release`
with the original child run IDs and an evidence output path before manually
recreating the workflow's draft, dependency evidence asset, proof section,
and publish step.
25. Run the post-published beta verification roster. First scan current `main`
for critical fixes that landed after the release branch cut; backport only
important low-risk fixes before starting expensive lanes, or increment to

View File

@@ -210,6 +210,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
FIREWORKS_API_KEY: ${{ secrets.FIREWORKS_API_KEY }}

View File

@@ -128,6 +128,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}

View File

@@ -113,6 +113,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}

View File

@@ -1358,8 +1358,6 @@ jobs:
- check_name: check-additional-boundaries-bcd
group: boundaries
boundary_shard: 2/4,3/4,4/4
- check_name: check-session-accessor-boundary
group: session-accessor-boundary
- check_name: check-additional-extension-channels
group: extension-channels
- check_name: check-additional-extension-bundled
@@ -1506,15 +1504,6 @@ jobs:
boundaries)
node scripts/run-additional-boundary-checks.mjs
;;
session-accessor-boundary)
if [ ! -f scripts/check-session-accessor-boundary.mjs ]; then
echo "[skip] session accessor boundary check is not present in this checkout"
elif ! node -e 'const pkg = require("./package.json"); process.exit(pkg.scripts?.["lint:tmp:session-accessor-boundary"] ? 0 : 1);'; then
echo "[skip] session accessor boundary script is not present in package.json"
else
run_check "lint:tmp:session-accessor-boundary" pnpm run lint:tmp:session-accessor-boundary
fi
;;
extension-channels)
run_check "lint:extensions:channels" pnpm run lint:extensions:channels
;;

View File

@@ -663,6 +663,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}

View File

@@ -229,6 +229,8 @@ on:
required: false
ANTHROPIC_API_TOKEN:
required: false
ANTHROPIC_OAUTH_TOKEN:
required: false
FACTORY_API_KEY:
required: false
BYTEPLUS_API_KEY:
@@ -519,6 +521,7 @@ jobs:
env:
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
OPENCLAW_LIVE_CACHE_TEST: "1"
OPENCLAW_LIVE_TEST: "1"
steps:
@@ -541,10 +544,13 @@ jobs:
echo "Missing OPENAI_API_KEY secret for live-cache validation." >&2
exit 1
fi
if [[ -z "${ANTHROPIC_API_KEY:-}" ]]; then
echo "Missing ANTHROPIC_API_KEY secret for live-cache validation." >&2
if [[ -z "${ANTHROPIC_OAUTH_TOKEN:-}" && -z "${ANTHROPIC_API_KEY:-}" ]]; then
echo "Missing ANTHROPIC_OAUTH_TOKEN or ANTHROPIC_API_KEY secret for live-cache validation." >&2
exit 1
fi
if [[ -n "${ANTHROPIC_OAUTH_TOKEN:-}" ]]; then
echo "ANTHROPIC_API_KEY=" >> "$GITHUB_ENV"
fi
- name: Verify live prompt cache floors
run: |
@@ -680,6 +686,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
@@ -944,6 +951,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
@@ -1655,6 +1663,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
@@ -1746,7 +1755,7 @@ jobs:
}
case "${LIVE_MODEL_PROVIDERS}" in
anthropic) require_any Anthropic ANTHROPIC_API_KEY ANTHROPIC_API_KEY_OLD ANTHROPIC_API_TOKEN ;;
anthropic) require_any Anthropic ANTHROPIC_OAUTH_TOKEN ANTHROPIC_API_KEY ANTHROPIC_API_KEY_OLD ANTHROPIC_API_TOKEN ;;
google) require_any Google GEMINI_API_KEY GOOGLE_API_KEY ;;
minimax) require_any MiniMax MINIMAX_API_KEY ;;
moonshot) require_any Moonshot MOONSHOT_API_KEY KIMI_API_KEY ;;
@@ -1778,6 +1787,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
@@ -1921,7 +1931,7 @@ jobs:
IFS=',' read -r -a providers <<<"${OPENCLAW_LIVE_PROVIDERS}"
for provider in "${providers[@]}"; do
case "$provider" in
anthropic) require_any Anthropic ANTHROPIC_API_KEY ANTHROPIC_API_KEY_OLD ANTHROPIC_API_TOKEN ;;
anthropic) require_any Anthropic ANTHROPIC_OAUTH_TOKEN ANTHROPIC_API_KEY ANTHROPIC_API_KEY_OLD ANTHROPIC_API_TOKEN ;;
google) require_any Google GEMINI_API_KEY GOOGLE_API_KEY ;;
minimax) require_any MiniMax MINIMAX_API_KEY ;;
moonshot) require_any Moonshot MOONSHOT_API_KEY KIMI_API_KEY ;;
@@ -2140,6 +2150,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
@@ -2222,7 +2233,11 @@ jobs:
case "${{ matrix.suite_id }}" in
live-cli-backend-docker)
echo "OPENCLAW_LIVE_CLI_BACKEND_MODEL=claude-cli/claude-sonnet-4-6" >> "$GITHUB_ENV"
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=api-key" >> "$GITHUB_ENV"
if [[ -n "${OPENCLAW_CLAUDE_CREDENTIALS_JSON:-}" || -n "${CLAUDE_CODE_OAUTH_TOKEN:-}" ]]; then
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=subscription" >> "$GITHUB_ENV"
else
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=api-key" >> "$GITHUB_ENV"
fi
echo "OPENCLAW_LIVE_CLI_BACKEND_DEBUG=1" >> "$GITHUB_ENV"
echo "OPENCLAW_CLI_BACKEND_LOG_OUTPUT=1" >> "$GITHUB_ENV"
echo "OPENCLAW_TEST_CONSOLE=1" >> "$GITHUB_ENV"
@@ -2356,6 +2371,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
@@ -2447,7 +2463,11 @@ jobs:
case "${{ matrix.suite_id }}" in
live-cli-backend-docker)
echo "OPENCLAW_LIVE_CLI_BACKEND_MODEL=claude-cli/claude-sonnet-4-6" >> "$GITHUB_ENV"
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=api-key" >> "$GITHUB_ENV"
if [[ -n "${OPENCLAW_CLAUDE_CREDENTIALS_JSON:-}" || -n "${CLAUDE_CODE_OAUTH_TOKEN:-}" ]]; then
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=subscription" >> "$GITHUB_ENV"
else
echo "OPENCLAW_LIVE_CLI_BACKEND_AUTH=api-key" >> "$GITHUB_ENV"
fi
echo "OPENCLAW_LIVE_CLI_BACKEND_DEBUG=1" >> "$GITHUB_ENV"
echo "OPENCLAW_CLI_BACKEND_LOG_OUTPUT=1" >> "$GITHUB_ENV"
echo "OPENCLAW_TEST_CONSOLE=1" >> "$GITHUB_ENV"
@@ -2568,6 +2588,7 @@ jobs:
OPENAI_BASE_URL: ${{ secrets.OPENAI_BASE_URL }}
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}

View File

@@ -631,6 +631,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}
@@ -724,6 +725,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}

View File

@@ -1112,13 +1112,14 @@ jobs:
}
append_release_proof_to_github_release() {
local release_version body_file notes_file tarball integrity telegram_line clawhub_line clawhub_bootstrap_line clawhub_runtime_state_path windows_line
local release_version body_file notes_file evidence_path tarball integrity telegram_line clawhub_line clawhub_bootstrap_line clawhub_runtime_state_path windows_line
release_version="${RELEASE_TAG#v}"
body_file="${RUNNER_TEMP}/release-body.md"
notes_file="${RUNNER_TEMP}/release-notes-with-proof.md"
tarball="$(npm view "openclaw@${release_version}" dist.tarball --json | jq -r '.')"
integrity="$(npm view "openclaw@${release_version}" dist.integrity --json | jq -r '.')"
evidence_path="${POSTPUBLISH_EVIDENCE_DIR}/release-postpublish-evidence.json"
tarball="$(jq -er '.openclawNpmTarball | select(type == "string" and length > 0)' "${evidence_path}")"
integrity="$(jq -er '.openclawNpmIntegrity | select(type == "string" and length > 0)' "${evidence_path}")"
gh release view "${RELEASE_TAG}" --repo "$GITHUB_REPOSITORY" --json body --jq .body > "${body_file}"
if [[ -n "${NPM_TELEGRAM_RUN_ID// }" ]]; then

View File

@@ -38,6 +38,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}

View File

@@ -203,6 +203,8 @@ on:
required: false
ANTHROPIC_API_TOKEN:
required: false
ANTHROPIC_OAUTH_TOKEN:
required: false
FACTORY_API_KEY:
required: false
BYTEPLUS_API_KEY:
@@ -588,6 +590,7 @@ jobs:
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
ANTHROPIC_API_KEY_OLD: ${{ secrets.ANTHROPIC_API_KEY_OLD }}
ANTHROPIC_API_TOKEN: ${{ secrets.ANTHROPIC_API_TOKEN }}
ANTHROPIC_OAUTH_TOKEN: ${{ secrets.ANTHROPIC_OAUTH_TOKEN }}
FACTORY_API_KEY: ${{ secrets.FACTORY_API_KEY }}
BYTEPLUS_API_KEY: ${{ secrets.BYTEPLUS_API_KEY }}
CEREBRAS_API_KEY: ${{ secrets.CEREBRAS_API_KEY }}

View File

@@ -6,30 +6,31 @@ Docs: https://docs.openclaw.ai
### Highlights
- Telegram and WhatsApp channel delivery are richer and less brittle: Telegram can send structured rich text with tables, lists, expandable blockquotes, prompt-preserving CLI backend delivery, retired native draft migration, and safer rich-media boundaries, while WhatsApp now honors configured ACP bindings. (#92679, #84082, #89421, #92513) Thanks @obviyus, @jzakirov, @spacegeologist, and @TurboTheTurtle.
- Agent and Gateway recovery is sharper across account-scoped DM sends, generated media completions, restart shutdown aborts, yielded subagent pauses, yielded cron media, heartbeat dedupe, session identity prompts, and unknown OpenAI agent selector rejection. (#92788, #91246, #91357, #92631, #92146, #91287, #92468, #92510) Thanks @yetval, @TurboTheTurtle, @ooiuuii, @openperf, @IWhatsskill, @ZengWen-DT, and @zhangguiping-xydt.
- Provider/model handling expands and tightens with GLM-5.2, Claude Haiku 4.5 catalog rows, OpenRouter and Google Vertex provider-prefix normalization, managed SecretRef auth, bounded model browse discovery, storeless OpenAI Responses replay gating, and Claude 4.5 Copilot tool-streaming safety. (#92796, #90116, #92627, #91218, #90686, #92247, #90706, #75393) Thanks @arkyu2077, @liuhao1024, @bymle, @rohitjavvadi, @samson910022, @snowzlm, and @Kailigithub.
- Telegram and WhatsApp channel delivery are richer and less brittle: Telegram can send structured rich text with tables, lists, expandable blockquotes, preserved intentional line breaks, prompt-preserving CLI backend delivery, retired native draft migration, and safer rich-media boundaries, while WhatsApp now honors configured ACP bindings. (#92679, #93164, #84082, #89421, #92513) Thanks @obviyus, @jzakirov, @spacegeologist, and @TurboTheTurtle.
- Agent and Gateway recovery is sharper across account-scoped DM sends, generated media completions, auto-reply message-tool final replies, reset archive fallback reads, restart shutdown aborts, yielded subagent pauses, trusted subagent thinking override fallback, yielded cron media, heartbeat dedupe, session identity prompts, and unknown OpenAI agent selector rejection. (#92788, #91246, #92879, #91357, #92631, #92412, #92146, #91287, #92468, #92510) Thanks @yetval, @TurboTheTurtle, @masatohoshino, @CadanHu, @ooiuuii, @openperf, @IWhatsskill, @ZengWen-DT, and @zhangguiping-xydt.
- Provider/model handling expands and tightens with GLM-5.2, Claude Haiku 4.5 catalog rows, OpenRouter and Google Vertex provider-prefix normalization, managed SecretRef auth, OAuth image-default routing through Codex, bounded model browse discovery, LM Studio binary thinking-off delivery, storeless OpenAI Responses replay gating, invalid OpenAI reasoning-signature and genericized Anthropic thinking-signature recovery, Claude 4.5 Copilot tool-streaming safety, and OpenAI/Anthropic-family payload quarantine for unreadable or post-hook tool schemas. (#92796, #90116, #92627, #91218, #90686, #92824, #92247, #92002, #90706, #92941, #92201, #92916, #75393, #92908, #92921, #92928) Thanks @arkyu2077, @liuhao1024, @bymle, @rohitjavvadi, @nxmxbbd, @bek91, @samson910022, @mmyzwl, @CarlCapital, @snowzlm, @Kailigithub, and @vincentkoc.
- `/usage` and reply payload hooks now have a native full footer renderer, default template, fixed-decimal formatting, credential-aware limits, better partial-count handling, and warnings for broken templates instead of silent bad output. (#92657, #89835, #89629) Thanks @Marvinthebored.
- UI and mobile flows are steadier: workspace files can collapse and start collapsed, WebChat backscroll survives streaming, the sidebar session picker remains interactive above the desktop workbench, reset soft args survive UI dispatch, stale dashboard session parent lineage is preserved, and iOS reconnects stale foreground gateways. (#92779, #92622, #92705, #91353, #90658, #92552) Thanks @shakkernerd, @TurboTheTurtle, @NianJiuZst, @zhouhe-xydt, @luoyanglang, and @Solvely-Colin.
- Memory, state, and diagnostics recover cleaner: oversized OpenAI embedding batches split before 431s, QMD memory search stays available in transient mode, SQLite avoids WAL on NFS state volumes, stuck-session recovery scheduling no longer resets warning backoff, and Infinity chunk limits stay genuinely unbounded. (#92650, #92618, #92639, #91247, #92752, #92735) Thanks @mushuiyu886, @TurboTheTurtle, @849261680, @gnanam1990, and @yhterrance.
- Memory, state, and diagnostics recover cleaner: oversized OpenAI embedding batches split before 431s, QMD memory search stays available in transient mode, SQLite avoids WAL on NFS state volumes, stuck-session recovery scheduling no longer resets warning backoff, full memory reindexes preserve rollback/cache recovery, raw Memory Wiki source pages stop looking malformed, and Infinity chunk limits stay genuinely unbounded. (#92650, #92618, #92639, #91247, #92752, #92881, #59137, #92876, #69700, #92735) Thanks @mushuiyu886, @TurboTheTurtle, @849261680, @gnanam1990, @TSHOGX, @arlen8411, and @yhterrance.
### Changes
- Providers/models: add GLM-5.2 support and Claude Haiku 4.5 catalog entries while keeping provider-qualified model IDs normalized across OpenRouter and Google Vertex paths. (#92796, #90116, #92627, #91218) Thanks @arkyu2077, @liuhao1024, and @bymle.
- Channel plugins: ship Telegram rich-message delivery and WhatsApp ACP binding support, including rich prompt handoff to CLI backends and transport fixtures for richer drafts. (#92679, #92513) Thanks @obviyus and @TurboTheTurtle.
- Channel plugins: ship Telegram rich-message delivery and WhatsApp ACP binding support, including preserved intentional line breaks, rich prompt handoff to CLI backends, and transport fixtures for richer drafts. (#92679, #93164, #92513) Thanks @obviyus and @TurboTheTurtle.
- Agent commands: support `/btw` in CLI-backed sessions and keep CLI usage-error exits classified as usage failures instead of successful runs. (#92669, #92162) Thanks @joshavant and @Pandah97.
- Usage hooks: add built-in full footer rendering, default footer templates, per-turn usage state, credential-aware limits, and fixed-decimal formatting for usage-bar templates. (#92657, #89835, #89629) Thanks @Marvinthebored.
- Docs and operator guidance: document node config examples, clarify before-install hook scope, correct agent default concurrency comments, refresh ZAI provider docs, and update channel/group docs for current Telegram and WhatsApp behavior. (#92677, #92766, #92695) Thanks @liuhao1024, @sallyom, and @ArielSmoliar.
### Fixes
- Channels and delivery: preserve account-scoped DM channel send policy, rich Telegram final replies, rich Telegram tables and lists, Telegram thread-create CLI remapping, Slack outbound `message_sent` hooks, contributed message-tool schema optionality, same-channel generated media completions, and channel chunking around surrogate pairs and Infinity limits. (#92788, #92679, #89421, #89943, #91137, #91246, #92735) Thanks @yetval, @obviyus, @spacegeologist, @rishitamrakar, @lundog, @TurboTheTurtle, and @yhterrance.
- Channels and delivery: preserve account-scoped DM channel send policy, intentional rich-message line breaks in Telegram and status output, rich Telegram final replies, rich Telegram tables and lists, Telegram thread-create CLI remapping, Feishu dynamic-agent routes after persisted binding reuse, Slack outbound `message_sent` hooks, contributed message-tool schema optionality, same-channel generated media completions, and channel chunking around surrogate pairs and Infinity limits. (#92788, #93164, #92679, #89421, #89943, #42837, #92814, #91137, #91246, #92735) Thanks @yetval, @obviyus, @spacegeologist, @rishitamrakar, @liuhao1024, @lundog, @TurboTheTurtle, and @yhterrance.
- Discord: give generated auto-thread titles a 60-second timeout and 4,096-token reasoning-model output budget, clamped to the selected model output cap. (#64734) Thanks @hanamizuki.
- Agent, cron, and Gateway runtime: mark active main sessions before restart shutdown aborts, pause yielded subagent runs whose terminal also signals abort, preserve yielded media completions, de-duplicate main-session heartbeat events, expose session identity in runtime prompts, reject unknown OpenAI agent selectors, keep generated media completions and slash-command block replies in WebChat, and require admin privileges for HTTP session/model override surfaces. (#91357, #92631, #92146, #91287, #92468, #92510, #91246, #92651, #92646) Thanks @ooiuuii, @openperf, @IWhatsskill, @ZengWen-DT, @zhangguiping-xydt, and @TurboTheTurtle.
- Providers and model replay: preserve storeless OpenAI Responses replay compatibility, avoid eager tool streaming for Claude 4.5 in Copilot, honor profile auth for SecretRef model entries, bound model browsing, strip provider prefixes where runtimes need bare IDs, and surface nested embedding fetch failures. (#90706, #75393, #90686, #92247, #92627, #91218, #92628) Thanks @snowzlm, @Kailigithub, @rohitjavvadi, @samson910022, @liuhao1024, @bymle, and @mushuiyu886.
- Memory, state, diagnostics, and config: split header-too-large embedding batches, keep QMD memory search enabled in transient mode, avoid SQLite WAL on NFS volumes, preserve recovery scheduling outside stuck-session warning backoff, and keep shell environment fallbacks contained in config write tests. (#92650, #92618, #92639, #91247, #92752) Thanks @mushuiyu886, @TurboTheTurtle, @849261680, and @gnanam1990.
- UI/mobile/TUI: preserve dashboard session parent lineage, WebChat backscroll, reset soft command args, sidebar session picker interactivity, collapsed workspace files, resolved `/model` confirmation refs, and stale foreground iOS Gateway reconnects. (#90658, #92622, #91353, #92705, #92779, #92773, #92552) Thanks @luoyanglang, @TurboTheTurtle, @zhouhe-xydt, @NianJiuZst, @shakkernerd, @NarahariRaghava, and @Solvely-Colin.
- Release and test reliability: extend slow Gateway/full-suite watchdogs, split local full-suite shards when throttled, stabilize plugin auth marker fixtures, avoid brittle provider-ref error text, and keep QA Lab bootstrap selection assertions aligned with flow-only scenarios. (#92652)
- Agent, cron, and Gateway runtime: mark active main sessions before restart shutdown aborts, pause yielded subagent runs whose terminal also signals abort, clamp trusted subagent thinking overrides through provider/model fallback, preserve yielded media completions, deliver channel message-tool final replies through auto-reply while hiding internal delivery hints, restore reset archive fallback reads when active async transcripts are missing, de-duplicate main-session heartbeat events, expose session identity in runtime prompts, reject unknown OpenAI agent selectors, keep generated media completions, slash-command block replies, and trajectory export commands in WebChat, and require admin privileges for HTTP session/model override surfaces. (#91357, #92631, #92412, #92146, #92879, #91287, #92468, #92510, #91246, #92651, #92646) Thanks @ooiuuii, @openperf, @IWhatsskill, @masatohoshino, @CadanHu, @ZengWen-DT, @zhangguiping-xydt, and @TurboTheTurtle.
- Providers and model replay: preserve storeless OpenAI Responses replay compatibility, recover invalid OpenAI reasoning signatures and genericized Anthropic thinking-signature replay errors, route OAuth image defaults through Codex for eligible OpenAI profiles, avoid eager tool streaming for Claude 4.5 in Copilot, quarantine unreadable and post-hook OpenAI/Anthropic-family tool schemas without broadening allowed tool choices, deliver explicit thinking-off requests to LM Studio binary-thinking models, honor profile auth for SecretRef model entries, bound model browsing, strip provider prefixes where runtimes need bare IDs, and surface nested embedding fetch failures. (#90706, #92941, #92201, #92916, #92824, #75393, #92908, #92921, #92928, #92002, #90686, #92247, #92627, #91218, #92628) Thanks @snowzlm, @mmyzwl, @CarlCapital, @bek91, @Kailigithub, @vincentkoc, @rohitjavvadi, @samson910022, @nxmxbbd, @liuhao1024, @bymle, and @mushuiyu886.
- Memory, state, diagnostics, and config: split header-too-large embedding batches, keep QMD memory search enabled in transient mode, avoid SQLite WAL on NFS volumes, preserve recovery scheduling outside stuck-session warning backoff, preserve full-reindex rollback/cache recovery, treat raw Memory Wiki source pages as source evidence, and keep shell environment fallbacks contained in config write tests. (#92650, #92618, #92639, #91247, #92752, #92881, #59137, #92876, #69700) Thanks @mushuiyu886, @TurboTheTurtle, @849261680, @gnanam1990, @TSHOGX, and @arlen8411.
- UI/mobile/TUI: preserve dashboard session parent lineage, WebChat backscroll, reset soft command args, sidebar session picker interactivity, collapsed workspace files, resolved `/model` confirmation refs, stale foreground iOS Gateway reconnects, and paused setup-parent stdin after inherited-stdio child exit. (#90658, #92622, #91353, #92705, #92779, #92773, #92552, #93159) Thanks @luoyanglang, @TurboTheTurtle, @zhouhe-xydt, @NianJiuZst, @shakkernerd, @NarahariRaghava, @Solvely-Colin, and @fuller-stack-dev.
- Plugins and updates: repair missing required platform packages during managed plugin installs and updates, including omitted Codex platform binaries.
- Release and test reliability: extend slow Gateway/full-suite watchdogs, split local full-suite shards when throttled, stabilize plugin auth marker fixtures, avoid brittle provider-ref error text, fold Telegram RTT sampling into live QA evidence, simplify QA scorecard mappings around canonical coverage IDs, keep QA Lab bootstrap selection assertions aligned with flow-only scenarios, skip QA coverage artifact consumers when runtime parity producer status is not green, keep Feishu lifecycle release checks pointed at the active fixture config, isolate trajectory-export live seed turns from Codex-native shell approvals, preserve release-check child refs while pinning expected SHAs, widen live OpenAI TTS budgets for slower provider responses, and avoid false downgrade prompts for unresolved latest-tag updates. (#92652, #92550, #92558, #92911) Thanks @RomneyDa and @Andy312432.
## 2026.6.6

View File

@@ -138,7 +138,7 @@ ARG OPENCLAW_BUNDLED_PLUGIN_DIR
# BuildKit cache mounts are not part of cached layers; seed tarballs for the
# installed prod graph in the same step that runs offline prune.
RUN --mount=type=cache,id=openclaw-pnpm-store,target=/root/.local/share/pnpm/store,sharing=locked \
pnpm list --prod --depth Infinity --json | node scripts/list-prod-store-packages.mjs | xargs -r pnpm store add && \
node scripts/list-prod-store-packages.mjs | xargs -r pnpm store add && \
CI=true pnpm prune --prod \
--config.offline=true \
--config.supportedArchitectures.os=linux \

View File

@@ -65,8 +65,8 @@ android {
applicationId = "ai.openclaw.app"
minSdk = 31
targetSdk = 36
versionCode = 2026060201
versionName = "2026.6.2"
versionCode = 2026060801
versionName = "2026.6.8"
ndk {
// Support all major ABIs — native libs are tiny (~47 KB per ABI)
abiFilters += listOf("armeabi-v7a", "arm64-v8a", "x86", "x86_64")

View File

@@ -1,5 +1,9 @@
# OpenClaw iOS Changelog
## 2026.6.8 - 2026-06-14
Maintenance update for the current OpenClaw beta release.
## 2026.6.2 - 2026-06-02
OpenClaw is now available on iPhone.

View File

@@ -2,8 +2,8 @@
// Source of truth: apps/ios/version.json
// Generated by scripts/ios-sync-versioning.ts.
OPENCLAW_IOS_VERSION = 2026.6.2
OPENCLAW_MARKETING_VERSION = 2026.6.2
OPENCLAW_IOS_VERSION = 2026.6.8
OPENCLAW_MARKETING_VERSION = 2026.6.8
OPENCLAW_BUILD_VERSION = 1
#include? "../build/Version.xcconfig"

View File

@@ -1,3 +1 @@
OpenClaw is now available on iPhone.
Connect to your OpenClaw Gateway to chat with your assistant, use realtime Talk mode, review approvals, share content from iOS, and bring device capabilities like camera, location, screen, and notifications into your private automation workflows.
Maintenance update for the current OpenClaw beta release.

View File

@@ -1,3 +1,3 @@
{
"version": "2026.6.2"
"version": "2026.6.8"
}

View File

@@ -15,9 +15,9 @@
<key>CFBundlePackageType</key>
<string>APPL</string>
<key>CFBundleShortVersionString</key>
<string>2026.6.2</string>
<string>2026.6.8</string>
<key>CFBundleVersion</key>
<string>2026060200</string>
<string>2026060800</string>
<key>CFBundleIconFile</key>
<string>OpenClaw</string>
<key>CFBundleURLTypes</key>

View File

@@ -2074,204 +2074,6 @@ public struct SessionsCompactionRestoreResult: Codable, Sendable {
}
}
public struct SessionFileBrowserEntry: Codable, Sendable {
public let path: String
public let name: String
public let kind: AnyCodable
public let sessionkind: SessionFileRelevance?
public let size: Int?
public let updatedatms: Int?
public init(
path: String,
name: String,
kind: AnyCodable,
sessionkind: SessionFileRelevance?,
size: Int?,
updatedatms: Int?)
{
self.path = path
self.name = name
self.kind = kind
self.sessionkind = sessionkind
self.size = size
self.updatedatms = updatedatms
}
private enum CodingKeys: String, CodingKey {
case path
case name
case kind
case sessionkind = "sessionKind"
case size
case updatedatms = "updatedAtMs"
}
}
public struct SessionFileBrowserResult: Codable, Sendable {
public let path: String
public let parentpath: String?
public let search: String?
public let entries: [SessionFileBrowserEntry]
public let truncated: Bool?
public init(
path: String,
parentpath: String?,
search: String?,
entries: [SessionFileBrowserEntry],
truncated: Bool?)
{
self.path = path
self.parentpath = parentpath
self.search = search
self.entries = entries
self.truncated = truncated
}
private enum CodingKeys: String, CodingKey {
case path
case parentpath = "parentPath"
case search
case entries
case truncated
}
}
public struct SessionFileEntry: Codable, Sendable {
public let path: String
public let name: String
public let kind: SessionFileKind
public let missing: Bool
public let size: Int?
public let updatedatms: Int?
public let content: String?
public init(
path: String,
name: String,
kind: SessionFileKind,
missing: Bool,
size: Int?,
updatedatms: Int?,
content: String?)
{
self.path = path
self.name = name
self.kind = kind
self.missing = missing
self.size = size
self.updatedatms = updatedatms
self.content = content
}
private enum CodingKeys: String, CodingKey {
case path
case name
case kind
case missing
case size
case updatedatms = "updatedAtMs"
case content
}
}
public struct SessionsFilesListParams: Codable, Sendable {
public let sessionkey: String
public let agentid: String?
public let path: String?
public let search: String?
public init(
sessionkey: String,
agentid: String? = nil,
path: String?,
search: String?)
{
self.sessionkey = sessionkey
self.agentid = agentid
self.path = path
self.search = search
}
private enum CodingKeys: String, CodingKey {
case sessionkey = "sessionKey"
case agentid = "agentId"
case path
case search
}
}
public struct SessionsFilesListResult: Codable, Sendable {
public let sessionkey: String
public let root: String?
public let files: [SessionFileEntry]
public let browser: SessionFileBrowserResult?
public init(
sessionkey: String,
root: String?,
files: [SessionFileEntry],
browser: SessionFileBrowserResult?)
{
self.sessionkey = sessionkey
self.root = root
self.files = files
self.browser = browser
}
private enum CodingKeys: String, CodingKey {
case sessionkey = "sessionKey"
case root
case files
case browser
}
}
public struct SessionsFilesGetParams: Codable, Sendable {
public let sessionkey: String
public let path: String
public let agentid: String?
public init(
sessionkey: String,
path: String,
agentid: String? = nil)
{
self.sessionkey = sessionkey
self.path = path
self.agentid = agentid
}
private enum CodingKeys: String, CodingKey {
case sessionkey = "sessionKey"
case path
case agentid = "agentId"
}
}
public struct SessionsFilesGetResult: Codable, Sendable {
public let sessionkey: String
public let root: String?
public let file: SessionFileEntry
public init(
sessionkey: String,
root: String?,
file: SessionFileEntry)
{
self.sessionkey = sessionkey
self.root = root
self.file = file
}
private enum CodingKeys: String, CodingKey {
case sessionkey = "sessionKey"
case root
case file
}
}
public struct SessionsCreateParams: Codable, Sendable {
public let key: String?
public let agentid: String?

View File

@@ -1,2 +1,2 @@
b121079a0912b3051a9fc319a675ef920da9db23364ca0c0ccd3c9f0a05a3a49 plugin-sdk-api-baseline.json
61a0108da670e0f44ba4b861c002eb6eaa5cf63e392d4e7e7de42044cbe7d115 plugin-sdk-api-baseline.jsonl
5500f1695f29e18f52f4c9c19fbf1093613a4e4596a3a32619fe7c9da68828fe plugin-sdk-api-baseline.json
0b58595cee12ac1330e0d5b07eb09954ad6959118b7d19d2abebbf5158b3f0a1 plugin-sdk-api-baseline.jsonl

View File

@@ -311,9 +311,7 @@ $OPENCLAW_STATE_DIR/tasks/runs.sqlite
The registry loads into memory at gateway start and syncs writes to SQLite for durability across restarts.
The Gateway keeps the SQLite write-ahead log bounded by using SQLite's default
autocheckpoint threshold plus periodic `PASSIVE` checkpoints. Shutdown and
explicit maintenance checkpoints still use `TRUNCATE` so normal closes can
reclaim WAL space without making the background sweeper wait on active readers.
autocheckpoint threshold plus periodic and shutdown `TRUNCATE` checkpoints.
### Automatic maintenance

View File

@@ -143,39 +143,12 @@ The native Codex app-server harness supports context engines that require
pre-prompt assembly. Generic CLI backends, including `codex-cli`, do not provide
that host capability.
Codex thread bindings live in OpenClaw's SQLite plugin state and use the stable
agent-scoped OpenClaw session key, or an opaque conversation-binding id, as
their owner. Physical session ids fence delayed cleanup but may rotate without
losing the Codex thread. Context-engine compaction adopts the successor id
before continuing native Codex compaction. The bounded store rejects a new
binding at its safety limit instead of evicting an existing thread's continuity
record.
Conversation binds create or resume their Codex thread on the first bound
message after channel approval; an abandoned approval consumes no thread row.
That first message carries the prepared thread directly into its turn.
Subsequent messages use a metadata-only resume to subscribe the shared client,
then unsubscribe after the turn completes.
The runtime does not poll transcript-adjacent binding files. Upgrades from
releases that used `*.jsonl.codex-app-server.json` sidecars migrate them during
normal startup preflight. `openclaw doctor --fix` can run the same migration
manually.
Successfully matched sidecars are archived before the new runtime resumes their
threads. Migration imports durable thread ownership only; it does not infer
Codex context usage from OpenClaw counters or crawl Codex rollout files. For
agent-session harness bindings, the next resume attempts to restore a cached
native snapshot when Codex has one, and ongoing turns persist the current-context
usage reported by app-server notifications, not the cumulative thread lifetime
total. Conversation bindings
keep metadata-only resumes and leave continuity and compaction with the native
Codex thread. Conflicting or ambiguous sidecars stay in place with a warning for
operator review.
For Codex-backed agents, `/compact` starts native Codex app-server compaction on
the bound thread. OpenClaw bounds the request-acceptance RPC but does not wait
for compaction completion, restart the shared app-server, or fall back to a
context-engine or public OpenAI summarizer. If the native Codex thread binding
is missing or stale, the command fails closed so the operator sees the real
runtime boundary instead of silently switching compaction backends.
the bound thread. OpenClaw does not wait for completion, impose an OpenClaw
timeout, restart the shared app-server, or fall back to a context-engine or
public OpenAI summarizer. If the native Codex thread binding is missing or
stale, the command fails closed so the operator sees the real runtime boundary
instead of silently switching compaction backends.
```json5
{

View File

@@ -1278,6 +1278,7 @@ Important examples:
| `openclaw.compat.pluginApi` | Minimum OpenClaw plugin API range required by this package, using a semver floor like `>=2026.5.27`. |
| `openclaw.install.expectedIntegrity` | Expected npm dist integrity string such as `sha512-...`; install and update flows verify the fetched artifact against it. |
| `openclaw.install.allowInvalidConfigRecovery` | Allows a narrow bundled-plugin reinstall recovery path when config is invalid. |
| `openclaw.install.requiredPlatformPackages` | npm package aliases that must materialize when their lockfile platform constraints match the current host. |
| `openclaw.startup.deferConfiguredChannelFullLoadUntilAfterListen` | Lets setup-runtime channel surfaces load before listen, then defers the full configured channel plugin until post-listen activation. |
Manifest metadata decides which provider/channel/setup choices appear in
@@ -1290,6 +1291,13 @@ registry loading for non-bundled plugin sources. Invalid values are rejected;
newer-but-valid values skip external plugins on older hosts. Bundled source
plugins are assumed to be co-versioned with the host checkout.
`openclaw.install.requiredPlatformPackages` is for npm packages that expose
required native binaries through optional, platform-specific aliases. List the
bare npm package name for every supported platform alias. During npm install,
OpenClaw verifies only the declared alias whose lockfile constraints match the
current host. If npm reports success but omits that alias, OpenClaw retries once
with a fresh cache and rolls back the install if the alias is still missing.
`openclaw.compat.pluginApi` is enforced during package install for non-bundled
plugin sources. Use it for the OpenClaw plugin SDK/runtime API floor that the
package was built against. It can be stricter than `minHostVersion` when a

View File

@@ -163,6 +163,7 @@ Example:
| `minHostVersion` | `string` | Minimum supported OpenClaw version in the form `>=x.y.z` or `>=x.y.z-prerelease`. |
| `expectedIntegrity` | `string` | Expected npm dist integrity string, usually `sha512-...`, for pinned installs. |
| `allowInvalidConfigRecovery` | `boolean` | Lets bundled-plugin reinstall flows recover from specific stale-config failures. |
| `requiredPlatformPackages` | `string[]` | Required platform-specific npm aliases verified during npm install. |
<AccordionGroup>
<Accordion title="Onboarding behavior">

View File

@@ -20,7 +20,6 @@ Scope includes:
- Thinking signature cleanup
- Image payload sanitization
- Blank text-block cleanup before provider replay
- Incomplete reasoning-only length-turn cleanup before provider replay
- User-input provenance tagging (for inter-session routed prompts)
- Empty assistant error-turn repair for Bedrock Converse replay
@@ -92,21 +91,6 @@ Implementation:
---
## Global rule: incomplete reasoning-only turns
Assistant turns that hit the provider output limit with only thinking or
redacted-thinking content are omitted from the in-memory replay copy. Such turns
contain incomplete provider state and may carry a partial thinking signature.
Empty length turns remain unchanged, as do length turns with visible text, tool
calls, or unknown content blocks. Stored transcripts are not rewritten.
Implementation:
- `normalizeAssistantReplayContent` in `src/agents/embedded-agent-runner/replay-history.ts`
---
## Global rule: inter-session input provenance
When an agent sends a prompt into another session via `sessions_send` (including

View File

@@ -13,12 +13,7 @@ CLI, and scripting patterns (snapshots, refs, waits, debug flows).
## Control API (optional)
For local integrations only, the Gateway exposes a small loopback HTTP API.
This standalone server is opt-in — set the environment variable
`OPENCLAW_EAGER_BROWSER_CONTROL_SERVER=1` in the gateway service environment
and restart the gateway before the HTTP endpoints become available. Without
this variable the browser control runtime still works through the CLI and
agent tools, but nothing listens on the loopback control port.
For local integrations only, the Gateway exposes a small loopback HTTP API:
- Status/start/stop: `GET /`, `POST /start`, `POST /stop`
- Tabs: `GET /tabs`, `POST /tabs/open`, `POST /tabs/focus`, `DELETE /tabs/:targetId`

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/acpx",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/acpx",
"version": "2026.6.2",
"version": "2026.6.8",
"dependencies": {
"@agentclientprotocol/claude-agent-acp": "0.39.0",
"@zed-industries/codex-acp": "0.15.0",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/acpx",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw ACP runtime backend with plugin-owned session and transport management.",
"repository": {
"type": "git",
@@ -26,10 +26,10 @@
"minHostVersion": ">=2026.4.25"
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2",
"openclawVersion": "2026.6.8",
"staticAssets": [
{
"source": "./src/runtime-internals/mcp-proxy.mjs",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/admin-http-rpc",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw admin HTTP RPC endpoint",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/alibaba-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Alibaba Model Studio video provider plugin",
"type": "module",

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/amazon-bedrock-mantle-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/amazon-bedrock-mantle-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"dependencies": {
"@anthropic-ai/sdk": "0.100.1",
"@aws/bedrock-token-generator": "1.1.0"

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/amazon-bedrock-mantle-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Amazon Bedrock Mantle provider plugin for OpenAI-compatible model routing.",
"repository": {
"type": "git",
@@ -24,10 +24,10 @@
"minHostVersion": ">=2026.5.12-beta.1"
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2",
"openclawVersion": "2026.6.8",
"bundledDist": false
},
"release": {

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/amazon-bedrock-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/amazon-bedrock-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"dependencies": {
"@aws-sdk/client-bedrock": "3.1056.0",
"@aws-sdk/client-bedrock-runtime": "3.1056.0",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/amazon-bedrock-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Amazon Bedrock provider plugin with model discovery, embeddings, and guardrail support.",
"repository": {
"type": "git",
@@ -28,10 +28,10 @@
"minHostVersion": ">=2026.5.12-beta.1"
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2",
"openclawVersion": "2026.6.8",
"bundledDist": false
},
"release": {

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/anthropic-vertex-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/anthropic-vertex-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"dependencies": {
"@anthropic-ai/vertex-sdk": "0.16.1"
}

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/anthropic-vertex-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Anthropic Vertex provider plugin for Claude models on Google Vertex AI.",
"repository": {
"type": "git",
@@ -23,10 +23,10 @@
"minHostVersion": ">=2026.5.12-beta.1"
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2",
"openclawVersion": "2026.6.8",
"bundledDist": false
},
"release": {

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/anthropic-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Anthropic provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/arcee-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Arcee provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/azure-speech",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Azure Speech plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/bonjour",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Bonjour/mDNS gateway discovery",
"type": "module",
"dependencies": {

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/brave-plugin",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/brave-plugin",
"version": "2026.6.2"
"version": "2026.6.8"
}
}
}

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/brave-plugin",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Brave Search provider plugin for web search.",
"repository": {
"type": "git",
@@ -21,10 +21,10 @@
"allowInvalidConfigRecovery": true
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2"
"openclawVersion": "2026.6.8"
},
"release": {
"publishToClawHub": true,

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/browser-plugin",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw browser tool plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/byteplus-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw BytePlus provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/canvas-plugin",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Canvas plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/cerebras-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Cerebras provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/chutes-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Chutes.ai provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/clickclack",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw ClickClack channel plugin",
"type": "module",
@@ -18,7 +18,7 @@
"openclaw": "2026.5.28"
},
"peerDependencies": {
"openclaw": ">=2026.6.2"
"openclaw": ">=2026.6.8"
},
"peerDependenciesMeta": {
"openclaw": {

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/cloudflare-ai-gateway-provider",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Cloudflare AI Gateway provider plugin",
"type": "module",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/codex-supervisor",
"version": "2026.6.2",
"version": "2026.6.8",
"private": true,
"description": "OpenClaw Codex app-server fleet supervision plugin.",
"type": "module",

View File

@@ -1,44 +1,6 @@
// Codex tests cover doctor contract api plugin behavior.
import fs from "node:fs/promises";
import os from "node:os";
import path from "node:path";
import type { PluginStateKeyedStore } from "openclaw/plugin-sdk/plugin-state-runtime";
import {
createPluginStateKeyedStoreForTests,
resetPluginStateStoreForTests,
} from "openclaw/plugin-sdk/plugin-state-test-runtime";
import type {
OpenKeyedStoreOptions,
PluginDoctorStateMigrationContext,
} from "openclaw/plugin-sdk/runtime-doctor";
import { afterEach, describe, expect, it } from "vitest";
import {
legacyConfigRules,
normalizeCompatibilityConfig,
stateMigrations,
} from "./doctor-contract-api.js";
import {
bindingStoreKey,
CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
CODEX_APP_SERVER_BINDING_NAMESPACE,
type StoredCodexAppServerBinding,
} from "./src/app-server/session-binding.js";
import { legacyCodexConversationBindingId } from "./src/conversation-binding-data.js";
function createDoctorContext(env: NodeJS.ProcessEnv): PluginDoctorStateMigrationContext {
return {
openPluginStateKeyedStore<T>(options: OpenKeyedStoreOptions) {
return createPluginStateKeyedStoreForTests<T>("codex", {
...options,
env: options.env ?? env,
});
},
};
}
afterEach(() => {
resetPluginStateStoreForTests();
});
import { describe, expect, it } from "vitest";
import { legacyConfigRules, normalizeCompatibilityConfig } from "./doctor-contract-api.js";
describe("codex doctor contract", () => {
it("reports the retired dynamic tools profile config key", () => {
@@ -80,856 +42,4 @@ describe("codex doctor contract", () => {
});
expect(original.plugins.entries.codex.config).toHaveProperty("codexDynamicToolsProfile");
});
it("imports shipped binding sidecars under session and legacy conversation identities", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "agents", "main", "sessions");
const transcriptPath = path.join(sessionsDir, "session-current.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
const legacyBinding = {
schemaVersion: 1,
threadId: "thread-1",
sessionFile: transcriptPath,
updatedAt: "2026-01-01T00:00:00.000Z",
};
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"session-current"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:session-1": {
sessionId: "session-current",
sessionFile: "session-current.jsonl",
totalTokens: 42_000,
totalTokensFresh: true,
contextTokens: 258_400,
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(sidecarPath, JSON.stringify(legacyBinding), "utf8");
const params = {
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
};
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(migration.detectLegacyState(params)).resolves.toMatchObject({
preview: [expect.stringContaining("legacy sidecar")],
});
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-current",
sessionKey: "agent:main:session-1",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-current",
binding: { threadId: "thread-1" },
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
}),
),
).resolves.toMatchObject({
state: "active",
binding: {
threadId: "thread-1",
cwd: "",
historyCoveredThrough: expect.any(String),
},
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
}),
),
).resolves.not.toHaveProperty("binding.nativeContextUsage");
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await expect(
fs.readFile(path.join(sessionsDir, "sessions.json"), "utf8").then(JSON.parse),
).resolves.toMatchObject({
"agent:main:session-1": { sessionId: "session-current", agentHarnessId: "codex" },
});
await fs.rm(`${sidecarPath}.migrated`);
await fs.writeFile(sidecarPath, JSON.stringify(legacyBinding), "utf8");
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
const resetTranscript = path.join(sessionsDir, "session-before-reset.jsonl");
const resetSidecar = `${resetTranscript}.codex-app-server.json`;
await fs.writeFile(resetTranscript, '{"type":"session","id":"session-before-reset"}\n', "utf8");
await fs.writeFile(
resetSidecar,
JSON.stringify({ schemaVersion: 1, threadId: "thread-before-reset" }),
"utf8",
);
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1 safe")],
warnings: [expect.stringContaining("session owner could not be resolved")],
});
await expect(fs.access(resetSidecar)).resolves.toBeUndefined();
await fs.rm(resetSidecar);
const conflictingTranscript = path.join(sessionsDir, "session-2.jsonl");
const conflictingSidecar = `${conflictingTranscript}.codex-app-server.json`;
await fs.writeFile(conflictingTranscript, '{"type":"session","id":"session-2"}\n', "utf8");
await fs.writeFile(
conflictingSidecar,
JSON.stringify({ schemaVersion: 1, threadId: "legacy-thread" }),
"utf8",
);
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:session-1": {
sessionId: "session-1",
sessionFile: "session-1.jsonl",
updatedAt: Date.now(),
},
"agent:main:session-2": {
sessionId: "session-2",
sessionFile: "session-2.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
const conflictingSessionKey = bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-2",
sessionKey: "agent:main:session-2",
});
await store.register(conflictingSessionKey, {
version: 1,
state: "active",
binding: {
threadId: "legacy-thread",
cwd: "/repo",
historyCoveredThrough: "2026-01-01T00:00:00.000Z",
},
});
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({
changes: [],
warnings: [
expect.stringContaining(`canonical plugin state changed at ${conflictingSessionKey}`),
],
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(conflictingTranscript),
}),
),
).resolves.toBeUndefined();
await expect(fs.access(conflictingSidecar)).resolves.toBeUndefined();
await fs.rm(conflictingSidecar);
const inverseTranscript = path.join(sessionsDir, "session-3.jsonl");
const inverseSidecar = `${inverseTranscript}.codex-app-server.json`;
const inverseConversationKey = bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(inverseTranscript),
});
await fs.writeFile(inverseTranscript, '{"type":"session","id":"session-3"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:session-3": {
sessionId: "session-3",
sessionFile: "session-3.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
inverseSidecar,
JSON.stringify({ schemaVersion: 1, threadId: "session-thread" }),
"utf8",
);
await store.register(inverseConversationKey, {
version: 1,
state: "active",
binding: { threadId: "conversation-thread", cwd: "/repo" },
});
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-3",
sessionKey: "agent:main:session-3",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-3",
binding: { threadId: "conversation-thread" },
});
await expect(store.lookup(inverseConversationKey)).resolves.toMatchObject({
state: "active",
binding: { threadId: "conversation-thread" },
});
await expect(fs.access(`${inverseSidecar}.migrated`)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("does not publish Codex session ownership before every binding row persists", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-order-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "agents", "main", "sessions");
const transcriptPath = path.join(sessionsDir, "session-order.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
const storePath = path.join(sessionsDir, "sessions.json");
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"session-order"}\n', "utf8");
await fs.writeFile(
storePath,
JSON.stringify({
"agent:main:order": {
sessionId: "session-order",
sessionFile: "session-order.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({ schemaVersion: 1, threadId: "thread-order" }),
"utf8",
);
const store = createPluginStateKeyedStoreForTests<StoredCodexAppServerBinding>("codex", {
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
env,
});
const registerIfAbsent = store.registerIfAbsent.bind(store);
let registerCalls = 0;
const failingStore: PluginStateKeyedStore<StoredCodexAppServerBinding> = {
...store,
async registerIfAbsent(key, value, opts) {
registerCalls++;
if (registerCalls === 2) {
throw new Error("injected session binding write failure");
}
return await registerIfAbsent(key, value, opts);
},
};
const failingContext: PluginDoctorStateMigrationContext = {
openPluginStateKeyedStore<T>() {
return failingStore as unknown as PluginStateKeyedStore<T>;
},
};
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(
migration.migrateLegacyState({
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: failingContext,
}),
).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1 safe")],
warnings: [expect.stringContaining("injected session binding write failure")],
});
await expect(fs.readFile(storePath, "utf8").then(JSON.parse)).resolves.toMatchObject({
"agent:main:order": { sessionId: "session-order" },
});
expect(
(JSON.parse(await fs.readFile(storePath, "utf8")) as Record<string, Record<string, unknown>>)[
"agent:main:order"
],
).not.toHaveProperty("agentHarnessId");
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-order",
sessionKey: "agent:main:order",
}),
),
).resolves.toBeUndefined();
await expect(fs.access(sidecarPath)).resolves.toBeUndefined();
await expect(
migration.migrateLegacyState({
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
await expect(fs.readFile(storePath, "utf8").then(JSON.parse)).resolves.toMatchObject({
"agent:main:order": {
sessionId: "session-order",
agentHarnessId: "codex",
},
});
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("retains a shipped binding when its session now belongs to another harness", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-owner-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "agents", "main", "sessions");
const transcriptPath = path.join(sessionsDir, "session-foreign.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"session-foreign"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:foreign": {
sessionId: "session-foreign",
sessionFile: "session-foreign.jsonl",
agentHarnessId: "openclaw",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({
schemaVersion: 1,
threadId: "thread-foreign",
sessionFile: transcriptPath,
}),
"utf8",
);
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(
migration.migrateLegacyState({
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toMatchObject({
changes: [],
warnings: [expect.stringContaining("owned by agent harness openclaw")],
});
await expect(fs.access(sidecarPath)).resolves.toBeUndefined();
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-foreign",
sessionKey: "agent:main:foreign",
}),
),
).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("imports sidecars from the pre-agent session directory before core moves it", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-legacy-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "sessions");
const transcriptPath = path.join(sessionsDir, "legacy-session.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"legacy-session"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:legacy": {
sessionId: "legacy-session",
sessionFile: "legacy-session.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({
schemaVersion: 1,
threadId: "legacy-thread",
sessionFile: transcriptPath,
}),
"utf8",
);
const params = {
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
};
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(migration.migrateLegacyState(params)).resolves.toMatchObject({ warnings: [] });
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "legacy-session",
sessionKey: "agent:main:legacy",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "legacy-session",
binding: { threadId: "legacy-thread" },
});
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await expect(
fs.readFile(path.join(sessionsDir, "sessions.json"), "utf8").then(JSON.parse),
).resolves.toMatchObject({
"agent:main:legacy": { sessionId: "legacy-session", agentHarnessId: "codex" },
});
});
it("uses the session index when a shipped sidecar transcript is missing", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "agents", "main", "sessions");
const transcriptPath = path.join(sessionsDir, "missing.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:missing": {
sessionId: "session-missing",
sessionFile: "missing.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({
schemaVersion: 1,
threadId: "thread-legacy-conversation",
sessionFile: transcriptPath,
}),
"utf8",
);
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(
migration.migrateLegacyState({
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
}),
),
).resolves.toMatchObject({
state: "active",
binding: { threadId: "thread-legacy-conversation" },
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-missing",
sessionKey: "agent:main:missing",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-missing",
binding: { threadId: "thread-legacy-conversation" },
});
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("imports a binding without crawling Codex rollout files", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-"));
const env = { ...process.env, OPENCLAW_STATE_DIR: stateDir };
const sessionsDir = path.join(stateDir, "agents", "main", "sessions");
const transcriptPath = path.join(sessionsDir, "session-fresh.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"session-fresh"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:main:fresh": {
sessionId: "session-fresh",
sessionFile: "session-fresh.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({ schemaVersion: 1, threadId: "thread-without-rollout" }),
"utf8",
);
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(
migration.migrateLegacyState({
config: {},
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toEqual({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
const targetKey = bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "main",
sessionId: "session-fresh",
sessionKey: "agent:main:fresh",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-fresh",
binding: { threadId: "thread-without-rollout" },
});
await expect(store.lookup(targetKey)).resolves.toMatchObject({
state: "active",
binding: { threadId: "thread-without-rollout" },
});
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("retains an ambiguous sidecar and converges after its owner resolves", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-"));
const env = { ...process.env, HOME: stateDir, OPENCLAW_STATE_DIR: stateDir };
const config = {
agents: { list: [{ id: "alpha" }, { id: "beta" }] },
session: { store: "~/shared/sessions.json" },
};
const sessionsDir = path.join(stateDir, "shared");
const transcriptPath = path.join(sessionsDir, "ambiguous.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(sessionsDir, { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"message"}\n', "utf8");
await fs.writeFile(
sidecarPath,
JSON.stringify({
schemaVersion: 1,
threadId: "thread-ambiguous",
sessionFile: transcriptPath,
}),
"utf8",
);
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await expect(
migration.migrateLegacyState({
config,
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1 safe")],
warnings: [expect.stringContaining("session owner could not be resolved")],
});
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
}),
),
).resolves.toMatchObject({ state: "active", binding: { threadId: "thread-ambiguous" } });
await expect(fs.access(sidecarPath)).resolves.toBeUndefined();
const conversationKey = bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
});
const imported = await store.lookup(conversationKey);
if (imported?.state !== "active") {
throw new Error("missing imported Codex conversation binding");
}
await store.register(conversationKey, {
...imported,
binding: { ...imported.binding, threadId: "thread-recovered" },
});
await expect(
migration.migrateLegacyState({
config,
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toEqual({
changes: [],
warnings: [expect.stringContaining("session owner could not be resolved")],
});
await expect(store.lookup(conversationKey)).resolves.toMatchObject({
state: "active",
binding: { threadId: "thread-recovered" },
});
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:alpha:ambiguous": {
sessionId: "session-ambiguous",
sessionFile: "ambiguous.jsonl",
totalTokens: 12_345,
totalTokensFresh: true,
contextTokens: 128_000,
updatedAt: Date.now(),
},
}),
"utf8",
);
await expect(
migration.migrateLegacyState({
config,
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
}),
).resolves.toMatchObject({
changes: [expect.stringContaining("Migrated 1")],
warnings: [],
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "alpha",
sessionId: "session-ambiguous",
sessionKey: "agent:alpha:ambiguous",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-ambiguous",
binding: { threadId: "thread-recovered" },
});
await expect(store.lookup(conversationKey)).resolves.toMatchObject({
state: "active",
binding: {
threadId: "thread-recovered",
},
});
await expect(store.lookup(conversationKey)).resolves.not.toHaveProperty(
"binding.nativeContextUsage",
);
await expect(fs.access(`${sidecarPath}.migrated`)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
});
it("uses canonical custom-store, agent, and nested transcript path resolution", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-doctor-"));
const customStoreRoot = await fs.mkdtemp(
path.join(os.tmpdir(), "openclaw-codex-custom-store-"),
);
const env = { ...process.env, HOME: stateDir, OPENCLAW_STATE_DIR: stateDir };
const config = {
agents: { list: [{ id: "alpha" }] },
session: { store: path.join(customStoreRoot, "{agentId}", "sessions.json") },
};
const sessionsDir = path.join(customStoreRoot, "alpha");
const transcriptPath = path.join(sessionsDir, "nested", "session-custom.jsonl");
const sidecarPath = `${transcriptPath}.codex-app-server.json`;
await fs.mkdir(path.dirname(transcriptPath), { recursive: true });
await fs.writeFile(transcriptPath, '{"type":"session","id":"session-custom"}\n', "utf8");
await fs.writeFile(
path.join(sessionsDir, "sessions.json"),
JSON.stringify({
"agent:alpha:custom": {
sessionId: "session-custom",
sessionFile: "nested/session-custom.jsonl",
updatedAt: Date.now(),
},
}),
"utf8",
);
await fs.writeFile(
sidecarPath,
JSON.stringify({ schemaVersion: 1, threadId: "thread-custom" }),
"utf8",
);
const unrelatedSidecar = path.join(
customStoreRoot,
"unrelated",
`not-a-session.jsonl.codex-app-server.json`,
);
await fs.mkdir(path.dirname(unrelatedSidecar), { recursive: true });
await fs.writeFile(
unrelatedSidecar,
JSON.stringify({ schemaVersion: 1, threadId: "unrelated-thread" }),
"utf8",
);
const migration = stateMigrations[0];
if (!migration) {
throw new Error("missing Codex binding migration");
}
await migration.migrateLegacyState({
config,
env,
stateDir,
oauthDir: path.join(stateDir, "oauth"),
context: createDoctorContext(env),
});
const store = createDoctorContext(env).openPluginStateKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
});
await expect(
store.lookup(
bindingStoreKey({
kind: "session",
agentId: "alpha",
sessionId: "session-custom",
sessionKey: "agent:alpha:custom",
}),
),
).resolves.toMatchObject({
state: "active",
sessionId: "session-custom",
binding: { threadId: "thread-custom" },
});
await expect(
store.lookup(
bindingStoreKey({
kind: "conversation",
bindingId: legacyCodexConversationBindingId(transcriptPath),
}),
),
).resolves.toMatchObject({
state: "active",
binding: { threadId: "thread-custom" },
});
await expect(fs.access(unrelatedSidecar)).resolves.toBeUndefined();
await fs.rm(stateDir, { recursive: true, force: true });
await fs.rm(customStoreRoot, { recursive: true, force: true });
});
});

View File

@@ -1,4 +1,7 @@
/** Doctor contract hooks for Codex config, state migration, and route ownership. */
/**
* Doctor contract hooks for Codex plugin config migrations and session-route
* ownership warnings.
*/
import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
import type { DoctorSessionRouteStateOwner } from "openclaw/plugin-sdk/runtime-doctor";
@@ -28,7 +31,9 @@ export const legacyConfigRules: LegacyConfigRule[] = [
},
];
/** Removes retired Codex plugin config keys while preserving unrelated config. */
/**
* Removes retired Codex plugin config keys while preserving unrelated config.
*/
export function normalizeCompatibilityConfig({ cfg }: { cfg: OpenClawConfig }): {
config: OpenClawConfig;
changes: string[];
@@ -42,9 +47,10 @@ export function normalizeCompatibilityConfig({ cfg }: { cfg: OpenClawConfig }):
const nextConfig = structuredClone(cfg) as OpenClawConfig & {
plugins?: Record<string, unknown>;
};
const nextPluginConfig = asRecord(
asRecord(asRecord(asRecord(nextConfig.plugins)?.entries)?.codex)?.config,
);
const nextPlugins = asRecord(nextConfig.plugins);
const nextEntries = asRecord(nextPlugins?.entries);
const nextEntry = asRecord(nextEntries?.codex);
const nextPluginConfig = asRecord(nextEntry?.config);
if (!nextPluginConfig) {
return { config: cfg, changes: [] };
}
@@ -69,5 +75,3 @@ export const sessionRouteStateOwners: DoctorSessionRouteStateOwner[] = [
authProfilePrefixes: ["codex:", "codex-cli:", "openai-codex:"],
},
];
export { stateMigrations } from "./src/migration/session-binding-sidecars.js";

View File

@@ -1,18 +1,9 @@
// Codex tests cover harness plugin behavior.
import fs from "node:fs/promises";
import os from "node:os";
import path from "node:path";
import { describe, expect, it } from "vitest";
import { createCodexAppServerAgentHarness } from "./harness.js";
import {
createCodexTestBindingStore,
testCodexAppServerBindingStore,
} from "./src/app-server/session-binding.test-helpers.js";
describe("Codex agent harness supports()", () => {
const harness = createCodexAppServerAgentHarness({
bindingStore: testCodexAppServerBindingStore,
});
const harness = createCodexAppServerAgentHarness();
it("supports the canonical codex virtual provider", () => {
expect(harness.supports({ provider: "codex", requestedRuntime: "codex" })).toEqual({
@@ -49,149 +40,8 @@ describe("Codex agent harness supports()", () => {
});
it("honors explicit provider id overrides", () => {
const narrowHarness = createCodexAppServerAgentHarness({
bindingStore: testCodexAppServerBindingStore,
providerIds: ["codex"],
});
const narrowHarness = createCodexAppServerAgentHarness({ providerIds: ["codex"] });
const result = narrowHarness.supports({ provider: "openai", requestedRuntime: "codex" });
expect(result.supported).toBe(false);
});
});
describe("Codex agent harness reset", () => {
it("uses the host agent for global session keys", async () => {
const bindingStore = createCodexTestBindingStore();
const harness = createCodexAppServerAgentHarness({ bindingStore });
const identity = {
kind: "session" as const,
agentId: "work",
sessionId: "session-1",
sessionKey: "global",
};
await bindingStore.mutate(identity, {
kind: "set",
binding: { threadId: "thread-work", cwd: "/repo" },
});
await harness.reset?.({
agentId: "work",
sessionId: "session-1",
sessionKey: "global",
reason: "reset",
});
await expect(bindingStore.read(identity)).resolves.toBeUndefined();
await expect(
bindingStore.mutate(identity, {
kind: "set",
binding: { threadId: "thread-stale", cwd: "/stale" },
}),
).resolves.toBe(false);
const nextIdentity = { ...identity, sessionId: "session-2" };
await expect(
bindingStore.mutate(nextIdentity, {
kind: "set",
binding: { threadId: "thread-next", cwd: "/next" },
}),
).resolves.toBe(false);
await expect(
bindingStore.mutate(nextIdentity, {
kind: "reclaim-generation",
expectedPreviousSessionId: identity.sessionId,
}),
).resolves.toBe(true);
await expect(
bindingStore.mutate(nextIdentity, {
kind: "set",
binding: { threadId: "thread-next", cwd: "/next" },
}),
).resolves.toBe(true);
await expect(bindingStore.read(nextIdentity)).resolves.toMatchObject({
threadId: "thread-next",
});
});
it("accepts an absent binding but rejects a mismatched reset generation", async () => {
const bindingStore = createCodexTestBindingStore();
const harness = createCodexAppServerAgentHarness({ bindingStore });
const current = {
kind: "session" as const,
agentId: "main",
sessionId: "session-1",
sessionKey: "agent:main:main",
};
await expect(
harness.reset?.({
agentId: "main",
sessionId: "missing-session",
sessionKey: "agent:main:missing",
reason: "reset",
}),
).resolves.toBeUndefined();
await bindingStore.mutate(current, {
kind: "set",
binding: { threadId: "thread-1", cwd: "/repo" },
});
await expect(
harness.reset?.({
agentId: "main",
sessionId: "session-2",
sessionKey: current.sessionKey,
reason: "reset",
}),
).rejects.toThrow("binding generation changed");
await expect(bindingStore.read(current)).resolves.toMatchObject({ threadId: "thread-1" });
});
it("reclaims a stale generation left while the Codex plugin was unavailable", async () => {
const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-reset-"));
const storePath = path.join(stateDir, "sessions.json");
const sessionKey = "agent:main:main";
await fs.writeFile(
storePath,
JSON.stringify({
[sessionKey]: {
sessionId: "session-2",
updatedAt: Date.now(),
},
}),
"utf8",
);
const bindingStore = createCodexTestBindingStore();
const harness = createCodexAppServerAgentHarness({
bindingStore,
resolveConfig: () => ({ session: { store: storePath } }),
});
const stale = {
kind: "session" as const,
agentId: "main",
sessionId: "session-1",
sessionKey,
};
await bindingStore.mutate(stale, {
kind: "set",
binding: { threadId: "thread-stale", cwd: "/repo" },
});
await expect(
harness.reset?.({
agentId: "main",
sessionId: "session-2",
sessionKey,
reason: "reset",
}),
).resolves.toBeUndefined();
const current = { ...stale, sessionId: "session-2" };
await expect(bindingStore.read(current)).resolves.toBeUndefined();
await expect(
bindingStore.mutate(current, {
kind: "set",
binding: { threadId: "thread-delayed", cwd: "/repo" },
}),
).resolves.toBe(false);
await fs.rm(stateDir, { recursive: true, force: true });
});
});

View File

@@ -7,13 +7,11 @@ import type {
AgentHarnessCompactResult,
ContextEngineHostCapability,
} from "openclaw/plugin-sdk/agent-harness-runtime";
import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
import type {
CodexAppServerListModelsOptions,
CodexAppServerModel,
CodexAppServerModelListResult,
} from "./src/app-server/models.js";
import type { CodexAppServerBindingStore } from "./src/app-server/session-binding.js";
const DEFAULT_CODEX_HARNESS_PROVIDER_IDS = new Set(["codex", "openai"]);
const CODEX_APP_SERVER_CONTEXT_ENGINE_HOST_CAPABILITIES = [
@@ -39,14 +37,12 @@ type CodexAppServerAgentHarness = AgentHarness & {
* Creates the Codex app-server harness used for attempts, side questions,
* compaction, reset, and disposal.
*/
export function createCodexAppServerAgentHarness(options: {
export function createCodexAppServerAgentHarness(options?: {
id?: string;
label?: string;
providerIds?: Iterable<string>;
pluginConfig?: unknown;
resolvePluginConfig?: () => unknown;
resolveConfig?: () => OpenClawConfig | undefined;
bindingStore: CodexAppServerBindingStore;
}): AgentHarness {
const providerIds = new Set(
[...(options?.providerIds ?? DEFAULT_CODEX_HARNESS_PROVIDER_IDS)].map((id) =>
@@ -75,7 +71,6 @@ export function createCodexAppServerAgentHarness(options: {
// cold provider catalog reads do not pull in the whole Codex runtime.
const { runCodexAppServerAttempt } = await import("./src/app-server/run-attempt.js");
return runCodexAppServerAttempt(params, {
bindingStore: options.bindingStore,
pluginConfig: options?.resolvePluginConfig?.() ?? options?.pluginConfig,
nativeHookRelay: { enabled: true },
});
@@ -83,7 +78,6 @@ export function createCodexAppServerAgentHarness(options: {
runSideQuestion: async (params) => {
const { runCodexAppServerSideQuestion } = await import("./src/app-server/side-question.js");
return runCodexAppServerSideQuestion(params, {
bindingStore: options.bindingStore,
pluginConfig: options?.resolvePluginConfig?.() ?? options?.pluginConfig,
nativeHookRelay: { enabled: true },
});
@@ -91,43 +85,20 @@ export function createCodexAppServerAgentHarness(options: {
compact: async (params) => {
const { maybeCompactCodexAppServerSession } = await import("./src/app-server/compact.js");
return maybeCompactCodexAppServerSession(params, {
bindingStore: options.bindingStore,
pluginConfig: options?.resolvePluginConfig?.() ?? options?.pluginConfig,
});
},
compactAfterContextEngine: async (params) => {
const { maybeCompactCodexAppServerSession } = await import("./src/app-server/compact.js");
return maybeCompactCodexAppServerSession(params, {
bindingStore: options.bindingStore,
pluginConfig: options?.resolvePluginConfig?.() ?? options?.pluginConfig,
allowNonManualNativeRequest: true,
});
},
reset: async (params) => {
if (params.sessionId) {
const { reclaimCurrentCodexSessionGeneration, sessionBindingIdentity } =
await import("./src/app-server/session-binding.js");
const identity = sessionBindingIdentity({
agentId: params.agentId,
sessionId: params.sessionId,
sessionKey: params.sessionKey,
});
let retired = await options.bindingStore.retireSessionGeneration(identity);
if (retired === "conflict") {
const reclaimed = await reclaimCurrentCodexSessionGeneration({
bindingStore: options.bindingStore,
identity,
config: options.resolveConfig?.(),
});
if (reclaimed) {
retired = await options.bindingStore.retireSessionGeneration(identity);
}
}
if (retired === "conflict") {
throw new Error(
`Codex binding generation changed before session ${params.sessionId} could reset`,
);
}
if (params.sessionFile) {
const { clearCodexAppServerBinding } = await import("./src/app-server/session-binding.js");
await clearCodexAppServerBinding(params.sessionFile);
}
},
dispose: async () => {

View File

@@ -4,30 +4,10 @@ import { createTestPluginApi } from "openclaw/plugin-sdk/plugin-test-api";
import { describe, expect, it, vi } from "vitest";
import { createCodexAppServerAgentHarness } from "./harness.js";
import plugin from "./index.js";
import {
createCodexAppServerBindingStore,
sessionBindingIdentity,
} from "./src/app-server/session-binding.js";
import {
createCodexTestBindingStateStore,
testCodexAppServerBindingStore,
} from "./src/app-server/session-binding.test-helpers.js";
const runCodexAppServerAttemptMock = vi.hoisted(() => vi.fn());
const runCodexAppServerSideQuestionMock = vi.hoisted(() => vi.fn());
function createCodexTestRuntime(
current?: () => unknown,
stateStore = createCodexTestBindingStateStore(),
) {
return {
...(current ? { config: { current } } : {}),
state: {
openSyncKeyedStore: () => stateStore,
},
} as never;
}
vi.mock("./src/app-server/run-attempt.js", () => ({
runCodexAppServerAttempt: runCodexAppServerAttemptMock,
}));
@@ -59,6 +39,7 @@ describe("codex plugin", () => {
const registerMigrationProvider = vi.fn();
const registerProvider = vi.fn();
const on = vi.fn();
const onConversationBindingResolved = vi.fn();
plugin.register(
createTestPluginApi({
@@ -67,13 +48,14 @@ describe("codex plugin", () => {
source: "test",
config: {},
pluginConfig: {},
runtime: createCodexTestRuntime(),
runtime: {} as never,
registerAgentHarness,
registerCommand,
registerMediaUnderstandingProvider,
registerMigrationProvider,
registerProvider,
on,
onConversationBindingResolved,
}),
);
@@ -83,6 +65,9 @@ describe("codex plugin", () => {
| Record<string, unknown>
| undefined;
const inboundClaimRegistration = mockCall(on) as [unknown, unknown] | undefined;
const bindingResolvedRegistration = mockCall(onConversationBindingResolved) as
| [unknown]
| undefined;
expect(providerRegistration.id).toBe("codex");
expect(providerRegistration.label).toBe("Codex");
@@ -109,12 +94,33 @@ describe("codex plugin", () => {
expect(migrationRegistration?.label).toBe("Codex");
expect(inboundClaimRegistration?.[0]).toBe("inbound_claim");
expect(typeof inboundClaimRegistration?.[1]).toBe("function");
expect(typeof bindingResolvedRegistration?.[0]).toBe("function");
});
it("registers with capture APIs that do not expose conversation binding hooks yet", () => {
const registerProvider = vi.fn();
const api = createTestPluginApi({
id: "codex",
name: "Codex",
source: "test",
config: {},
pluginConfig: {},
runtime: {} as never,
registerAgentHarness: vi.fn(),
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
registerProvider,
on: vi.fn(),
});
delete (api as { onConversationBindingResolved?: unknown }).onConversationBindingResolved;
plugin.register(api);
expect(registerProvider).toHaveBeenCalledTimes(1);
expect((mockCallArg(registerProvider) as { id?: string } | undefined)?.id).toBe("codex");
});
it("claims the Codex routing providers by default", () => {
const harness = createCodexAppServerAgentHarness({
bindingStore: testCodexAppServerBindingStore,
});
const harness = createCodexAppServerAgentHarness();
expect(harness.deliveryDefaults?.sourceVisibleReplies).toBe("message_tool");
expect(
@@ -135,196 +141,8 @@ describe("codex plugin", () => {
expect(unsupported.supported).toBe(false);
});
it("clears only ended session binding rows in the owning agent scope", async () => {
const stateStore = createCodexTestBindingStateStore();
const bindingStore = createCodexAppServerBindingStore(stateStore);
const on = vi.fn();
plugin.register(
createTestPluginApi({
id: "codex",
name: "Codex",
source: "test",
config: {},
pluginConfig: {},
runtime: createCodexTestRuntime(undefined, stateStore),
registerAgentHarness: vi.fn(),
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
registerMigrationProvider: vi.fn(),
registerProvider: vi.fn(),
on,
}),
);
const sessionEnd = on.mock.calls.find(([name]) => name === "session_end")?.[1] as
| ((
event: { sessionId: string; sessionKey?: string; reason?: string },
ctx: { agentId?: string; sessionId: string; sessionKey?: string },
) => Promise<void>)
| undefined;
if (!sessionEnd) {
throw new Error("missing Codex session_end hook");
}
const identity = sessionBindingIdentity({
agentId: "worker",
sessionId: "session-1",
sessionKey: "agent:worker:session-1",
});
const setBinding = () =>
bindingStore.mutate(identity, {
kind: "set",
binding: { threadId: "thread-1", cwd: "/repo" },
});
for (const reason of ["shutdown", "restart", "compaction", "unknown"] as const) {
await setBinding();
await sessionEnd(
{ sessionId: "session-1", sessionKey: "agent:worker:session-1", reason },
{ agentId: "worker", sessionId: "session-1" },
);
await expect(bindingStore.read(identity)).resolves.toMatchObject({
threadId: "thread-1",
});
}
for (const reason of ["new", "reset", "idle", "daily", "deleted"] as const) {
await setBinding();
await sessionEnd(
{ sessionId: "session-1", sessionKey: "agent:worker:session-1", reason },
{ agentId: "worker", sessionId: "session-1" },
);
await expect(bindingStore.read(identity)).resolves.toBeUndefined();
}
});
it("adopts compaction successors before delayed lifecycle cleanup", async () => {
const stateStore = createCodexTestBindingStateStore();
const bindingStore = createCodexAppServerBindingStore(stateStore);
const on = vi.fn();
plugin.register(
createTestPluginApi({
id: "codex",
name: "Codex",
source: "test",
config: {},
pluginConfig: {},
runtime: createCodexTestRuntime(undefined, stateStore),
registerAgentHarness: vi.fn(),
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
registerMigrationProvider: vi.fn(),
registerProvider: vi.fn(),
on,
}),
);
const afterCompaction = on.mock.calls.find(([name]) => name === "after_compaction")?.[1] as
| ((
event: {
messageCount: number;
compactedCount: number;
previousSessionId?: string;
},
ctx: { agentId?: string; sessionId?: string; sessionKey?: string },
) => Promise<void>)
| undefined;
const sessionEnd = on.mock.calls.find(([name]) => name === "session_end")?.[1] as
| ((
event: { sessionId: string; sessionKey?: string; reason?: string },
ctx: { agentId?: string; sessionId: string; sessionKey?: string },
) => Promise<void>)
| undefined;
if (!afterCompaction || !sessionEnd) {
throw new Error("missing Codex compaction lifecycle hooks");
}
const sessionKey = "agent:worker:telegram:chat-1";
const previous = sessionBindingIdentity({
agentId: "worker",
sessionId: "session-1",
sessionKey,
});
const successor = sessionBindingIdentity({
agentId: "worker",
sessionId: "session-2",
sessionKey,
});
const newest = sessionBindingIdentity({
agentId: "worker",
sessionId: "session-3",
sessionKey,
});
await bindingStore.mutate(previous, {
kind: "set",
binding: { threadId: "thread-1", cwd: "/repo" },
});
await afterCompaction(
{ messageCount: 1, compactedCount: 1, previousSessionId: "session-1" },
{ agentId: "worker", sessionId: "session-2", sessionKey },
);
await expect(bindingStore.read(previous)).resolves.toBeUndefined();
await expect(bindingStore.read(successor)).resolves.toMatchObject({ threadId: "thread-1" });
await afterCompaction(
{ messageCount: 1, compactedCount: 1, previousSessionId: "session-2" },
{ agentId: "worker", sessionId: "session-3", sessionKey },
);
await afterCompaction(
{ messageCount: 1, compactedCount: 1, previousSessionId: "session-1" },
{ agentId: "worker", sessionId: "session-2", sessionKey },
);
await expect(bindingStore.read(successor)).resolves.toBeUndefined();
await expect(bindingStore.read(newest)).resolves.toMatchObject({ threadId: "thread-1" });
await sessionEnd(
{ sessionId: "session-1", sessionKey, reason: "reset" },
{ agentId: "worker", sessionId: "session-1", sessionKey },
);
await sessionEnd(
{ sessionId: "session-2", sessionKey, reason: "compaction" },
{ agentId: "worker", sessionId: "session-2", sessionKey },
);
await expect(bindingStore.read(newest)).resolves.toMatchObject({ threadId: "thread-1" });
expect(stateStore.entries()).toHaveLength(1);
});
it("ignores compaction for a session without a Codex binding", async () => {
const warn = vi.fn();
const on = vi.fn();
plugin.register(
createTestPluginApi({
id: "codex",
name: "Codex",
source: "test",
config: {},
pluginConfig: {},
logger: { debug: vi.fn(), info: vi.fn(), warn, error: vi.fn() },
runtime: createCodexTestRuntime(),
registerAgentHarness: vi.fn(),
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
registerMigrationProvider: vi.fn(),
registerProvider: vi.fn(),
on,
}),
);
const afterCompaction = on.mock.calls.find(([name]) => name === "after_compaction")?.[1] as
| ((event: object, ctx: { sessionId?: string; sessionKey?: string }) => Promise<void>)
| undefined;
if (!afterCompaction) {
throw new Error("missing Codex after_compaction hook");
}
await afterCompaction(
{ previousSessionId: "session-1" },
{ sessionId: "session-2", sessionKey: "agent:main:main" },
);
expect(warn).not.toHaveBeenCalled();
});
it("enables the native hook relay for public Codex app-server attempts", async () => {
const harness = createCodexAppServerAgentHarness({
bindingStore: testCodexAppServerBindingStore,
pluginConfig: { appServer: {} },
});
const harness = createCodexAppServerAgentHarness({ pluginConfig: { appServer: {} } });
const result = { success: true };
runCodexAppServerAttemptMock.mockResolvedValueOnce(result);
@@ -333,7 +151,6 @@ describe("codex plugin", () => {
expect(runCodexAppServerAttemptMock).toHaveBeenCalledWith(
{ prompt: "hello" },
{
bindingStore: testCodexAppServerBindingStore,
pluginConfig: { appServer: {} },
nativeHookRelay: { enabled: true },
},
@@ -368,7 +185,11 @@ describe("codex plugin", () => {
source: "test",
config: {},
pluginConfig: { codexPlugins: { enabled: false } },
runtime: createCodexTestRuntime(() => liveConfig),
runtime: {
config: {
current: () => liveConfig,
},
} as never,
registerAgentHarness,
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
@@ -388,49 +209,14 @@ describe("codex plugin", () => {
expect(runCodexAppServerAttemptMock).toHaveBeenCalledWith(
{ prompt: "calendar" },
{
bindingStore: expect.any(Object),
pluginConfig: liveConfig.plugins.entries.codex.config,
nativeHookRelay: { enabled: true },
},
);
});
it("does not resurrect startup Codex config after the live entry is removed", async () => {
const registerAgentHarness = vi.fn();
plugin.register(
createTestPluginApi({
id: "codex",
name: "Codex",
source: "test",
config: {},
pluginConfig: { appServer: { mode: "yolo" } },
runtime: createCodexTestRuntime(() => ({ plugins: { entries: {} } })),
registerAgentHarness,
registerCommand: vi.fn(),
registerMediaUnderstandingProvider: vi.fn(),
registerMigrationProvider: vi.fn(),
registerProvider: vi.fn(),
on: vi.fn(),
}),
);
const harness = mockCallArg(registerAgentHarness) as ReturnType<
typeof createCodexAppServerAgentHarness
>;
runCodexAppServerAttemptMock.mockResolvedValueOnce({ success: true });
await harness.runAttempt({ prompt: "default policy" } as never);
expect(runCodexAppServerAttemptMock).toHaveBeenCalledWith(
{ prompt: "default policy" },
expect.objectContaining({ pluginConfig: undefined }),
);
});
it("enables the native hook relay for public Codex side questions", async () => {
const harness = createCodexAppServerAgentHarness({
bindingStore: testCodexAppServerBindingStore,
pluginConfig: { appServer: {} },
});
const harness = createCodexAppServerAgentHarness({ pluginConfig: { appServer: {} } });
const runSideQuestion = harness["runSideQuestion"];
const result = { text: "ok" };
runCodexAppServerSideQuestionMock.mockResolvedValueOnce(result);
@@ -443,7 +229,6 @@ describe("codex plugin", () => {
expect(runCodexAppServerSideQuestionMock).toHaveBeenCalledWith(
{ question: "btw" },
{
bindingStore: testCodexAppServerBindingStore,
pluginConfig: { appServer: {} },
nativeHookRelay: { enabled: true },
},

View File

@@ -4,71 +4,47 @@
*/
import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
import { mutateConfigFile } from "openclaw/plugin-sdk/config-mutation";
import {
resolveLivePluginConfigObject,
resolvePluginConfigObject,
} from "openclaw/plugin-sdk/plugin-config-runtime";
import { resolveLivePluginConfigObject } from "openclaw/plugin-sdk/plugin-config-runtime";
import { definePluginEntry } from "openclaw/plugin-sdk/plugin-entry";
import { createCodexAppServerAgentHarness } from "./harness.js";
import { buildCodexMediaUnderstandingProvider } from "./media-understanding-provider.js";
import { buildCodexProvider } from "./provider.js";
import {
CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
CODEX_APP_SERVER_BINDING_NAMESPACE,
createLazyCodexAppServerBindingStore,
type StoredCodexAppServerBinding,
} from "./src/app-server/session-binding-store.js";
import type { CodexPluginsConfigBlock } from "./src/command-plugins-management.js";
import { createCodexCommand } from "./src/commands.js";
import {
handleCodexConversationBindingResolved,
handleCodexConversationInboundClaim,
} from "./src/conversation-binding.js";
import { buildCodexMigrationProvider } from "./src/migration/provider.js";
import {
createCodexCliSessionNodeHostCommands,
createCodexCliSessionNodeInvokePolicies,
} from "./src/node-cli-session-registration.js";
const ENDED_SESSION_REASONS: ReadonlySet<string> = new Set([
"new",
"reset",
"idle",
"daily",
"deleted",
]);
listCodexCliSessionsOnNode,
resumeCodexCliSessionOnNode,
resolveCodexCliSessionForBindingOnNode,
} from "./src/node-cli-sessions.js";
export default definePluginEntry({
id: "codex",
name: "Codex",
description: "Codex app-server harness and Codex-managed GPT model catalog.",
register(api) {
const runtimeConfigLoader = api.runtime.config?.current
? () => api.runtime.config?.current() as OpenClawConfig
: undefined;
const resolveCurrentConfig = () => runtimeConfigLoader?.();
const loadNodeCliSessions = () => import("./src/node-cli-sessions.js");
const resolveCurrentConfig = () =>
api.runtime.config?.current ? (api.runtime.config.current() as OpenClawConfig) : undefined;
const resolveCurrentPluginConfig = () =>
// Codex plugin config can change at runtime; resolve from live config for
// harness attempts and binding claims instead of keeping startup values.
resolveLivePluginConfigObject(
runtimeConfigLoader,
resolveCurrentConfig,
"codex",
api.pluginConfig as Record<string, unknown>,
);
const bindingStore = createLazyCodexAppServerBindingStore(
api.runtime.state.openSyncKeyedStore<StoredCodexAppServerBinding>({
namespace: CODEX_APP_SERVER_BINDING_NAMESPACE,
maxEntries: CODEX_APP_SERVER_BINDING_MAX_ENTRIES,
overflowPolicy: "reject-new",
}),
);
) ?? api.pluginConfig;
api.registerAgentHarness(
createCodexAppServerAgentHarness({
bindingStore,
resolveConfig: resolveCurrentConfig,
resolvePluginConfig: resolveCurrentPluginConfig,
}),
createCodexAppServerAgentHarness({ resolvePluginConfig: resolveCurrentPluginConfig }),
);
api.registerProvider(buildCodexProvider({ pluginConfig: api.pluginConfig }));
api.registerMediaUnderstandingProvider(
buildCodexMediaUnderstandingProvider({ resolvePluginConfig: resolveCurrentPluginConfig }),
buildCodexMediaUnderstandingProvider({ pluginConfig: api.pluginConfig }),
);
api.registerMigrationProvider(buildCodexMigrationProvider({ runtime: api.runtime }));
for (const command of createCodexCliSessionNodeHostCommands()) {
@@ -79,43 +55,43 @@ export default definePluginEntry({
}
api.registerCommand(
createCodexCommand({
resolvePluginConfig: resolveCurrentPluginConfig,
pluginConfig: api.pluginConfig,
deps: {
bindingStore,
listCodexCliSessionsOnNode: async (params) =>
await (
await loadNodeCliSessions()
).listCodexCliSessionsOnNode({
runtime: api.runtime,
...params,
}),
resolveCodexCliSessionForBindingOnNode: async (params) =>
await (
await loadNodeCliSessions()
).resolveCodexCliSessionForBindingOnNode({
runtime: api.runtime,
...params,
}),
listCodexCliSessionsOnNode: (params) =>
listCodexCliSessionsOnNode({ runtime: api.runtime, ...params }),
resolveCodexCliSessionForBindingOnNode: (params) =>
resolveCodexCliSessionForBindingOnNode({ runtime: api.runtime, ...params }),
codexPluginsManagementIo: {
readConfig: () => {
const current = (api.runtime.config?.current?.() ?? {}) as OpenClawConfig;
const codexPlugins = resolvePluginConfigObject(current, "codex")?.codexPlugins;
if (
!codexPlugins ||
typeof codexPlugins !== "object" ||
Array.isArray(codexPlugins)
) {
const plugins = (current as Record<string, unknown>).plugins;
if (!plugins || typeof plugins !== "object") {
return Promise.resolve({});
}
const block = codexPlugins as Record<string, unknown>;
const declared = block.plugins;
const entries = (plugins as Record<string, unknown>).entries;
if (!entries || typeof entries !== "object") {
return Promise.resolve({});
}
const codexEntry = (entries as Record<string, unknown>).codex;
if (!codexEntry || typeof codexEntry !== "object") {
return Promise.resolve({});
}
const config = (codexEntry as Record<string, unknown>).config;
if (!config || typeof config !== "object") {
return Promise.resolve({});
}
const codexPlugins = (config as Record<string, unknown>).codexPlugins;
if (!codexPlugins || typeof codexPlugins !== "object") {
return Promise.resolve({});
}
const declared = (codexPlugins as Record<string, unknown>).plugins;
if (!declared || typeof declared !== "object") {
return Promise.resolve({
enabled: block.enabled === true,
enabled: (codexPlugins as Record<string, unknown>).enabled === true,
});
}
return Promise.resolve({
enabled: block.enabled === true,
enabled: (codexPlugins as Record<string, unknown>).enabled === true,
plugins: declared as Record<string, never>,
});
},
@@ -125,12 +101,17 @@ export default definePluginEntry({
// Create the nested plugin config path on demand so codex
// plugin commands can enable/update Codex-managed plugins.
const root = draft as Record<string, unknown>;
const pluginsBlock = (root.plugins ??= {}) as Record<string, unknown>;
const entries = (pluginsBlock.entries ??= {}) as Record<string, unknown>;
const codexEntry = (entries.codex ??= {}) as Record<string, unknown>;
const config = (codexEntry.config ??= {}) as Record<string, unknown>;
const codexPlugins = (config.codexPlugins ??= {}) as Record<string, unknown>;
codexPlugins.plugins ??= {};
root.plugins = (root.plugins ?? {}) as Record<string, unknown>;
const pluginsBlock = root.plugins as Record<string, unknown>;
pluginsBlock.entries = (pluginsBlock.entries ?? {}) as Record<string, unknown>;
const entries = pluginsBlock.entries as Record<string, unknown>;
entries.codex = (entries.codex ?? {}) as Record<string, unknown>;
const codexEntry = entries.codex as Record<string, unknown>;
codexEntry.config = (codexEntry.config ?? {}) as Record<string, unknown>;
const config = codexEntry.config as Record<string, unknown>;
config.codexPlugins = (config.codexPlugins ?? {}) as Record<string, unknown>;
const codexPlugins = config.codexPlugins as Record<string, unknown>;
codexPlugins.plugins = (codexPlugins.plugins ?? {}) as Record<string, unknown>;
update(codexPlugins as CodexPluginsConfigBlock);
},
});
@@ -139,58 +120,14 @@ export default definePluginEntry({
},
}),
);
api.on("inbound_claim", async (event, ctx) => {
const { handleCodexConversationInboundClaim } = await import("./src/conversation-binding.js");
return await handleCodexConversationInboundClaim(event, ctx, {
bindingStore,
api.on("inbound_claim", (event, ctx) =>
handleCodexConversationInboundClaim(event, ctx, {
pluginConfig: resolveCurrentPluginConfig(),
config: resolveCurrentConfig(),
resumeCodexCliSessionOnNode: async (params) =>
await (
await loadNodeCliSessions()
).resumeCodexCliSessionOnNode({
runtime: api.runtime,
...params,
}),
});
});
api.on("after_compaction", async (event, ctx) => {
const previousSessionId = event.previousSessionId?.trim();
const sessionId = ctx.sessionId?.trim();
if (!previousSessionId || !sessionId || previousSessionId === sessionId) {
return;
}
const config = resolveCurrentConfig();
const sessionKey = ctx.sessionKey?.trim();
const { sessionBindingIdentity } = await import("./src/app-server/session-binding.js");
const identity = sessionBindingIdentity({
sessionId,
...(sessionKey ? { sessionKey } : {}),
...(ctx.agentId ? { agentId: ctx.agentId } : {}),
...(config ? { config } : {}),
});
const adopted = await bindingStore.adoptSessionGeneration(identity, previousSessionId);
if (adopted === "conflict") {
api.logger.warn?.(
`codex: could not adopt compacted session generation ${sessionId} (${adopted}); secondary native compaction will skip`,
);
}
});
api.on("session_end", async (event, ctx) => {
if (!event.reason || !ENDED_SESSION_REASONS.has(event.reason)) {
return;
}
const sessionKey = event.sessionKey ?? ctx.sessionKey;
const config = resolveCurrentConfig();
const { sessionBindingIdentity } = await import("./src/app-server/session-binding.js");
await bindingStore.retireSessionGeneration(
sessionBindingIdentity({
sessionId: event.sessionId,
...(sessionKey ? { sessionKey } : {}),
...(ctx.agentId ? { agentId: ctx.agentId } : {}),
...(config ? { config } : {}),
}),
);
});
resumeCodexCliSessionOnNode: (params) =>
resumeCodexCliSessionOnNode({ runtime: api.runtime, ...params }),
}),
);
api.onConversationBindingResolved?.(handleCodexConversationBindingResolved);
},
});

View File

@@ -2,25 +2,8 @@
import { MAX_TIMER_TIMEOUT_MS } from "openclaw/plugin-sdk/number-runtime";
import { afterEach, describe, expect, it, vi } from "vitest";
import { buildCodexMediaUnderstandingProvider } from "./media-understanding-provider.js";
import { CodexAppServerRpcError, type CodexAppServerClient } from "./src/app-server/client.js";
import type { CodexAppServerClient } from "./src/app-server/client.js";
import type { CodexServerNotification, JsonValue } from "./src/app-server/protocol.js";
import { adaptCodexTestClientFactory } from "./src/app-server/test-support.js";
const EXPECTED_MEDIA_THREAD_CONFIG = {
project_doc_max_bytes: 0,
web_search: "disabled",
"tools.experimental_request_user_input.enabled": false,
"features.hooks": false,
"features.multi_agent": false,
"features.apps": false,
"features.plugins": false,
"features.image_generation": false,
"features.skill_mcp_dependency_install": false,
"features.memories": false,
"features.goals": false,
"features.code_mode": false,
"features.code_mode_only": false,
};
const sharedClientMocks = vi.hoisted(() => ({
createIsolatedCodexAppServerClient: vi.fn(),
@@ -102,15 +85,13 @@ function createFakeClient(options?: {
inputModalities?: string[];
completeWithItems?: boolean;
notifyError?: string;
approvalRequestMethod?: string;
responseText?: string;
turnStartError?: Error;
preBindNotificationCount?: number;
interruptError?: Error;
unsubscribeError?: Error;
}) {
const notifications = new Set<(notification: CodexServerNotification) => void>();
const closeHandlers = new Set<() => void>();
const requestHandlers = new Set<(request: { method: string }) => JsonValue | undefined>();
const requests: Array<{ method: string; params?: JsonValue }> = [];
const approvalResponses: JsonValue[] = [];
const request = vi.fn(async (method: string, params?: JsonValue) => {
requests.push({ method, params });
if (method === "model/list") {
@@ -123,60 +104,51 @@ function createFakeClient(options?: {
return threadStartResult();
}
if (method === "turn/start") {
if (options?.turnStartError) {
throw options.turnStartError;
}
if (options?.preBindNotificationCount) {
for (let index = 0; index < options.preBindNotificationCount; index += 1) {
for (const notify of notifications) {
notify({
method: "item/started",
params: { threadId: "thread-1", turnId: "turn-1" },
});
if (options?.approvalRequestMethod) {
for (const handler of requestHandlers) {
const response = handler({ method: options.approvalRequestMethod });
if (response !== undefined) {
approvalResponses.push(response);
}
}
return turnStartResult();
}
const emitTurnNotifications = () => {
if (options?.notifyError) {
for (const notify of notifications) {
notify({
method: "error",
params: {
threadId: "thread-1",
turnId: "turn-1",
error: {
message: options.notifyError,
codexErrorInfo: null,
additionalDetails: null,
},
willRetry: false,
if (options?.notifyError) {
for (const notify of notifications) {
notify({
method: "error",
params: {
threadId: "thread-1",
turnId: "turn-1",
error: {
message: options.notifyError,
codexErrorInfo: null,
additionalDetails: null,
},
});
}
} else if (!options?.completeWithItems) {
for (const notify of notifications) {
notify({
method: "item/agentMessage/delta",
params: {
threadId: "thread-1",
turnId: "turn-1",
itemId: "msg-1",
delta: options?.responseText ?? "A red square.",
},
});
notify({
method: "turn/completed",
params: {
threadId: "thread-1",
turnId: "turn-1",
turn: turnStartResult("completed").turn,
},
});
}
willRetry: false,
},
});
}
};
emitTurnNotifications();
} else if (!options?.completeWithItems) {
for (const notify of notifications) {
notify({
method: "item/agentMessage/delta",
params: {
threadId: "thread-1",
turnId: "turn-1",
itemId: "msg-1",
delta: options?.responseText ?? "A red square.",
},
});
notify({
method: "turn/completed",
params: {
threadId: "thread-1",
turnId: "turn-1",
turn: turnStartResult("completed").turn,
},
});
}
}
return turnStartResult(
options?.completeWithItems ? "completed" : "inProgress",
options?.completeWithItems
@@ -192,12 +164,6 @@ function createFakeClient(options?: {
: [],
);
}
if (method === "turn/interrupt" && options?.interruptError) {
throw options.interruptError;
}
if (method === "thread/unsubscribe" && options?.unsubscribeError) {
throw options.unsubscribeError;
}
return {};
});
@@ -207,17 +173,14 @@ function createFakeClient(options?: {
notifications.add(handler);
return () => notifications.delete(handler);
},
addRequestHandler() {
return () => undefined;
},
addCloseHandler(handler: () => void) {
closeHandlers.add(handler);
return () => closeHandlers.delete(handler);
addRequestHandler(handler: (request: { method: string }) => JsonValue | undefined) {
requestHandlers.add(handler);
return () => requestHandlers.delete(handler);
},
close: vi.fn(),
} as unknown as CodexAppServerClient;
return { client, requests };
return { client, requests, approvalResponses };
}
describe("codex media understanding provider", () => {
@@ -229,9 +192,11 @@ describe("codex media understanding provider", () => {
it("runs image understanding through a bounded Codex app-server turn", async () => {
const { client, requests } = createFakeClient();
const clientFactory = vi.fn(async () => client);
const clientFactory = vi.fn(
async (_startOptions, _authProfileId, _agentDir, _config) => client,
);
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(clientFactory),
clientFactory,
});
const cfg = {
auth: {
@@ -254,33 +219,35 @@ describe("codex media understanding provider", () => {
});
expect(result).toEqual({ text: "A red square.", model: "gpt-5.4" });
expect(requests.map((entry) => entry.method)).toEqual([
"model/list",
"thread/start",
"turn/start",
]);
expect(clientFactory).toHaveBeenCalledWith(
expect.any(Object),
undefined,
"/tmp/openclaw-agent",
cfg,
expect.objectContaining({ timeoutMs: 30_000 }),
);
expect(requests.map((entry) => entry.method)).toEqual([
"model/list",
"thread/start",
"turn/start",
"thread/unsubscribe",
]);
expect(requests[0]?.params).toEqual({ limit: 100, cursor: null, includeHidden: true });
expect(requests[1]?.params).toEqual({
model: "gpt-5.4",
modelProvider: "openai",
cwd: "/tmp/openclaw-agent/codex-media-home",
approvalPolicy: "never",
cwd: "/tmp/openclaw-agent",
approvalPolicy: "on-request",
sandbox: "read-only",
serviceName: "OpenClaw",
personality: "none",
developerInstructions:
"You are OpenClaw's bounded image-understanding worker. Describe only the provided image content. Do not call tools, edit files, or ask follow-up questions.",
config: EXPECTED_MEDIA_THREAD_CONFIG,
config: {
"features.code_mode": false,
"features.code_mode_only": false,
},
environments: [],
dynamicTools: [],
experimentalRawEvents: true,
ephemeral: true,
persistExtendedHistory: false,
});
expect(requests[2]?.params).toEqual({
threadId: "thread-1",
@@ -288,6 +255,9 @@ describe("codex media understanding provider", () => {
{ type: "text", text: "Describe briefly.", text_elements: [] },
{ type: "image", url: "data:image/png;base64,aW1hZ2UtYnl0ZXM=" },
],
cwd: "/tmp/openclaw-agent",
approvalPolicy: "on-request",
model: "gpt-5.4",
effort: "low",
});
});
@@ -295,12 +265,8 @@ describe("codex media understanding provider", () => {
it("treats a blank agent directory as absent when starting the app-server", async () => {
const { client, requests } = createFakeClient();
const clientFactory = vi.fn(async () => client);
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(clientFactory),
});
const cfg = {
agents: { list: [{ id: "main", agentDir: "/tmp/openclaw-default-agent" }] },
};
const provider = buildCodexMediaUnderstandingProvider({ clientFactory });
const cfg = {};
await provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
@@ -313,16 +279,9 @@ describe("codex media understanding provider", () => {
agentDir: " ",
});
expect(clientFactory).toHaveBeenCalledWith(
expect.any(Object),
undefined,
"/tmp/openclaw-default-agent",
cfg,
expect.any(Object),
);
expect(requests[1]?.params).toEqual(
expect.objectContaining({ cwd: "/tmp/openclaw-default-agent/codex-media-home" }),
);
expect(clientFactory).toHaveBeenCalledWith(expect.any(Object), undefined, undefined, cfg);
expect(requests[1]?.params).toEqual(expect.objectContaining({ cwd: process.cwd() }));
expect(requests[2]?.params).toEqual(expect.objectContaining({ cwd: process.cwd() }));
});
it("passes the scoped auth store into isolated app-server startup", async () => {
@@ -364,7 +323,7 @@ describe("codex media understanding provider", () => {
try {
const { client } = createFakeClient();
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
const result = await provider.describeImage?.({
@@ -387,97 +346,33 @@ describe("codex media understanding provider", () => {
}
});
it("starts the media deadline before client acquisition", async () => {
vi.useFakeTimers();
it("declines approval requests during image understanding", async () => {
const { client, approvalResponses } = createFakeClient({
approvalRequestMethod: "item/permissions/requestApproval",
});
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(
async () => await new Promise<CodexAppServerClient>(() => {}),
),
clientFactory: async () => client,
});
const description = provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
timeoutMs: 100,
cfg: {},
agentDir: "/tmp/openclaw-agent",
});
const rejected = expect(description).rejects.toThrow(
"Codex app-server image understanding timed out",
);
await vi.advanceTimersByTimeAsync(100);
await rejected;
});
it("retires a media client lease that resolves after its deadline", async () => {
let resolveLease!: (lease: {
client: CodexAppServerClient;
release: () => void;
abandon: () => Promise<void>;
}) => void;
const pendingLease = new Promise<{
client: CodexAppServerClient;
release: () => void;
abandon: () => Promise<void>;
}>((resolve) => {
resolveLease = resolve;
});
const clientLeaseFactory = vi.fn(async () => await pendingLease);
const provider = buildCodexMediaUnderstandingProvider({ clientLeaseFactory });
const description = provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
timeoutMs: 5,
cfg: {},
agentDir: "/tmp/openclaw-agent",
});
await expect(description).rejects.toThrow("Codex app-server image understanding timed out");
const { client } = createFakeClient();
const release = vi.fn();
const abandon = vi.fn(async () => undefined);
resolveLease({ client, release, abandon });
await vi.waitFor(() => expect(abandon).toHaveBeenCalledOnce());
expect(release).not.toHaveBeenCalled();
});
it("releases the bounded route between isolated media calls", async () => {
const { client, requests } = createFakeClient();
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
});
const request = {
await provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
prompt: "Describe briefly.",
timeoutMs: 30_000,
cfg: {},
agentDir: "/tmp/openclaw-agent",
};
});
const first = await provider.describeImage?.(request);
const second = await provider.describeImage?.(request);
expect(first?.text).toBe("A red square.");
expect(second?.text).toBe("A red square.");
expect(requests.filter((entry) => entry.method === "model/list")).toHaveLength(2);
expect(requests.filter((entry) => entry.method === "thread/start")).toHaveLength(2);
expect(approvalResponses).toEqual([{ permissions: {}, scope: "turn" }]);
});
it("extracts text from terminal turn items", async () => {
const { client } = createFakeClient({ completeWithItems: true });
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
const result = await provider.describeImages?.({
@@ -496,7 +391,7 @@ describe("codex media understanding provider", () => {
it("rejects text-only Codex app-server models before starting a turn", async () => {
const { client, requests } = createFakeClient({ inputModalities: ["text"] });
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
await expect(
@@ -517,7 +412,7 @@ describe("codex media understanding provider", () => {
it("surfaces Codex app-server turn errors", async () => {
const { client } = createFakeClient({ notifyError: "vision unavailable" });
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
await expect(
@@ -534,107 +429,12 @@ describe("codex media understanding provider", () => {
).rejects.toThrow("vision unavailable");
});
it.each([
{
name: "structured rejection",
error: new CodexAppServerRpcError({ message: "turn rejected" }, "turn/start"),
abandonCount: 0,
},
{
name: "ambiguous timeout",
error: new Error("turn/start timed out"),
abandonCount: 1,
},
])("handles $name with exact media lease ownership", async ({ error, abandonCount }) => {
const { client } = createFakeClient({ turnStartError: error });
const release = vi.fn();
const abandon = vi.fn(async () => undefined);
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: async () => ({ client, release, abandon }),
});
await expect(
provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
timeoutMs: 30_000,
cfg: {},
agentDir: "/tmp/openclaw-agent",
}),
).rejects.toBe(error);
expect(abandon).toHaveBeenCalledTimes(abandonCount);
expect(release).toHaveBeenCalledTimes(1);
});
it("retires the media client when thread cleanup is unconfirmed", async () => {
const { client } = createFakeClient({ unsubscribeError: new Error("unsubscribe failed") });
const release = vi.fn();
const abandon = vi.fn(async () => undefined);
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: async () => ({ client, release, abandon }),
});
await expect(
provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
timeoutMs: 30_000,
cfg: {},
agentDir: "/tmp/openclaw-agent",
}),
).resolves.toEqual({ text: "A red square.", model: "gpt-5.4" });
expect(abandon).toHaveBeenCalledOnce();
expect(release).not.toHaveBeenCalled();
});
it("retires the media client when an accepted turn cannot be interrupted", async () => {
const { client, requests } = createFakeClient({
preBindNotificationCount: 257,
interruptError: new Error("interrupt timeout"),
});
const release = vi.fn();
const abandon = vi.fn(async () => undefined);
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: async () => ({ client, release, abandon }),
});
await expect(
provider.describeImage?.({
buffer: Buffer.from("image-bytes"),
fileName: "image.png",
mime: "image/png",
provider: "codex",
model: "gpt-5.4",
timeoutMs: 30_000,
cfg: {},
agentDir: "/tmp/openclaw-agent",
}),
).rejects.toThrow("pre-bind notification buffer exceeded 256 entries");
expect(requests.map((entry) => entry.method)).toEqual([
"model/list",
"thread/start",
"turn/start",
"turn/interrupt",
]);
expect(abandon).toHaveBeenCalledOnce();
expect(release).not.toHaveBeenCalled();
});
it("runs structured extraction through the same bounded Codex app-server path", async () => {
const { client, requests } = createFakeClient({
responseText: '{"summary":"red square","tags":["shape"]}',
});
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
const result = await provider.extractStructured?.({
@@ -675,21 +475,25 @@ describe("codex media understanding provider", () => {
"model/list",
"thread/start",
"turn/start",
"thread/unsubscribe",
]);
expect(requests[1]?.params).toEqual({
model: "gpt-5.4",
modelProvider: "openai",
cwd: "/tmp/openclaw-agent/codex-media-home",
approvalPolicy: "never",
cwd: "/tmp/openclaw-agent",
approvalPolicy: "on-request",
sandbox: "read-only",
serviceName: "OpenClaw",
personality: "none",
developerInstructions:
"You are OpenClaw's bounded structured-extraction worker. Return only the requested extraction. Do not call tools, edit files, ask follow-up questions, or include secrets.",
config: EXPECTED_MEDIA_THREAD_CONFIG,
config: {
"features.code_mode": false,
"features.code_mode_only": false,
},
environments: [],
dynamicTools: [],
experimentalRawEvents: true,
ephemeral: true,
persistExtendedHistory: false,
});
const turnParams = requests[2]?.params as
| {
@@ -702,9 +506,9 @@ describe("codex media understanding provider", () => {
}
| undefined;
expect(turnParams?.threadId).toBe("thread-1");
expect(turnParams?.approvalPolicy).toBeUndefined();
expect(turnParams?.model).toBeUndefined();
expect(turnParams?.cwd).toBeUndefined();
expect(turnParams?.approvalPolicy).toBe("on-request");
expect(turnParams?.model).toBe("gpt-5.4");
expect(turnParams?.cwd).toBe("/tmp/openclaw-agent");
expect(turnParams?.effort).toBe("low");
expect(turnParams?.input).toHaveLength(3);
expect(turnParams?.input?.[0]?.type).toBe("text");
@@ -727,7 +531,7 @@ describe("codex media understanding provider", () => {
responseText: '{"summary":"only text"}',
});
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
await expect(
@@ -747,7 +551,7 @@ describe("codex media understanding provider", () => {
it("returns a controlled error when structured JSON parsing fails", async () => {
const { client } = createFakeClient({ responseText: "not json" });
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
await expect(
@@ -776,7 +580,7 @@ describe("codex media understanding provider", () => {
responseText: '{"summary":123,"tags":["shape"]}',
});
const provider = buildCodexMediaUnderstandingProvider({
clientLeaseFactory: adaptCodexTestClientFactory(async () => client),
clientFactory: async () => client,
});
await expect(

View File

@@ -1,35 +1,549 @@
/** Lazy registration facade for Codex-backed media understanding. */
import type { MediaUnderstandingProvider } from "openclaw/plugin-sdk/media-understanding";
/**
* Codex-backed media understanding provider for bounded image description and
* structured extraction turns.
*/
import {
type JsonSchemaObject,
validateJsonSchemaValue,
} from "openclaw/plugin-sdk/json-schema-runtime";
import type {
ImagesDescriptionRequest,
ImagesDescriptionResult,
MediaUnderstandingProvider,
StructuredExtractionRequest,
StructuredExtractionResult,
} from "openclaw/plugin-sdk/media-understanding";
import { resolveTimerTimeoutMs } from "openclaw/plugin-sdk/number-runtime";
import { CODEX_PROVIDER_ID, FALLBACK_CODEX_MODELS } from "./provider-catalog.js";
import type { CodexAppServerClientLeaseFactory } from "./src/app-server/shared-client.js";
import type { CodexAppServerClientFactory } from "./src/app-server/client-factory.js";
import type { CodexAppServerClient } from "./src/app-server/client.js";
import { resolveCodexAppServerRuntimeOptions } from "./src/app-server/config.js";
import { readModelListResult } from "./src/app-server/models.js";
import {
assertCodexThreadStartResponse,
assertCodexTurnStartResponse,
readCodexErrorNotification,
readCodexTurnCompletedNotification,
} from "./src/app-server/protocol-validators.js";
import {
isJsonObject,
type CodexServerNotification,
type CodexThreadItem,
type CodexThreadStartParams,
type CodexTurn,
type CodexTurnStartParams,
type CodexUserInput,
type JsonObject,
type JsonValue,
} from "./src/app-server/protocol.js";
import { buildCodexRuntimeThreadConfig } from "./src/app-server/thread-lifecycle.js";
const DEFAULT_CODEX_IMAGE_MODEL =
FALLBACK_CODEX_MODELS.find((model) => model.inputModalities.includes("image"))?.id ??
FALLBACK_CODEX_MODELS[0]?.id;
const DEFAULT_CODEX_IMAGE_PROMPT = "Describe the image.";
/** Dependencies and plugin config for Codex media-understanding calls. */
export type CodexMediaUnderstandingProviderOptions = {
pluginConfig?: unknown;
resolvePluginConfig?: () => unknown;
clientLeaseFactory?: CodexAppServerClientLeaseFactory;
clientFactory?: CodexAppServerClientFactory;
};
/** Builds a provider whose app-server implementation loads on first use. */
/**
* Builds the media-understanding provider that delegates image tasks to an
* isolated Codex app-server session.
*/
export function buildCodexMediaUnderstandingProvider(
options: CodexMediaUnderstandingProviderOptions = {},
): MediaUnderstandingProvider {
let runtime: Promise<typeof import("./src/media-understanding-provider.runtime.js")> | undefined;
const load = () => (runtime ??= import("./src/media-understanding-provider.runtime.js"));
return {
id: CODEX_PROVIDER_ID,
capabilities: ["image"],
...(DEFAULT_CODEX_IMAGE_MODEL ? { defaultModels: { image: DEFAULT_CODEX_IMAGE_MODEL } } : {}),
describeImage: async ({ buffer, fileName, mime, ...request }) =>
await (
await load()
).describeCodexImages({ ...request, images: [{ buffer, fileName, mime }] }, options),
describeImages: async (request) => await (await load()).describeCodexImages(request, options),
extractStructured: async (request) =>
await (await load()).extractCodexStructured(request, options),
describeImage: async (req) =>
describeCodexImages(
{
images: [
{
buffer: req.buffer,
fileName: req.fileName,
mime: req.mime,
},
],
provider: req.provider,
model: req.model,
prompt: req.prompt,
maxTokens: req.maxTokens,
timeoutMs: req.timeoutMs,
profile: req.profile,
preferredProfile: req.preferredProfile,
authStore: req.authStore,
agentDir: req.agentDir,
cfg: req.cfg,
},
options,
),
describeImages: async (req) => describeCodexImages(req, options),
extractStructured: async (req) => extractCodexStructured(req, options),
};
}
async function describeCodexImages(
req: ImagesDescriptionRequest,
options: CodexMediaUnderstandingProviderOptions,
): Promise<ImagesDescriptionResult> {
const model = req.model.trim();
if (!model) {
throw new Error("Codex image understanding requires model id.");
}
const text = await runBoundedCodexVisionTurn({
model,
profile: req.profile,
timeoutMs: req.timeoutMs,
agentDir: req.agentDir,
authStore: req.authStore,
cfg: req.cfg,
options,
taskLabel: "image understanding",
developerInstructions:
"You are OpenClaw's bounded image-understanding worker. Describe only the provided image content. Do not call tools, edit files, or ask follow-up questions.",
input: [
{ type: "text", text: buildCodexImagePrompt(req), text_elements: [] },
...req.images.map((image) => ({
type: "image" as const,
url: `data:${image.mime ?? "image/png"};base64,${image.buffer.toString("base64")}`,
})),
],
requiredModalities: ["text", "image"],
});
return { text, model };
}
type BoundedCodexVisionTurnParams = {
model: string;
profile?: string;
timeoutMs: number;
agentDir?: string;
authStore?: ImagesDescriptionRequest["authStore"];
cfg: ImagesDescriptionRequest["cfg"];
options: CodexMediaUnderstandingProviderOptions;
taskLabel: string;
developerInstructions: string;
input: CodexUserInput[];
requiredModalities: string[];
};
async function runBoundedCodexVisionTurn(params: BoundedCodexVisionTurnParams): Promise<string> {
const appServer = resolveCodexAppServerRuntimeOptions({
pluginConfig: params.options.pluginConfig,
});
const timeoutMs = resolveTimerTimeoutMs(params.timeoutMs, 100, 100);
const agentDir = params.agentDir?.trim() || undefined;
const cwd = agentDir ?? process.cwd();
const ownsClient = !params.options.clientFactory;
// Tests inject a client factory; production creates an isolated app-server
// client so media tasks cannot reuse the interactive attempt session.
const client = params.options.clientFactory
? await params.options.clientFactory(appServer.start, params.profile, agentDir, params.cfg)
: await import("./src/app-server/shared-client.js").then(
({ createIsolatedCodexAppServerClient }) =>
createIsolatedCodexAppServerClient({
startOptions: appServer.start,
timeoutMs,
authProfileId: params.profile,
agentDir,
authProfileStore: params.authStore,
config: params.cfg,
}),
);
const abortController = new AbortController();
const timeout = setTimeout(() => abortController.abort("timeout"), timeoutMs);
timeout.unref?.();
try {
await assertCodexModelSupportsInput({
client,
model: params.model,
requiredModalities: params.requiredModalities,
timeoutMs,
signal: abortController.signal,
});
const thread = assertCodexThreadStartResponse(
await client.request<unknown>(
"thread/start",
{
model: params.model,
modelProvider: "openai",
cwd,
approvalPolicy: "on-request",
sandbox: "read-only",
serviceName: "OpenClaw",
developerInstructions: params.developerInstructions,
// Media workers are bounded read-only turns; native code mode and
// dynamic tools stay disabled to avoid side effects while inspecting media.
config: buildCodexRuntimeThreadConfig(undefined, { nativeCodeModeEnabled: false }),
environments: [],
dynamicTools: [],
experimentalRawEvents: true,
persistExtendedHistory: false,
ephemeral: true,
} satisfies CodexThreadStartParams,
{ timeoutMs, signal: abortController.signal },
),
);
const collector = createCodexTurnCollector(thread.thread.id, params.taskLabel);
const cleanup = client.addNotificationHandler(collector.handleNotification);
const requestCleanup = client.addRequestHandler(denyCodexImageApprovalRequest);
try {
const turn = assertCodexTurnStartResponse(
await client.request<unknown>(
"turn/start",
{
threadId: thread.thread.id,
input: params.input,
cwd,
approvalPolicy: "on-request",
model: params.model,
effort: "low",
} satisfies CodexTurnStartParams,
{ timeoutMs, signal: abortController.signal },
),
);
const text = await collector.collect(turn.turn, {
timeoutMs,
signal: abortController.signal,
});
return text;
} finally {
requestCleanup();
cleanup();
}
} finally {
clearTimeout(timeout);
if (ownsClient) {
client.close();
}
}
}
async function extractCodexStructured(
req: StructuredExtractionRequest,
options: CodexMediaUnderstandingProviderOptions,
): Promise<StructuredExtractionResult> {
const model = req.model.trim();
if (!model) {
throw new Error("Codex structured extraction requires model id.");
}
const instructions = req.instructions.trim();
if (!instructions) {
throw new Error("Codex structured extraction requires instructions.");
}
if (req.input.length === 0) {
throw new Error("Codex structured extraction requires at least one input.");
}
if (!req.input.some((entry) => entry.type === "image")) {
throw new Error("Codex structured extraction requires at least one image input.");
}
const text = await runBoundedCodexVisionTurn({
model,
profile: req.profile,
timeoutMs: req.timeoutMs,
agentDir: req.agentDir,
authStore: req.authStore,
cfg: req.cfg,
options,
taskLabel: "structured extraction",
developerInstructions:
"You are OpenClaw's bounded structured-extraction worker. Return only the requested extraction. Do not call tools, edit files, ask follow-up questions, or include secrets.",
input: buildCodexStructuredInput(req),
requiredModalities: requiredStructuredModalities(),
});
return normalizeStructuredExtractionResult({ text, model, provider: req.provider, req });
}
function denyCodexImageApprovalRequest(request: { method: string }): JsonValue | undefined {
if (
request.method === "item/commandExecution/requestApproval" ||
request.method === "item/fileChange/requestApproval"
) {
return {
decision: "decline",
reason: "OpenClaw Codex image understanding does not grant tool or file approvals.",
};
}
if (request.method === "item/permissions/requestApproval") {
return { permissions: {}, scope: "turn" };
}
if (request.method.includes("requestApproval")) {
return {
decision: "decline",
reason: "OpenClaw Codex image understanding does not grant native approvals.",
};
}
if (request.method === "mcpServer/elicitation/request") {
return { action: "decline" };
}
return undefined;
}
async function assertCodexModelSupportsInput(params: {
client: CodexAppServerClient;
model: string;
requiredModalities: string[];
timeoutMs: number;
signal: AbortSignal;
}): Promise<void> {
const result = await params.client.request<unknown>(
"model/list",
{ limit: 100, cursor: null, includeHidden: false },
{ timeoutMs: Math.min(params.timeoutMs, 5_000), signal: params.signal },
);
const listed = readModelListResult(result).models;
const match = listed.find((entry) => entry.model === params.model || entry.id === params.model);
if (!match) {
throw new Error(`Codex app-server model not found: ${params.model}`);
}
if (params.requiredModalities.includes("image") && !match.inputModalities.includes("image")) {
throw new Error(`Codex app-server model does not support images: ${params.model}`);
}
if (params.requiredModalities.includes("text") && !match.inputModalities.includes("text")) {
throw new Error(`Codex app-server model does not support text: ${params.model}`);
}
}
function buildCodexImagePrompt(req: ImagesDescriptionRequest): string {
const prompt = req.prompt?.trim() || DEFAULT_CODEX_IMAGE_PROMPT;
if (req.images.length <= 1) {
return prompt;
}
return `${prompt}\n\nAnalyze all ${req.images.length} images together.`;
}
function requiredStructuredModalities(): string[] {
return ["text", "image"];
}
function buildCodexStructuredInput(req: StructuredExtractionRequest): CodexUserInput[] {
return [
{ type: "text", text: buildStructuredExtractionPrompt(req), text_elements: [] },
...req.input.map((entry) => {
if (entry.type === "text") {
return { type: "text" as const, text: entry.text, text_elements: [] };
}
return {
type: "image" as const,
url: `data:${entry.mime ?? "image/png"};base64,${entry.buffer.toString("base64")}`,
};
}),
];
}
function buildStructuredExtractionPrompt(req: StructuredExtractionRequest): string {
return [
req.instructions.trim(),
req.schemaName ? `Schema name: ${req.schemaName}` : undefined,
req.jsonSchema ? `JSON schema:\n${JSON.stringify(req.jsonSchema)}` : undefined,
req.jsonMode === false
? "Return the extraction as concise text."
: "Return valid JSON only. Do not wrap the JSON in Markdown fences.",
]
.filter((part): part is string => Boolean(part))
.join("\n\n");
}
function isJsonSchemaObject(value: unknown): value is JsonSchemaObject {
return typeof value === "object" && value !== null && !Array.isArray(value);
}
function normalizeStructuredExtractionResult(params: {
text: string;
model: string;
provider: string;
req: StructuredExtractionRequest;
}): StructuredExtractionResult {
const result: StructuredExtractionResult = {
text: params.text,
model: params.model,
provider: params.provider,
contentType: params.req.jsonMode === false ? "text" : "json",
};
if (params.req.jsonMode !== false) {
try {
result.parsed = JSON.parse(params.text);
} catch {
throw new Error("Codex structured extraction returned invalid JSON.");
}
if (isJsonSchemaObject(params.req.jsonSchema)) {
const validation = validateJsonSchemaValue({
schema: params.req.jsonSchema,
cacheKey: "codex.media-understanding.extractStructured",
value: result.parsed,
cache: false,
});
if (!validation.ok) {
const message = validation.errors.map((error) => error.text).join("; ") || "invalid";
throw new Error(`Codex structured extraction JSON did not match schema: ${message}`);
}
result.parsed = validation.value;
}
}
return result;
}
function createCodexTurnCollector(threadId: string, taskLabel: string) {
let turnId: string | undefined;
let completedTurn: CodexTurn | undefined;
let promptError: string | undefined;
const pending: CodexServerNotification[] = [];
const assistantTextByItem = new Map<string, string>();
const assistantItemOrder: string[] = [];
let resolveCompletion: (() => void) | undefined;
const completion = new Promise<void>((resolve) => {
resolveCompletion = resolve;
});
const rememberAssistantText = (itemId: string, text: string) => {
if (!text) {
return;
}
if (!assistantTextByItem.has(itemId)) {
assistantItemOrder.push(itemId);
}
assistantTextByItem.set(itemId, text);
};
const handleNotification = (notification: CodexServerNotification): void => {
const params = isJsonObject(notification.params) ? notification.params : undefined;
if (!params || readString(params, "threadId") !== threadId) {
return;
}
if (!turnId) {
pending.push(notification);
return;
}
const notificationTurnId = readNotificationTurnId(params);
if (notificationTurnId !== turnId) {
return;
}
if (notification.method === "item/agentMessage/delta") {
const itemId = readString(params, "itemId") ?? readString(params, "id") ?? "assistant";
const delta = readString(params, "delta") ?? "";
rememberAssistantText(itemId, `${assistantTextByItem.get(itemId) ?? ""}${delta}`);
return;
}
if (notification.method === "turn/completed") {
completedTurn =
readCodexTurnCompletedNotification(notification.params)?.turn ?? completedTurn;
resolveCompletion?.();
return;
}
if (notification.method === "error") {
promptError =
readCodexErrorNotification(notification.params)?.error.message ??
`codex app-server ${taskLabel} turn failed`;
resolveCompletion?.();
}
};
return {
handleNotification,
async collect(
startedTurn: CodexTurn,
options: { timeoutMs: number; signal: AbortSignal },
): Promise<string> {
turnId = startedTurn.id;
if (isTerminalTurn(startedTurn)) {
completedTurn = startedTurn;
}
for (const notification of pending.splice(0)) {
handleNotification(notification);
}
if (!completedTurn && !promptError) {
await waitForTurnCompletion({
completion,
timeoutMs: options.timeoutMs,
signal: options.signal,
taskLabel,
});
}
if (promptError) {
throw new Error(promptError);
}
if (completedTurn?.status === "failed") {
throw new Error(
completedTurn.error?.message ?? `codex app-server ${taskLabel} turn failed`,
);
}
const itemText = collectAssistantTextFromItems(completedTurn?.items);
const deltaText = assistantItemOrder
.map((itemId) => assistantTextByItem.get(itemId)?.trim())
.filter((text): text is string => Boolean(text))
.join("\n\n")
.trim();
const text = (itemText || deltaText).trim();
if (!text) {
throw new Error(`Codex app-server ${taskLabel} turn returned no text.`);
}
return text;
},
};
}
async function waitForTurnCompletion(params: {
completion: Promise<void>;
timeoutMs: number;
signal: AbortSignal;
taskLabel: string;
}): Promise<void> {
let timeout: ReturnType<typeof setTimeout> | undefined;
let cleanupAbort: (() => void) | undefined;
try {
await Promise.race([
params.completion,
new Promise<never>((_, reject) => {
timeout = setTimeout(
() => reject(new Error(`codex app-server ${params.taskLabel} turn timed out`)),
params.timeoutMs,
);
timeout.unref?.();
const abortListener = () =>
reject(new Error(`codex app-server ${params.taskLabel} turn aborted`));
params.signal.addEventListener("abort", abortListener, { once: true });
cleanupAbort = () => params.signal.removeEventListener("abort", abortListener);
}),
]);
} finally {
if (timeout) {
clearTimeout(timeout);
}
cleanupAbort?.();
}
}
function collectAssistantTextFromItems(items: CodexThreadItem[] | undefined): string {
return (items ?? [])
.filter((item) => item.type === "agentMessage")
.map((item) => item.text.trim())
.filter(Boolean)
.join("\n\n")
.trim();
}
function readNotificationTurnId(record: JsonObject): string | undefined {
const direct = readString(record, "turnId");
if (direct) {
return direct;
}
return isJsonObject(record.turn) ? readString(record.turn, "id") : undefined;
}
function readString(record: JsonObject, key: string): string | undefined {
const value = record[key];
return typeof value === "string" ? value : undefined;
}
function isTerminalTurn(turn: CodexTurn): boolean {
return turn.status === "completed" || turn.status === "interrupted" || turn.status === "failed";
}

View File

@@ -1,12 +1,12 @@
{
"name": "@openclaw/codex",
"version": "2026.6.2",
"version": "2026.6.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "@openclaw/codex",
"version": "2026.6.2",
"version": "2026.6.8",
"dependencies": {
"@openai/codex": "0.139.0",
"typebox": "1.1.39",

View File

@@ -1,6 +1,6 @@
{
"name": "@openclaw/codex",
"version": "2026.6.2",
"version": "2026.6.8",
"description": "OpenClaw Codex app-server harness and model provider plugin with a Codex-managed GPT catalog.",
"repository": {
"type": "git",
@@ -23,13 +23,21 @@
"install": {
"npmSpec": "@openclaw/codex",
"defaultChoice": "npm",
"minHostVersion": ">=2026.5.1-beta.1"
"minHostVersion": ">=2026.5.1-beta.1",
"requiredPlatformPackages": [
"@openai/codex-linux-x64",
"@openai/codex-linux-arm64",
"@openai/codex-darwin-x64",
"@openai/codex-darwin-arm64",
"@openai/codex-win32-x64",
"@openai/codex-win32-arm64"
]
},
"compat": {
"pluginApi": ">=2026.6.2"
"pluginApi": ">=2026.6.8"
},
"build": {
"openclawVersion": "2026.6.2"
"openclawVersion": "2026.6.8"
},
"release": {
"publishToClawHub": true,

View File

@@ -4,10 +4,10 @@ import { CODEX_GPT5_BEHAVIOR_CONTRACT } from "./prompt-overlay.js";
import { codexProviderDiscovery } from "./provider-discovery.js";
import { buildCodexProvider, buildCodexProviderCatalog } from "./provider.js";
import { CodexAppServerClient } from "./src/app-server/client.js";
import type { listAllCodexAppServerModels } from "./src/app-server/models.js";
import type { listCodexAppServerModels } from "./src/app-server/models.js";
import {
createIsolatedCodexAppServerClient,
leaseSharedCodexAppServerClient,
getSharedCodexAppServerClient,
resetSharedCodexAppServerClientForTests,
} from "./src/app-server/shared-client.js";
@@ -26,8 +26,7 @@ function createFakeCodexClient(): CodexAppServerClient {
return {
initialize: vi.fn(async () => undefined),
request: vi.fn(async () => ({ data: [] })),
addNotificationHandler: vi.fn(() => () => undefined),
addRequestHandler: vi.fn(() => () => undefined),
setActiveSharedLeaseCountProviderForUnscopedNotifications: vi.fn(),
addCloseHandler: vi.fn(() => () => undefined),
close: vi.fn(),
} as unknown as CodexAppServerClient;
@@ -40,7 +39,7 @@ const TEST_CODEX_APP_SERVER_CONFIG = {
};
async function listTestCodexAppServerModels(
options: Parameters<typeof listAllCodexAppServerModels>[0] = {},
options: Parameters<typeof listCodexAppServerModels>[0] = {},
) {
expect(options.sharedClient).toBe(false);
const client = await createIsolatedCodexAppServerClient({
@@ -184,33 +183,45 @@ describe("codex provider", () => {
expect(resultProvider?.models.map((model) => model.id)).toEqual(["gpt-5.4"]);
});
it("delegates all-page discovery to one model lister call", async () => {
const listModels = vi.fn(async () => ({
models: [
{
id: "gpt-5.4",
model: "gpt-5.4",
hidden: false,
inputModalities: ["text", "image"],
supportedReasoningEfforts: ["medium"],
},
{
id: "gpt-5.5",
model: "gpt-5.5",
hidden: false,
inputModalities: ["text"],
supportedReasoningEfforts: [],
},
],
}));
it("pages through live discovery before building the provider catalog", async () => {
const listModels = vi
.fn()
.mockResolvedValueOnce({
models: [
{
id: "gpt-5.4",
model: "gpt-5.4",
hidden: false,
inputModalities: ["text", "image"],
supportedReasoningEfforts: ["medium"],
},
],
nextCursor: "page-2",
})
.mockResolvedValueOnce({
models: [
{
id: "gpt-5.5",
model: "gpt-5.5",
hidden: false,
inputModalities: ["text"],
supportedReasoningEfforts: [],
},
],
});
const result = await buildCodexProviderCatalog({
env: {},
listModels,
});
expect(listModels).toHaveBeenCalledTimes(1);
expectRecordFields(mockCallArg(listModels, 0), {
cursor: undefined,
limit: 100,
sharedClient: false,
});
expectRecordFields(mockCallArg(listModels, 1), {
cursor: "page-2",
limit: 100,
sharedClient: false,
});
@@ -266,7 +277,7 @@ describe("codex provider", () => {
.mockReturnValueOnce(activeClient)
.mockReturnValueOnce(discoveryClient);
await leaseSharedCodexAppServerClient({
await getSharedCodexAppServerClient({
startOptions: {
transport: "stdio",
command: "/tmp/openclaw-test-codex",

View File

@@ -18,11 +18,16 @@ import {
CODEX_PROVIDER_ID,
FALLBACK_CODEX_MODELS,
} from "./provider-catalog.js";
import type { CodexAppServerStartOptions } from "./src/app-server/config.js";
import {
type CodexAppServerStartOptions,
readCodexPluginConfig,
resolveCodexAppServerRuntimeOptions,
} from "./src/app-server/config.js";
import type {
CodexAppServerModel,
CodexAppServerModelListResult,
} from "./src/app-server/models.js";
import { buildCodexAppServerUsageSnapshot } from "./src/app-server/rate-limits.js";
const DEFAULT_DISCOVERY_TIMEOUT_MS = 2500;
const LIVE_DISCOVERY_ENV = "OPENCLAW_CODEX_DISCOVERY_LIVE";
@@ -34,6 +39,7 @@ const codexCatalogLog = createSubsystemLogger("codex/catalog");
type CodexModelLister = (options: {
timeoutMs: number;
limit?: number;
cursor?: string;
startOptions?: CodexAppServerStartOptions;
sharedClient?: boolean;
}) => Promise<CodexAppServerModelListResult>;
@@ -117,11 +123,6 @@ export function buildCodexProvider(options: BuildCodexProviderOptions = {}): Pro
}
const runtimePluginConfig = resolvePluginConfigObject(ctx.config, CODEX_PROVIDER_ID);
const pluginConfig = runtimePluginConfig ?? (ctx.config ? undefined : options.pluginConfig);
const [{ resolveCodexAppServerRuntimeOptions }, { buildCodexAppServerUsageSnapshot }] =
await Promise.all([
import("./src/app-server/config.js"),
import("./src/app-server/rate-limits.js"),
]);
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig });
const rateLimits = await (options.readRateLimits ?? requestCodexAppServerRateLimitsLazy)({
timeoutMs: ctx.timeoutMs,
@@ -155,15 +156,13 @@ export function buildCodexProvider(options: BuildCodexProviderOptions = {}): Pro
export async function buildCodexProviderCatalog(
options: BuildCatalogOptions = {},
): Promise<{ provider: ModelProviderConfig }> {
const { readCodexPluginConfig, resolveCodexAppServerRuntimeOptions } =
await import("./src/app-server/config.js");
const config = readCodexPluginConfig(options.pluginConfig);
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig: options.pluginConfig });
const timeoutMs = normalizeTimeoutMs(config.discovery?.timeoutMs);
let discovered: CodexAppServerModel[] = [];
if (config.discovery?.enabled !== false && !shouldSkipLiveDiscovery(options.env)) {
discovered = await listModelsBestEffort({
listModels: options.listModels ?? listAllCodexAppServerModelsLazy,
listModels: options.listModels ?? listCodexAppServerModelsLazy,
timeoutMs,
startOptions: appServer.start,
onDiscoveryFailure: options.onDiscoveryFailure,
@@ -201,14 +200,22 @@ async function listModelsBestEffort(params: {
onDiscoveryFailure?: (error: unknown) => void;
}): Promise<CodexAppServerModel[]> {
try {
// The all-pages helper keeps one app-server client alive across pagination.
const result = await params.listModels({
timeoutMs: params.timeoutMs,
limit: MODEL_DISCOVERY_PAGE_LIMIT,
startOptions: params.startOptions,
sharedClient: false,
});
return result.models.filter((model) => !model.hidden);
const models: CodexAppServerModel[] = [];
let cursor: string | undefined;
do {
// App-server model listing is paginated; collect every visible model so
// aliases and picker rows match the current Codex account.
const result = await params.listModels({
timeoutMs: params.timeoutMs,
limit: MODEL_DISCOVERY_PAGE_LIMIT,
cursor,
startOptions: params.startOptions,
sharedClient: false,
});
models.push(...result.models.filter((model) => !model.hidden));
cursor = result.nextCursor;
} while (cursor);
return models;
} catch (error) {
params.onDiscoveryFailure?.(error);
codexCatalogLog.debug("codex model discovery failed; using fallback catalog", {
@@ -218,14 +225,15 @@ async function listModelsBestEffort(params: {
}
}
async function listAllCodexAppServerModelsLazy(options: {
async function listCodexAppServerModelsLazy(options: {
timeoutMs: number;
limit?: number;
cursor?: string;
startOptions?: CodexAppServerStartOptions;
sharedClient?: boolean;
}): Promise<CodexAppServerModelListResult> {
const { listAllCodexAppServerModels } = await import("./src/app-server/models.js");
return listAllCodexAppServerModels(options);
const { listCodexAppServerModels } = await import("./src/app-server/models.js");
return listCodexAppServerModels(options);
}
async function requestCodexAppServerRateLimitsLazy(options: {

View File

@@ -1,6 +1,9 @@
// Codex tests cover app server policy plugin behavior.
import { describe, expect, it } from "vitest";
import { resolveCodexAppServerForOpenClawToolPolicy } from "./app-server-policy.js";
import {
resolveCodexAppServerForModelProvider,
resolveCodexAppServerForOpenClawToolPolicy,
} from "./app-server-policy.js";
import { readCodexPluginConfig, resolveCodexAppServerRuntimeOptions } from "./config.js";
describe("Codex app-server policy", () => {
@@ -66,4 +69,143 @@ describe("Codex app-server policy", () => {
expect(explicitEnv.approvalPolicy).toBe("never");
expect(explicitRequirements.approvalPolicy).toBe("never");
});
it("keeps model-backed reviewers for explicit OpenAI model providers", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
execMode: "auto",
modelProvider: "openai",
});
expect(
resolveCodexAppServerForModelProvider({
appServer,
provider: "codex",
model: "openai/gpt-5.5",
}).approvalsReviewer,
).toBe("auto_review");
expect(
resolveCodexAppServerForModelProvider({
appServer,
provider: "codex",
model: "gpt-5.5",
}).approvalsReviewer,
).toBe("user");
expect(
resolveCodexAppServerForModelProvider({ appServer, provider: "openai" }).approvalsReviewer,
).toBe("auto_review");
});
it("uses human approval for OpenAI-compatible custom endpoints", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
execMode: "auto",
modelProvider: "openai",
model: "gpt-5.5",
config: {
models: {
providers: {
openai: {
baseUrl: "http://localhost:8080/v1",
models: [],
},
},
},
},
});
expect(appServer.approvalsReviewer).toBe("user");
expect(
resolveCodexAppServerForModelProvider({
appServer,
provider: "openai",
model: "gpt-5.5",
config: {
models: {
providers: {
openai: {
baseUrl: "http://localhost:8080/v1",
models: [],
},
},
},
},
}).approvalsReviewer,
).toBe("user");
});
it("uses human approval instead of Codex Guardian for custom model providers", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
execMode: "auto",
modelProvider: "openai",
});
const resolved = resolveCodexAppServerForModelProvider({
appServer,
provider: "lmstudio",
});
const vendorPrefixedModel = resolveCodexAppServerForModelProvider({
appServer,
provider: "openrouter",
model: "openai/gpt-5.5",
});
expect(appServer.approvalsReviewer).toBe("auto_review");
expect(resolved.approvalPolicy).toBe("on-request");
expect(resolved.sandbox).toBe("workspace-write");
expect(resolved.approvalsReviewer).toBe("user");
expect(vendorPrefixedModel.approvalsReviewer).toBe("user");
});
it("infers custom providers from provider-qualified model refs", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
execMode: "auto",
});
expect(
resolveCodexAppServerForModelProvider({
appServer,
model: "lmstudio/local-model",
}).approvalsReviewer,
).toBe("user");
});
it("uses provider-qualified model refs to override broad native provider wrappers", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
execMode: "auto",
});
expect(
resolveCodexAppServerForModelProvider({
appServer,
provider: "codex",
model: "lmstudio/local-model",
}).approvalsReviewer,
).toBe("user");
});
it("downgrades legacy guardian_subagent for custom model providers", () => {
const appServer = resolveCodexAppServerRuntimeOptions({
env: {},
requirementsToml: null,
pluginConfig: {
appServer: {
mode: "guardian",
approvalsReviewer: "guardian_subagent",
},
},
});
expect(
resolveCodexAppServerForModelProvider({ appServer, provider: "local" }).approvalsReviewer,
).toBe("user");
});
});

View File

@@ -2,10 +2,11 @@
* Policy promotion for Codex app-server runs that can safely use OpenClaw tool
* approvals.
*/
import type {
CodexAppServerRuntimeOptions,
CodexPluginConfig,
OpenClawExecPolicyForCodexAppServer,
import {
canUseCodexModelBackedApprovalsReviewerForModel,
type CodexAppServerRuntimeOptions,
type CodexPluginConfig,
type OpenClawExecPolicyForCodexAppServer,
} from "./config.js";
/**
@@ -44,6 +45,35 @@ export function resolveCodexAppServerForOpenClawToolPolicy(params: {
};
}
export function resolveCodexAppServerForModelProvider(params: {
appServer: CodexAppServerRuntimeOptions;
provider?: string;
model?: string;
config?: Parameters<typeof canUseCodexModelBackedApprovalsReviewerForModel>[0]["config"];
env?: NodeJS.ProcessEnv;
agentDir?: string;
codexConfigToml?: string | null;
}): CodexAppServerRuntimeOptions {
const explicitProvider = normalizeModelBackedReviewerProvider(params.provider);
if (
!isCodexModelBackedApprovalsReviewer(params.appServer.approvalsReviewer) ||
canUseCodexModelBackedApprovalsReviewerForModel({
modelProvider: explicitProvider,
model: params.model,
config: params.config,
env: params.env,
agentDir: params.agentDir,
codexConfigToml: params.codexConfigToml,
})
) {
return params.appServer;
}
return {
...params.appServer,
approvalsReviewer: "user",
};
}
function isCodexAppServerPolicyMode(value: unknown): boolean {
return value === "guardian" || value === "yolo";
}
@@ -53,3 +83,12 @@ function isCodexAppServerApprovalPolicy(value: unknown): boolean {
value === "never" || value === "on-request" || value === "on-failure" || value === "untrusted"
);
}
function isCodexModelBackedApprovalsReviewer(value: string): boolean {
return value === "auto_review" || value === "guardian_subagent";
}
function normalizeModelBackedReviewerProvider(provider: string | undefined): string | undefined {
const normalized = provider?.trim().toLowerCase();
return normalized || undefined;
}

View File

@@ -285,7 +285,8 @@ function matchesCurrentTurn(
if (!requestParams) {
return false;
}
const requestThreadId = readString(requestParams, "threadId");
const requestThreadId =
readString(requestParams, "threadId") ?? readString(requestParams, "conversationId");
const requestTurnId = readString(requestParams, "turnId");
return requestThreadId === threadId && requestTurnId === turnId;
}

View File

@@ -2,41 +2,10 @@
import { describe, expect, it, vi } from "vitest";
import {
interruptCodexTurnBestEffort,
runCodexTurnStartWithLease,
settleCodexAppServerClientLease,
unsubscribeCodexThreadBestEffort,
validateCodexThreadCreationResponse,
} from "./attempt-client-cleanup.js";
import { CodexAppServerRpcError } from "./client.js";
describe("Codex app-server attempt client cleanup", () => {
it("keeps the client lease after a structured turn-start rejection", async () => {
const abandon = vi.fn(async () => undefined);
const error = new CodexAppServerRpcError({ message: "turn rejected" }, "turn/start");
await expect(
runCodexTurnStartWithLease({ abandon } as never, async () => {
throw error;
}),
).rejects.toBe(error);
expect(abandon).not.toHaveBeenCalled();
});
it("abandons only the exact client lease after an ambiguous turn-start timeout", async () => {
const abandon = vi.fn(async () => undefined);
const otherAbandon = vi.fn(async () => undefined);
await expect(
runCodexTurnStartWithLease({ abandon } as never, async () => {
throw new Error("turn/start timed out");
}),
).rejects.toThrow("turn/start timed out");
expect(abandon).toHaveBeenCalledTimes(1);
expect(otherAbandon).not.toHaveBeenCalled();
});
it("interrupts turns with optional request timeout", () => {
const request = vi.fn(async () => ({}));
@@ -53,58 +22,7 @@ describe("Codex app-server attempt client cleanup", () => {
);
});
it("unsubscribes a retained thread when its create response is malformed", async () => {
const request = vi.fn(async () => ({}));
const abandon = vi.fn(async () => undefined);
const invalidResponse = { thread: { id: "thread-1" } };
await expect(
validateCodexThreadCreationResponse(
{ client: { request } as never, abandon },
invalidResponse,
() => {
throw new Error("invalid thread/start response");
},
),
).rejects.toThrow("invalid thread/start response");
expect(request).toHaveBeenCalledWith(
"thread/unsubscribe",
{ threadId: "thread-1" },
{ timeoutMs: 5_000 },
);
expect(abandon).not.toHaveBeenCalled();
});
it.each([
["omits the retained thread id", {}, vi.fn(async () => ({}))],
[
"cannot confirm unsubscribe",
{ thread: { id: "thread-1" } },
vi.fn(async () => {
throw new Error("connection lost");
}),
],
])(
"retires the client when a malformed create response %s",
async (_label, response, request) => {
const abandon = vi.fn(async () => undefined);
await expect(
validateCodexThreadCreationResponse(
{ client: { request } as never, abandon },
response,
() => {
throw new Error("invalid thread/start response");
},
),
).rejects.toThrow("subscription could not be released");
expect(abandon).toHaveBeenCalledOnce();
},
);
it("reports unsubscribe cleanup failures", async () => {
it("swallows unsubscribe cleanup failures", async () => {
const request = vi.fn(async () => {
throw new Error("already gone");
});
@@ -114,7 +32,7 @@ describe("Codex app-server attempt client cleanup", () => {
threadId: "thread-1",
timeoutMs: 123,
}),
).resolves.toBe(false);
).resolves.toBeUndefined();
expect(request).toHaveBeenCalledWith(
"thread/unsubscribe",
@@ -122,31 +40,4 @@ describe("Codex app-server attempt client cleanup", () => {
{ timeoutMs: 123 },
);
});
it("returns leases only after thread cleanup is confirmed", async () => {
const release = vi.fn();
const abandon = vi.fn(async () => undefined);
await settleCodexAppServerClientLease(
{ client: { request: vi.fn(async () => ({})) }, release, abandon } as never,
{ threadId: "thread-ok", timeoutMs: 123 },
);
expect(release).toHaveBeenCalledOnce();
expect(abandon).not.toHaveBeenCalled();
release.mockClear();
await settleCodexAppServerClientLease(
{
client: {
request: vi.fn(async () => {
throw new Error("unsubscribe failed");
}),
},
release,
abandon,
} as never,
{ threadId: "thread-stale", timeoutMs: 123 },
);
expect(release).not.toHaveBeenCalled();
expect(abandon).toHaveBeenCalledOnce();
});
});

View File

@@ -2,126 +2,14 @@
* Best-effort cleanup helpers for timed-out or aborted Codex app-server turns.
*/
import { embeddedAgentLog } from "openclaw/plugin-sdk/agent-harness-runtime";
import { CodexAppServerRpcError, type CodexAppServerClient } from "./client.js";
import { isJsonObject, readCodexThreadCreationResponseId } from "./protocol.js";
import type { CodexAppServerClientLease } from "./shared-client.js";
import type { CodexAppServerClient } from "./client.js";
import { retireSharedCodexAppServerClientIfCurrent } from "./shared-client.js";
/** Timeout for best-effort app-server turn interruption during cleanup. */
export const CODEX_APP_SERVER_INTERRUPT_TIMEOUT_MS = 5_000;
/** Timeout for best-effort thread unsubscribe during cleanup. */
export const CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS = 5_000;
/** The connection's thread-subscription ownership can no longer be proven. */
export class CodexAppServerUnsafeSubscriptionError extends Error {
constructor(message: string, options?: ErrorOptions) {
super(message, options);
this.name = "CodexAppServerUnsafeSubscriptionError";
}
}
export function isCodexAppServerUnsafeSubscriptionError(
error: unknown,
): error is CodexAppServerUnsafeSubscriptionError {
return error instanceof CodexAppServerUnsafeSubscriptionError;
}
/** A resume response may only describe the thread this connection retained. */
export function assertCodexThreadResumeSubscription(
requestedThreadId: string,
returnedThreadId: string,
): void {
if (returnedThreadId !== requestedThreadId) {
throw new CodexAppServerUnsafeSubscriptionError(
`Codex thread/resume returned ${returnedThreadId} for ${requestedThreadId}`,
);
}
}
/** Retires the exact client lease when turn acceptance is ambiguous. */
export async function runCodexTurnStartWithLease<T>(
lease: CodexAppServerClientLease,
startTurn: () => Promise<T>,
): Promise<T> {
try {
return await startTurn();
} catch (error) {
// Structured RPC rejection happens before Codex accepts the turn. Transport,
// timeout, and abort failures may hide an accepted turn with an unknown id.
if (!(error instanceof CodexAppServerRpcError)) {
await lease.abandon();
}
throw error;
}
}
/** Retries once when native work wins the race immediately before turn/start. */
export async function runCodexTurnStartWithNativeTurnRetry<T>(params: {
startTurn: () => Promise<T>;
waitForActiveTurnCompletion: () => Promise<boolean>;
afterActiveTurnCompletion?: () => Promise<void>;
onRetry?: () => void;
}): Promise<T> {
try {
return await params.startTurn();
} catch (error) {
if (!isCodexActiveTurnNotSteerableError(error)) {
throw error;
}
params.onRetry?.();
if (!(await params.waitForActiveTurnCompletion())) {
throw error;
}
await params.afterActiveTurnCompletion?.();
return await params.startTurn();
}
}
/** True for Codex's structured rejection when native work already owns the thread. */
export function isCodexActiveTurnNotSteerableError(error: unknown): boolean {
if (!(error instanceof CodexAppServerRpcError) || !isJsonObject(error.data)) {
return false;
}
const info = error.data.codexErrorInfo;
return isJsonObject(info) && isJsonObject(info.activeTurnNotSteerable);
}
/** Validates a create response and retires the client unless cleanup is confirmed. */
export async function validateCodexThreadCreationResponse<T>(
owner: {
client: CodexAppServerClient;
abandon: () => Promise<void>;
},
response: unknown,
validate: (value: unknown) => T,
): Promise<T> {
try {
return validate(response);
} catch (error) {
const threadId = readCodexThreadCreationResponseId(response);
const released = threadId
? await unsubscribeCodexThreadBestEffort(owner.client, {
threadId,
timeoutMs: CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
})
: false;
if (released) {
throw error;
}
try {
await owner.abandon();
} catch (abandonError) {
throw new CodexAppServerUnsafeSubscriptionError(
"Codex thread creation response was invalid and its client could not be retired",
{ cause: abandonError },
);
}
throw new CodexAppServerUnsafeSubscriptionError(
"Codex thread creation response was invalid and its subscription could not be released",
{ cause: error },
);
}
}
/** Sends a turn interrupt without blocking abort cleanup on app-server errors. */
export function interruptCodexTurnBestEffort(
client: CodexAppServerClient,
@@ -148,56 +36,28 @@ export function interruptCodexTurnBestEffort(
}
}
/** Unsubscribes from a thread and reports whether wire cleanup was confirmed. */
/** Unsubscribes from a thread while swallowing cleanup-only failures. */
export async function unsubscribeCodexThreadBestEffort(
client: CodexAppServerClient,
params: {
threadId: string;
timeoutMs: number;
},
): Promise<boolean> {
): Promise<void> {
try {
await client.request(
"thread/unsubscribe",
{ threadId: params.threadId },
{ timeoutMs: params.timeoutMs },
);
return true;
} catch (error) {
embeddedAgentLog.debug("codex app-server thread unsubscribe cleanup failed", {
threadId: params.threadId,
error,
});
return false;
}
}
/** Returns one exact client lease to the pool only after subscription cleanup succeeds. */
export async function settleCodexAppServerClientLease(
lease: CodexAppServerClientLease,
params: {
threadId?: string;
timeoutMs: number;
abandon?: boolean;
},
): Promise<void> {
if (params.abandon) {
await lease.abandon();
return;
}
if (
params.threadId &&
!(await unsubscribeCodexThreadBestEffort(lease.client, {
threadId: params.threadId,
timeoutMs: params.timeoutMs,
}))
) {
await lease.abandon();
return;
}
lease.release();
}
/**
* Retires the shared client after a timed-out turn so later runs do not reuse a
* potentially wedged app-server connection.
@@ -208,9 +68,10 @@ export async function retireCodexAppServerClientAfterTimedOutTurn(
threadId: string;
turnId: string;
reason: string;
abandonClientLease: () => Promise<void>;
},
): Promise<void> {
const retiredSharedClient = retireSharedCodexAppServerClientIfCurrent(client);
const detachedSharedClient = Boolean(retiredSharedClient);
interruptCodexTurnBestEffort(client, {
threadId: params.threadId,
turnId: params.turnId,
@@ -220,10 +81,28 @@ export async function retireCodexAppServerClientAfterTimedOutTurn(
threadId: params.threadId,
timeoutMs: CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
});
await params.abandonClientLease();
let closedClient = retiredSharedClient?.closed ?? false;
if (!detachedSharedClient) {
const close = (client as { close?: () => void }).close;
if (typeof close === "function") {
try {
close.call(client);
closedClient = true;
} catch (error) {
embeddedAgentLog.debug("codex app-server client close failed during timeout cleanup", {
threadId: params.threadId,
turnId: params.turnId,
error,
});
}
}
}
embeddedAgentLog.warn("codex app-server client retired after timed-out turn", {
threadId: params.threadId,
turnId: params.turnId,
reason: params.reason,
detachedSharedClient,
closedClient,
activeSharedClientLeases: retiredSharedClient?.activeLeases ?? 0,
});
}

View File

@@ -9,6 +9,7 @@ import {
isFileChangePatchUpdatedNotification,
isAssistantCommentaryCompletionNotification,
isNativeToolProgressNotification,
isNativeResponseStreamDeltaNotification,
isPendingOpenClawDynamicToolCompletionNotification,
isRawAssistantProgressNotification,
isRawReasoningCompletionNotification,
@@ -16,6 +17,7 @@ import {
isReasoningProgressNotification,
isReasoningItemCompletionNotification,
isRetryableErrorNotification,
isTurnNotification,
readCodexNotificationItem,
readNotificationItemId,
shouldDisarmAssistantCompletionIdleWatch,
@@ -23,7 +25,6 @@ import {
} from "./attempt-notifications.js";
import { CODEX_POST_REASONING_REPLY_IDLE_TIMEOUT_MS } from "./attempt-timeouts.js";
import type { CodexAttemptTurnWatchController } from "./attempt-turn-watches.js";
import { isCodexNotificationForTurn } from "./notification-correlation.js";
import type { CodexServerNotification } from "./protocol.js";
type CodexExecutionPhase =
@@ -69,7 +70,7 @@ export function isTerminalCodexTurnNotificationForTurn(params: {
turnId: string;
currentPromptTexts: string[];
}): boolean {
if (!isCodexNotificationForTurn(params.notification.params, params.threadId, params.turnId)) {
if (!isTurnNotification(params.notification.params, params.threadId, params.turnId)) {
return false;
}
return (
@@ -104,15 +105,16 @@ export function applyCodexTurnNotificationState(params: {
turnCrossedToolHandoff: boolean;
} {
const { notification, turnWatches } = params;
const isCurrentTurnNotification = isCodexNotificationForTurn(
const isCurrentTurnNotification = isTurnNotification(
notification.params,
params.threadId,
params.turnId,
);
const isTurnCompletion = notification.method === "turn/completed" && isCurrentTurnNotification;
const isNativeResponseStreamDelta = isNativeResponseStreamDeltaNotification(notification);
let turnCrossedToolHandoff = params.turnCrossedToolHandoff;
if (isCurrentTurnNotification) {
if (isCurrentTurnNotification && !isNativeResponseStreamDelta) {
turnWatches.touchActivity(`notification:${notification.method}`, {
details: describeNotificationActivity(notification),
attemptProgress: true,
@@ -248,6 +250,7 @@ export function applyCodexTurnNotificationState(params: {
!turnWatches.isCompletionIdleWatchPinnedByTerminalError() &&
notification.method !== "turn/completed" &&
isCurrentTurnNotification &&
!isNativeResponseStreamDelta &&
!trackedDynamicToolCompletion &&
!rawToolOutputCompletion &&
!postToolProgressNeedsTerminalGuard &&

View File

@@ -1,6 +1,11 @@
/**
* Predicates and readers for Codex app-server notification envelopes.
*/
import { asBoolean } from "openclaw/plugin-sdk/string-coerce-runtime";
import {
describeCodexNotificationCorrelation,
isCodexNotificationForTurn,
} from "./notification-correlation.js";
import {
isJsonObject,
type CodexServerNotification,
@@ -211,6 +216,13 @@ export function isNativeToolProgressNotification(notification: CodexServerNotifi
}
}
/** Returns true for raw native response stream delta events. */
export function isNativeResponseStreamDeltaNotification(
notification: CodexServerNotification,
): boolean {
return notification.method.startsWith("response.") && notification.method.endsWith(".delta");
}
/** Returns true for file-change patch update notifications. */
export function isFileChangePatchUpdatedNotification(
notification: CodexServerNotification,
@@ -265,9 +277,74 @@ function readRawAssistantTextPreview(item: JsonObject): string | undefined {
return text.length > 240 ? `${text.slice(0, 237)}...` : text;
}
/** Returns true when notification params correlate to a specific thread/turn. */
export function isTurnNotification(
value: JsonValue | undefined,
threadId: string,
turnId: string,
): boolean {
return isCodexNotificationForTurn(value, threadId, turnId);
}
/** Returns true when a correlated notification belongs to another active run. */
export function isCodexNotificationOutsideActiveRun(
correlation: ReturnType<typeof describeCodexNotificationCorrelation>,
): boolean {
const hasThreadScope = Boolean(correlation.threadId || correlation.nestedTurnThreadId);
if (!hasThreadScope) {
return false;
}
if (!correlation.matchesActiveThread) {
return true;
}
const hasTurnScope = Boolean(correlation.turnId || correlation.nestedTurnId);
return hasTurnScope && correlation.matchesActiveTurn === false;
}
/** Checks request params that must contain the current thread and turn ids. */
export function isCurrentThreadTurnRequestParams(
value: JsonValue | undefined,
threadId: string,
turnId: string,
): boolean {
if (!isJsonObject(value)) {
return false;
}
return readString(value, "threadId") === threadId && readString(value, "turnId") === turnId;
}
/** Checks approval request params, accepting `conversationId` as thread id. */
export function isCurrentApprovalTurnRequestParams(
value: JsonValue | undefined,
threadId: string,
turnId: string,
): boolean {
if (!isJsonObject(value)) {
return false;
}
const requestThreadId = readString(value, "threadId") ?? readString(value, "conversationId");
return requestThreadId === threadId && readString(value, "turnId") === turnId;
}
/** Checks request params where `turnId` may be omitted or null for the thread. */
export function isCurrentThreadOptionalTurnRequestParams(
value: JsonValue | undefined,
threadId: string,
turnId: string,
): boolean {
if (!isJsonObject(value) || readString(value, "threadId") !== threadId) {
return false;
}
const requestTurnId = value.turnId;
return requestTurnId === null || requestTurnId === undefined || requestTurnId === turnId;
}
/** Returns true for app-server error notifications that will retry. */
export function isRetryableErrorNotification(value: JsonValue | undefined): boolean {
return isJsonObject(value) && value.willRetry === true;
if (!isJsonObject(value)) {
return false;
}
return readBoolean(value, "willRetry") === true || readBoolean(value, "will_retry") === true;
}
/** Returns true for terminal app-server thread status strings. */
@@ -342,6 +419,10 @@ function readString(record: JsonObject, key: string): string | undefined {
return typeof value === "string" ? value : undefined;
}
function readBoolean(record: JsonObject, key: string): boolean | undefined {
return asBoolean(record[key]);
}
/** Reads a typed Codex item from notification params when id/type are present. */
export function readCodexNotificationItem(
params: JsonValue | undefined,

View File

@@ -9,16 +9,13 @@ import type {
} from "openclaw/plugin-sdk/agent-harness-runtime";
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
import { startCodexAttemptThread } from "./attempt-startup.js";
import { defaultLeasedCodexAppServerClientFactory } from "./client-factory.js";
import { CodexAppServerClient } from "./client.js";
import { type CodexPluginConfig, resolveCodexAppServerRuntimeOptions } from "./config.js";
import { threadStartResult } from "./run-attempt-test-harness.js";
import {
resetCodexTestBindingStore,
testCodexAppServerBindingStore,
} from "./session-binding.test-helpers.js";
import {
leaseSharedCodexAppServerClient,
resetSharedCodexAppServerClientForTests,
clearSharedCodexAppServerClient,
getLeasedSharedCodexAppServerClient,
releaseLeasedSharedCodexAppServerClient,
} from "./shared-client.js";
import { createClientHarness, createCodexTestModel } from "./test-support.js";
@@ -88,10 +85,12 @@ function startThreadWithHarness(
signal = new AbortController().signal,
overrides?: {
pluginConfig?: CodexPluginConfig;
attemptClientFactory?: (
harness: ClientHarness,
) => Parameters<typeof startCodexAttemptThread>[0]["attemptClientFactory"];
harness?: ClientHarness;
paths?: AttemptPaths;
skipStartSpy?: boolean;
onThreadReserved?: Parameters<typeof startCodexAttemptThread>[0]["onThreadReserved"];
},
) {
const harness = overrides?.harness ?? createClientHarness();
@@ -102,7 +101,8 @@ function startThreadWithHarness(
const effectivePluginConfig = overrides?.pluginConfig ?? pluginConfig;
const run = startCodexAttemptThread({
bindingStore: testCodexAppServerBindingStore,
attemptClientFactory:
overrides?.attemptClientFactory?.(harness) ?? defaultLeasedCodexAppServerClientFactory,
appServer: resolveCodexAppServerRuntimeOptions({ pluginConfig: effectivePluginConfig }),
pluginConfig: effectivePluginConfig,
computerUseConfig: effectivePluginConfig.computerUse ?? { enabled: false },
@@ -123,11 +123,10 @@ function startThreadWithHarness(
sandboxExecServerEnabled: false,
sandbox: null,
contextEngineProjection: undefined,
startupTokenGuard: {},
startupTimeoutMs,
signal,
onStartupTimeout: vi.fn(),
onThreadReserved: overrides?.onThreadReserved,
spawnedBy: undefined,
});
return { harness, run };
@@ -169,13 +168,12 @@ describe("startCodexAttemptThread", () => {
vi.useRealTimers();
vi.stubEnv("CODEX_API_KEY", "");
vi.stubEnv("OPENAI_API_KEY", "");
resetCodexTestBindingStore();
resetSharedCodexAppServerClientForTests();
clearSharedCodexAppServerClient();
});
afterEach(async () => {
vi.useRealTimers();
resetSharedCodexAppServerClientForTests();
clearSharedCodexAppServerClient();
vi.restoreAllMocks();
vi.unstubAllEnvs();
for (const root of tempRoots) {
@@ -184,7 +182,7 @@ describe("startCodexAttemptThread", () => {
tempRoots.clear();
});
it("keeps the shared app-server reusable after a structured startup rejection", async () => {
it("clears the shared app-server when top-level thread startup fails with an app error", async () => {
const { harness, run } = startThreadWithHarness(5_000);
await answerInitialize(harness);
const threadStart = await waitForThreadStart(harness);
@@ -194,57 +192,25 @@ describe("startCodexAttemptThread", () => {
});
await expect(run).rejects.toThrow("Invalid bearer token");
expect(harness.process.stdin.destroyed).toBe(false);
});
it("retires the client when malformed startup cleanup cannot be confirmed", async () => {
const { harness, run } = startThreadWithHarness(5_000);
await answerInitialize(harness);
const threadStart = await waitForThreadStart(harness);
harness.send({ id: threadStart.id, result: { thread: { id: "thread-malformed" } } });
const unsubscribe = await waitForRequest(harness, "thread/unsubscribe");
harness.send({
id: unsubscribe.id,
error: { code: -32000, message: "unsubscribe failed" },
});
await expect(run).rejects.toThrow("subscription could not be released");
expect(harness.process.stdin.destroyed).toBe(true);
});
it("retires the client when route cleanup cannot release the subscription", async () => {
const { harness, run } = startThreadWithHarness(5_000, undefined, {
onThreadReserved: () => {
throw new Error("route integration failed");
},
});
await answerInitialize(harness);
const threadStart = await waitForThreadStart(harness);
harness.send({ id: threadStart.id, result: threadStartResult("thread-route-failed") });
const unsubscribe = await waitForRequest(harness, "thread/unsubscribe");
harness.send({
id: unsubscribe.id,
error: { code: -32000, message: "unsubscribe failed" },
});
await expect(run).rejects.toThrow("Codex startup subscription cleanup failed");
expect(harness.process.stdin.destroyed).toBe(true);
});
it("does not retire a peer-owned client after a structured startup rejection", async () => {
it("retires a failed startup client after another active lease releases", async () => {
const retained = createClientHarness();
const startSpy = vi.spyOn(CodexAppServerClient, "start").mockReturnValue(retained.client);
const replacement = createClientHarness();
const startSpy = vi
.spyOn(CodexAppServerClient, "start")
.mockReturnValueOnce(retained.client)
.mockReturnValueOnce(replacement.client);
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig });
const paths = createAttemptPaths();
const retainedLeasePromise = leaseSharedCodexAppServerClient({
const retainedLease = getLeasedSharedCodexAppServerClient({
startOptions: appServer.start,
agentDir: paths.agentDir,
preparedAuth: {},
});
await answerInitialize(retained);
const retainedLease = await retainedLeasePromise;
expect(retainedLease.client).toBe(retained.client);
await expect(retainedLease).resolves.toBe(retained.client);
const { run } = startThreadWithHarness(5_000, new AbortController().signal, {
harness: retained,
@@ -260,16 +226,17 @@ describe("startCodexAttemptThread", () => {
await expect(run).rejects.toThrow("Invalid bearer token");
expect(retained.process.stdin.destroyed).toBe(false);
retainedLease.release();
const nextLeasePromise = leaseSharedCodexAppServerClient({
expect(releaseLeasedSharedCodexAppServerClient(retained.client)).toBe(true);
await vi.waitFor(() => expect(retained.process.stdin.destroyed).toBe(true));
const replacementLease = getLeasedSharedCodexAppServerClient({
startOptions: appServer.start,
agentDir: paths.agentDir,
preparedAuth: {},
});
const nextLease = await nextLeasePromise;
expect(nextLease.client).toBe(retained.client);
expect(startSpy).toHaveBeenCalledTimes(1);
nextLease.release();
await answerInitialize(replacement);
await expect(replacementLease).resolves.toBe(replacement.client);
expect(startSpy).toHaveBeenCalledTimes(2);
expect(releaseLeasedSharedCodexAppServerClient(replacement.client)).toBe(true);
});
it("clears the shared app-server when startup abandons an in-flight thread request", async () => {
@@ -291,20 +258,18 @@ describe("startCodexAttemptThread", () => {
expect(harness.stdinDestroyed).toBe(true);
});
it("retires abandoned thread startup even when another lease shares the client", async () => {
it("aborts abandoned thread startup when another lease keeps the shared app-server alive", async () => {
const retained = createClientHarness();
vi.spyOn(CodexAppServerClient, "start").mockReturnValue(retained.client);
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig });
const paths = createAttemptPaths();
const retainedLeasePromise = leaseSharedCodexAppServerClient({
const retainedLease = getLeasedSharedCodexAppServerClient({
startOptions: appServer.start,
agentDir: paths.agentDir,
preparedAuth: {},
});
await answerInitialize(retained);
const retainedLease = await retainedLeasePromise;
expect(retainedLease.client).toBe(retained.client);
await expect(retainedLease).resolves.toBe(retained.client);
const { run } = startThreadWithHarness(100, new AbortController().signal, {
harness: retained,
@@ -315,9 +280,11 @@ describe("startCodexAttemptThread", () => {
const threadStart = await waitForThreadStart(retained);
await rejected;
expect(threadStart.id).toBeDefined();
expect(retained.process.stdin.destroyed).toBe(true);
retainedLease.release();
expect(retained.process.stdin.destroyed).toBe(false);
retained.send({ id: threadStart.id, result: { threadId: "late-thread" } });
expect(releaseLeasedSharedCodexAppServerClient(retained.client)).toBe(true);
await vi.waitFor(() => expect(retained.process.stdin.destroyed).toBe(true));
});
it("closes the shared app-server when startup times out during initialize", async () => {
@@ -342,37 +309,45 @@ describe("startCodexAttemptThread", () => {
).toBe(false);
});
it("releases a late startup lease without retiring a peer-owned initializing client", async () => {
const harness = createClientHarness();
const startSpy = vi.spyOn(CodexAppServerClient, "start").mockReturnValue(harness.client);
const paths = createAttemptPaths();
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig });
const peerPromise = leaseSharedCodexAppServerClient({
startOptions: appServer.start,
agentDir: paths.agentDir,
preparedAuth: {},
it("closes a startup client that arrives after startup timeout", async () => {
let observedFactoryOptions:
| {
onStartedClient?: (client: CodexAppServerClient) => void;
abandonSignal?: AbortSignal;
}
| undefined;
let resolveFactoryDone: () => void = () => undefined;
const factoryDone = new Promise<void>((resolve) => {
resolveFactoryDone = resolve;
});
const { run } = startThreadWithHarness(100, new AbortController().signal, {
harness,
paths,
skipStartSpy: true,
const { harness, run } = startThreadWithHarness(100, new AbortController().signal, {
attemptClientFactory:
(factoryHarness) => async (_startOptions, _authProfileId, _agentDir, _config, options) => {
try {
observedFactoryOptions = options;
await new Promise<void>((resolve) => {
setTimeout(resolve, 250);
});
options?.onStartedClient?.(factoryHarness.client);
return factoryHarness.client;
} finally {
resolveFactoryDone();
}
},
});
const rejected = expect(run).rejects.toThrow("codex app-server startup timed out");
await expect(run).rejects.toThrow("codex app-server startup timed out");
expect(harness.stdinDestroyed).toBe(false);
await answerInitialize(harness);
const peer = await peerPromise;
expect(peer.client).toBe(harness.client);
await new Promise<void>((resolve) => {
setImmediate(resolve);
await rejected;
await factoryDone;
await vi.waitFor(() => expect(harness.stdinDestroyed).toBe(true), {
interval: 1,
timeout: 2_000,
});
expect(startSpy).toHaveBeenCalledTimes(1);
expect(
readHarnessMessages(harness.writes).some((write) => write.method === "thread/start"),
).toBe(false);
await peer.abandon();
expect(harness.stdinDestroyed).toBe(true);
expect(observedFactoryOptions?.onStartedClient).toBeTypeOf("function");
expect(observedFactoryOptions?.abandonSignal?.aborted).toBe(true);
});
it("clears the shared app-server when cancellation abandons an in-flight thread request", async () => {

View File

@@ -11,15 +11,9 @@ import {
type resolveSandboxContext,
} from "openclaw/plugin-sdk/agent-harness-runtime";
import { defaultCodexAppInventoryCache } from "./app-inventory-cache.js";
import {
CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
CodexAppServerUnsafeSubscriptionError,
isCodexAppServerUnsafeSubscriptionError,
unsubscribeCodexThreadBestEffort,
} from "./attempt-client-cleanup.js";
import { buildCodexPluginThreadConfigEligibilityLogData } from "./attempt-diagnostics.js";
import { withCodexStartupTimeout } from "./attempt-timeouts.js";
import { ensureCodexAppServerClientRuntime } from "./client-runtime.js";
import type { CodexAppServerClientFactory } from "./client-factory.js";
import { isCodexAppServerConnectionClosedError, type CodexAppServerClient } from "./client.js";
import { ensureCodexComputerUse } from "./computer-use.js";
import {
@@ -54,23 +48,17 @@ import {
releaseCodexSandboxExecServerEnvironment,
type CodexSandboxExecEnvironment,
} from "./sandbox-exec-server.js";
import type { CodexAppServerBindingStore } from "./session-binding.js";
import {
leaseSharedCodexAppServerClient,
type CodexAppServerClientLease,
type CodexAppServerClientLeaseFactory,
clearSharedCodexAppServerClientIfCurrentAndUnclaimed,
clearSharedCodexAppServerClientIfCurrent,
releaseLeasedSharedCodexAppServerClient,
retireSharedCodexAppServerClientIfCurrent,
} from "./shared-client.js";
import type { CodexAppServerStartupTokenGuard } from "./startup-binding.js";
import {
startOrResumeThread,
type CodexAppServerThreadLifecycleBinding,
type CodexContextEngineThreadBootstrapProjection,
} from "./thread-lifecycle.js";
import {
getCodexAppServerTurnRouter,
type CodexAppServerTurnRouter,
type CodexThreadRouteReservation,
} from "./turn-router.js";
const CODEX_APP_SERVER_STARTUP_CONNECTION_CLOSE_MAX_ATTEMPTS = 3;
@@ -78,15 +66,14 @@ type CodexSandboxContext = Awaited<ReturnType<typeof resolveSandboxContext>>;
/** Resources and bindings returned after a Codex attempt thread starts. */
export type StartCodexAttemptThreadResult = {
turnRouter: CodexAppServerTurnRouter;
turnRoute: CodexThreadRouteReservation;
client: CodexAppServerClient;
thread: CodexAppServerThreadLifecycleBinding;
pluginAppServer: CodexAppServerRuntimeOptions;
sandboxEnvironment: CodexSandboxExecEnvironment | undefined;
environmentSelection: CodexTurnEnvironmentParams[] | undefined;
executionCwd: string;
sandboxPolicy: CodexSandboxPolicy | undefined;
clientLease: CodexAppServerClientLease;
mcpElicitationDelegationRequired: boolean;
releaseSharedClientLease: () => void;
restartContextEngineCodexThread: () => Promise<CodexAppServerThreadLifecycleBinding>;
};
@@ -95,8 +82,7 @@ export type StartCodexAttemptThreadResult = {
* run loop must later release.
*/
export async function startCodexAttemptThread(params: {
bindingStore: CodexAppServerBindingStore;
clientLeaseFactory?: CodexAppServerClientLeaseFactory;
attemptClientFactory: CodexAppServerClientFactory;
appServer: CodexAppServerRuntimeOptions;
pluginConfig: CodexPluginConfig;
computerUseConfig: CodexComputerUseConfig;
@@ -119,26 +105,18 @@ export async function startCodexAttemptThread(params: {
sandboxExecServerEnabled: boolean;
sandbox: CodexSandboxContext;
contextEngineProjection: CodexContextEngineThreadBootstrapProjection | undefined;
expectedResumeThreadId?: string;
startupTokenGuard: CodexAppServerStartupTokenGuard;
startupTimeoutMs: number;
signal: AbortSignal;
onStartupTimeout: () => void | Promise<void>;
onThreadReserved?: (client: CodexAppServerClient, threadId: string) => () => void;
spawnedBy: EmbeddedRunAttemptParams["spawnedBy"];
}): Promise<StartCodexAttemptThreadResult> {
let mcpElicitationDelegationRequired = false;
let sharedClientLease: CodexAppServerClientLease | undefined;
let pluginAppServer = params.appServer;
let releaseSharedClientLease: (() => void) | undefined;
let startupClientForAbandonedRequestCleanup: CodexAppServerClient | undefined;
let releaseStartupResourcesOnTimeout: (() => Promise<void>) | undefined;
let startupAbandoned = false;
const startupAbandonController = new AbortController();
const abandonStartupAcquire = () => startupAbandonController.abort();
const abandonStartupClient = async () => {
const lease = sharedClientLease;
sharedClientLease = undefined;
if (lease) {
await lease.abandon();
}
};
params.signal.addEventListener("abort", abandonStartupAcquire, { once: true });
try {
const startupResult = await withCodexStartupTimeout({
@@ -149,7 +127,10 @@ export async function startCodexAttemptThread(params: {
startupAbandonController.abort();
await params.onStartupTimeout();
await releaseStartupResourcesOnTimeout?.();
await abandonStartupClient();
releaseSharedClientLease?.();
releaseSharedClientLease = undefined;
await closeAbandonedStartupClient(startupClientForAbandonedRequestCleanup);
startupClientForAbandonedRequestCleanup = undefined;
},
operation: async () => {
const threadConfig = mergeCodexThreadConfigs(
@@ -180,9 +161,8 @@ export async function startCodexAttemptThread(params: {
const resolvedPluginPolicy = pluginThreadConfigRequired
? resolveCodexPluginsPolicy(pluginThreadConfigPluginConfig)
: undefined;
const computerUseMcpElicitationDelegationRequired =
params.computerUseConfig.enabled === true;
mcpElicitationDelegationRequired =
const computerUseMcpElicitationDelegationRequired = params.computerUseConfig.enabled;
const mcpElicitationDelegationRequired =
resolvedPluginPolicy?.enabled === true || computerUseMcpElicitationDelegationRequired;
const enabledPluginConfigKeys = resolvedPluginPolicy
? resolvedPluginPolicy.pluginPolicies
@@ -204,48 +184,55 @@ export async function startCodexAttemptThread(params: {
appServer: params.appServer,
}),
);
const pluginAppServer = mcpElicitationDelegationRequired
pluginAppServer = mcpElicitationDelegationRequired
? {
...params.appServer,
approvalPolicy: withMcpElicitationsApprovalPolicy(params.appServer.approvalPolicy),
}
: params.appServer;
let attemptedClientAbandoned = false;
let attemptedClient: CodexAppServerClient | undefined;
const startupAttempt = async () => {
let startupClientLease: CodexAppServerClientLease | undefined;
let clientWorkStarted = false;
attemptedClientAbandoned = false;
let startupClientLease: (() => void) | undefined;
let startupClient: CodexAppServerClient | undefined;
let startupAttemptError: unknown;
let startupAttemptSucceeded = false;
try {
startupClientLease = await (
params.clientLeaseFactory ?? leaseSharedCodexAppServerClient
)({
startOptions: params.appServer.start,
authProfileId: params.startupAuthProfileId,
agentDir: params.agentDir,
config: params.config,
preparedAuth: {
profileId: params.startupAuthProfileId,
cacheKey: params.startupAuthAccountCacheKey ?? params.startupEnvApiKeyCacheKey,
startupClient = await params.attemptClientFactory(
params.appServer.start,
params.startupAuthProfileId,
params.agentDir,
params.config,
{
onStartedClient: (client) => {
// Timeout cleanup may fire before the client factory resolves;
// close any late-arriving client instead of leaking a lease.
startupClientForAbandonedRequestCleanup = client;
if (startupAbandoned || startupAbandonController.signal.aborted) {
void closeAbandonedStartupClient(client);
}
},
abandonSignal: startupAbandonController.signal,
},
abandonSignal: startupAbandonController.signal,
});
const activeStartupLease = startupClientLease;
const activeStartupClient = activeStartupLease.client;
sharedClientLease = startupClientLease;
);
const activeStartupClient = startupClient;
let startupClientLeaseReleased = false;
startupClientLease = () => {
if (startupClientLeaseReleased) {
return;
}
startupClientLeaseReleased = true;
releaseLeasedSharedCodexAppServerClient(activeStartupClient);
};
releaseSharedClientLease = startupClientLease;
attemptedClient = activeStartupClient;
startupClientForAbandonedRequestCleanup = activeStartupClient;
if (startupAbandoned) {
throw new Error("codex app-server startup timed out");
}
if (startupAbandonController.signal.aborted) {
throw new Error("codex app-server startup aborted");
}
clientWorkStarted = true;
ensureCodexAppServerClientRuntime(activeStartupClient, {
agentDir: params.agentDir,
authProfileId: params.startupAuthProfileId,
config: params.config,
});
const turnRouter = getCodexAppServerTurnRouter(activeStartupClient);
await ensureCodexComputerUse({
client: activeStartupClient,
pluginConfig: params.pluginConfig,
@@ -277,6 +264,7 @@ export async function startCodexAttemptThread(params: {
: undefined;
startupSandboxEnvironmentAcquired = Boolean(startupSandboxEnvironment);
if (startupAbandonController.signal.aborted) {
await releaseStartupSandboxEnvironment();
throw new Error("codex app-server startup aborted");
}
if (
@@ -305,57 +293,9 @@ export async function startCodexAttemptThread(params: {
const startupSandboxPolicy = startupSandboxEnvironment
? resolveCodexExternalSandboxPolicyForOpenClawSandbox(params.sandbox)
: undefined;
let startupReservation:
| { route: CodexThreadRouteReservation; release: () => void }
| undefined;
const reserveStartupThread = (threadId: string) => {
if (startupReservation) {
if (startupReservation.route.threadId !== threadId) {
throw new Error(
`codex app-server reserved ${startupReservation.route.threadId} but started ${threadId}`,
);
}
return { release: startupReservation.release };
}
const route = turnRouter.reserveThread({
threadId,
releaseOn: params.signal,
});
let releaseIntegration: (() => void) | undefined;
try {
releaseIntegration = params.onThreadReserved?.(activeStartupClient, threadId);
} catch (error) {
route.release();
throw error;
}
let released = false;
const release = () => {
if (released) {
return;
}
released = true;
if (startupReservation?.route === route) {
startupReservation = undefined;
}
route.release();
releaseIntegration?.();
};
startupReservation = { route, release };
return { release };
};
const releaseStartupResources = async () => {
startupReservation?.release();
await releaseStartupSandboxEnvironment();
};
releaseStartupResourcesOnTimeout = releaseStartupResources;
const buildThreadLifecycleParams = (
signal: AbortSignal,
options: { freshStartOnly?: boolean } = {},
) =>
const buildThreadLifecycleParams = (signal: AbortSignal) =>
({
client: activeStartupClient,
abandonClient: activeStartupLease.abandon,
bindingStore: params.bindingStore,
params: params.buildAttemptParams(),
agentId: params.sessionAgentId,
cwd: startupExecutionCwd,
@@ -373,13 +313,7 @@ export async function startCodexAttemptThread(params: {
mcpServersFingerprintEvaluated: params.bundleMcpThreadConfig.evaluated,
environmentSelection: startupEnvironmentSelection,
contextEngineProjection: params.contextEngineProjection,
freshStartOnly: options.freshStartOnly,
expectedResumeThreadId: options.freshStartOnly
? undefined
: params.expectedResumeThreadId,
signal,
reserveResumeThread: options.freshStartOnly ? undefined : reserveStartupThread,
startupTokenGuard: params.startupTokenGuard,
pluginThreadConfig: pluginThreadConfigRequired
? {
enabled: true,
@@ -403,65 +337,57 @@ export async function startCodexAttemptThread(params: {
const startupThread = await startOrResumeThread(
buildThreadLifecycleParams(startupAbandonController.signal),
);
try {
reserveStartupThread(startupThread.threadId);
} catch (error) {
const unsubscribed = await unsubscribeCodexThreadBestEffort(activeStartupClient, {
threadId: startupThread.threadId,
timeoutMs: CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
});
if (!unsubscribed) {
throw new CodexAppServerUnsafeSubscriptionError(
"Codex startup subscription cleanup failed",
{ cause: error },
);
}
throw error;
}
if (startupAbandonController.signal.aborted) {
await releaseStartupSandboxEnvironment();
throw new Error("codex app-server startup aborted");
}
if (!startupReservation) {
throw new Error("codex app-server startup did not reserve its thread route");
}
startupSandboxEnvironmentAcquired = false;
startupAttemptSucceeded = true;
return {
turnRouter,
turnRoute: startupReservation.route,
client: activeStartupClient,
thread: startupThread,
sandboxEnvironment: startupSandboxEnvironment,
environmentSelection: startupEnvironmentSelection,
executionCwd: startupExecutionCwd,
sandboxPolicy: startupSandboxPolicy,
restartContextEngineCodexThread: () =>
startOrResumeThread(
buildThreadLifecycleParams(params.signal, { freshStartOnly: true }),
),
startOrResumeThread(buildThreadLifecycleParams(params.signal)),
};
} catch (error) {
await releaseStartupResources();
await releaseStartupSandboxEnvironment();
throw error;
} finally {
if (releaseStartupResourcesOnTimeout === releaseStartupResources) {
if (releaseStartupResourcesOnTimeout === releaseStartupSandboxEnvironment) {
releaseStartupResourcesOnTimeout = undefined;
}
}
} catch (error) {
if (sharedClientLease === startupClientLease) {
sharedClientLease = undefined;
}
const shouldAbandonStartupClient =
clientWorkStarted &&
(startupAbandoned ||
params.signal.aborted ||
isIndeterminateCodexStartupFailure(error));
if (shouldAbandonStartupClient) {
attemptedClientAbandoned = true;
await startupClientLease?.abandon();
} else {
startupClientLease?.release();
}
startupAttemptError = error;
throw error;
} finally {
if (!startupAttemptSucceeded) {
if (releaseSharedClientLease === startupClientLease) {
releaseSharedClientLease = undefined;
}
startupClientLease?.();
if (startupAbandoned || params.signal.aborted) {
if (startupClientForAbandonedRequestCleanup === startupClient) {
startupClientForAbandonedRequestCleanup = undefined;
}
await closeAbandonedStartupClient(startupClient);
} else if (
shouldClearSharedClientAfterStartupRace(startupAttemptError) ||
shouldClearSharedClientAfterStartupFailure({
error: startupAttemptError,
spawnedBy: params.spawnedBy,
})
) {
if (startupClientForAbandonedRequestCleanup === startupClient) {
startupClientForAbandonedRequestCleanup = undefined;
}
await evictFailedStartupClient(startupClient);
}
}
}
};
@@ -476,13 +402,18 @@ export async function startCodexAttemptThread(params: {
if (params.signal.aborted || !isCodexAppServerConnectionClosedError(error)) {
throw error;
}
const failedClient = attemptedClient;
const clearedSharedClient = clearSharedCodexAppServerClientIfCurrent(failedClient);
if (startupClientForAbandonedRequestCleanup === failedClient) {
startupClientForAbandonedRequestCleanup = undefined;
}
if (attempt >= CODEX_APP_SERVER_STARTUP_CONNECTION_CLOSE_MAX_ATTEMPTS) {
embeddedAgentLog.warn(
"codex app-server connection closed during startup; retries exhausted",
{
attempt,
maxAttempts: CODEX_APP_SERVER_STARTUP_CONNECTION_CLOSE_MAX_ATTEMPTS,
abandonedSharedClient: attemptedClientAbandoned,
clearedSharedClient,
error: formatErrorMessage(error),
},
);
@@ -494,7 +425,7 @@ export async function startCodexAttemptThread(params: {
attempt,
nextAttempt: attempt + 1,
maxAttempts: CODEX_APP_SERVER_STARTUP_CONNECTION_CLOSE_MAX_ATTEMPTS,
abandonedSharedClient: attemptedClientAbandoned,
clearedSharedClient,
error: formatErrorMessage(error),
},
);
@@ -503,21 +434,32 @@ export async function startCodexAttemptThread(params: {
throw new Error("codex app-server startup retry loop exited unexpectedly");
},
});
const completedSharedClientLease = sharedClientLease;
if (!completedSharedClientLease) {
startupClientForAbandonedRequestCleanup = undefined;
if (!releaseSharedClientLease) {
throw new Error("codex app-server startup succeeded without a shared client lease");
}
sharedClientLease = undefined;
return {
...startupResult,
mcpElicitationDelegationRequired,
clientLease: completedSharedClientLease,
pluginAppServer,
releaseSharedClientLease,
};
} catch (error) {
const shouldAbandonStartupClient =
params.signal.aborted || isIndeterminateCodexStartupFailure(error);
if (shouldAbandonStartupClient) {
await abandonStartupClient();
if (params.signal.aborted || shouldClearSharedClientAfterStartupAbandon(error)) {
releaseSharedClientLease?.();
releaseSharedClientLease = undefined;
await closeAbandonedStartupClient(startupClientForAbandonedRequestCleanup);
startupClientForAbandonedRequestCleanup = undefined;
} else if (
shouldClearSharedClientAfterStartupRace(error) ||
shouldClearSharedClientAfterStartupFailure({
error,
spawnedBy: params.spawnedBy,
})
) {
releaseSharedClientLease?.();
releaseSharedClientLease = undefined;
await evictFailedStartupClient(startupClientForAbandonedRequestCleanup);
startupClientForAbandonedRequestCleanup = undefined;
}
throw error;
} finally {
@@ -525,13 +467,104 @@ export async function startCodexAttemptThread(params: {
}
}
function isIndeterminateCodexStartupFailure(error: unknown): boolean {
async function closeAbandonedStartupClient(
client: CodexAppServerClient | undefined,
): Promise<void> {
if (!client) {
return;
}
const unclaimedSharedClient = clearSharedCodexAppServerClientIfCurrentAndUnclaimed(client);
if (unclaimedSharedClient.closed) {
await closeClientAndWaitIfAvailable(client);
return;
}
if (unclaimedSharedClient.found) {
const retired = retireSharedCodexAppServerClientIfCurrent(client);
if (retired?.closed) {
await closeClientAndWaitIfAvailable(client);
}
return;
}
const retiredSharedClient = retireSharedCodexAppServerClientIfCurrent(client);
if (retiredSharedClient) {
if (retiredSharedClient.closed) {
await closeClientAndWaitIfAvailable(client);
}
return;
}
if (clearSharedCodexAppServerClientIfCurrent(client)) {
await closeClientAndWaitIfAvailable(client);
return;
}
await closeClientAndWaitIfAvailable(client);
}
async function closeClientAndWaitIfAvailable(client: CodexAppServerClient): Promise<void> {
const closeable = client as {
close?: CodexAppServerClient["close"];
closeAndWait?: CodexAppServerClient["closeAndWait"];
};
if (typeof closeable.closeAndWait === "function") {
await closeable.closeAndWait();
return;
}
closeable.close?.();
}
async function evictFailedStartupClient(client: CodexAppServerClient | undefined): Promise<void> {
if (!client) {
return;
}
const unclaimedSharedClient = clearSharedCodexAppServerClientIfCurrentAndUnclaimed(client);
if (unclaimedSharedClient.closed) {
await closeClientAndWaitIfAvailable(client);
return;
}
if (unclaimedSharedClient.found) {
const retired = retireSharedCodexAppServerClientIfCurrent(client);
if (retired?.closed) {
await closeClientAndWaitIfAvailable(client);
}
return;
}
const retiredSharedClient = retireSharedCodexAppServerClientIfCurrent(client);
if (retiredSharedClient) {
if (retiredSharedClient.closed) {
await closeClientAndWaitIfAvailable(client);
}
return;
}
if (clearSharedCodexAppServerClientIfCurrent(client)) {
await closeClientAndWaitIfAvailable(client);
return;
}
await closeClientAndWaitIfAvailable(client);
}
function shouldClearSharedClientAfterStartupAbandon(error: unknown): boolean {
return (
isCodexAppServerUnsafeSubscriptionError(error) ||
isCodexAppServerConnectionClosedError(error) ||
(error instanceof Error &&
(error.message.endsWith(" timed out") ||
error.message.endsWith(" aborted") ||
error.message.includes("write EPIPE")))
error instanceof Error &&
(error.message === "codex app-server startup timed out" ||
error.message === "codex app-server startup aborted")
);
}
function shouldClearSharedClientAfterStartupRace(error: unknown): boolean {
return (
error instanceof Error &&
(shouldClearSharedClientAfterStartupAbandon(error) || error.message.endsWith(" timed out"))
);
}
function shouldClearSharedClientAfterStartupFailure(params: {
error: unknown;
spawnedBy: EmbeddedRunAttemptParams["spawnedBy"];
}): boolean {
if (!(params.error instanceof Error)) {
return !params.spawnedBy;
}
if (params.error.message.includes("write EPIPE")) {
return true;
}
return !params.spawnedBy;
}

View File

@@ -159,39 +159,6 @@ describe("Codex app-server attempt timeouts", () => {
expect(events).toEqual(["cleanup-start", "cleanup-done"]);
});
it("keeps the timeout result when startup resolves during timeout cleanup", async () => {
vi.useFakeTimers();
const events: string[] = [];
let resolveOperation!: (value: string) => void;
let finishCleanup!: () => void;
const run = withCodexStartupTimeout({
timeoutMs: 10,
signal: new AbortController().signal,
onTimeout: async () => {
events.push("cleanup-start");
await new Promise<void>((resolve) => {
finishCleanup = resolve;
});
events.push("cleanup-done");
},
operation: () =>
new Promise<string>((resolve) => {
resolveOperation = resolve;
}),
});
const rejected = expect(run).rejects.toThrow("codex app-server startup timed out");
await vi.advanceTimersByTimeAsync(10);
expect(events).toEqual(["cleanup-start"]);
resolveOperation("late-ready");
await Promise.resolve();
expect(events).toEqual(["cleanup-start"]);
finishCleanup();
await rejected;
expect(events).toEqual(["cleanup-start", "cleanup-done"]);
});
it("rejects startup timeout when aborted before completion", async () => {
vi.useFakeTimers();
const controller = new AbortController();

View File

@@ -52,13 +52,13 @@ export async function withCodexStartupTimeout<T>(params: {
};
timeout = setTimeout(() => {
timeoutError = new Error("codex app-server startup timed out");
rejectOnce(timeoutError);
timeoutCleanup = Promise.resolve()
.then(() => params.onTimeout?.())
.then(
() => undefined,
() => undefined,
);
timeoutCleanup = Promise.resolve(params.onTimeout?.()).then(
() => undefined,
() => undefined,
);
void timeoutCleanup.finally(() => {
rejectOnce(timeoutError!);
});
}, params.timeoutMs);
const abortListener = () => rejectOnce(new Error("codex app-server startup aborted"));
params.signal.addEventListener("abort", abortListener, { once: true });

View File

@@ -29,7 +29,7 @@ describe("Codex app-server attempt turn watches", () => {
const progress: string[] = [];
const diagnostics: string[] = [];
const controller = createCodexAttemptTurnWatchController({
getThreadId: () => "thread-1",
threadId: "thread-1",
signal: abortController.signal,
getTurnId: () => "turn-1",
isCompleted: () => completed,

View File

@@ -29,7 +29,7 @@ export type CodexAttemptTurnWatchController = ReturnType<
* notifications and tool handoffs progress.
*/
export function createCodexAttemptTurnWatchController(params: {
getThreadId: () => string;
threadId: string;
signal: AbortSignal;
getTurnId: () => string | undefined;
isCompleted: () => boolean;
@@ -79,7 +79,6 @@ export function createCodexAttemptTurnWatchController(params: {
const turnTerminalIdleTimeoutMs = resolveTimerTimeoutMs(params.turnTerminalIdleTimeoutMs, 1);
const interruptTimeoutMs = resolveTimerTimeoutMs(params.interruptTimeoutMs, 1);
const resolveWatchTimeoutMs = (timeoutMs: number) => resolveTimerTimeoutMs(timeoutMs, 1);
const currentThreadId = () => params.getThreadId();
const clearCompletionIdleTimer = () => {
if (completionIdleTimer) {
@@ -228,7 +227,7 @@ export function createCodexAttemptTurnWatchController(params: {
clearTerminalIdleTimer();
const turnId = params.getTurnId();
params.onRecordEvent("turn.assistant_completion_idle_release", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId,
idleMs,
timeoutMs: turnAssistantCompletionIdleTimeoutMs,
@@ -237,7 +236,7 @@ export function createCodexAttemptTurnWatchController(params: {
embeddedAgentLog.warn(
"codex app-server turn released after completed assistant item without terminal event",
{
threadId: currentThreadId(),
threadId: params.threadId,
turnId,
idleMs,
timeoutMs: turnAssistantCompletionIdleTimeoutMs,
@@ -246,7 +245,7 @@ export function createCodexAttemptTurnWatchController(params: {
);
if (turnId) {
params.onInterruptTurn({
threadId: currentThreadId(),
threadId: params.threadId,
turnId,
timeoutMs: interruptTimeoutMs,
});
@@ -279,7 +278,7 @@ export function createCodexAttemptTurnWatchController(params: {
params.onTimeout(timeout);
params.onMarkTimedOut();
params.onRecordEvent("turn.progress_idle_timeout", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs: timeout.timeoutMs,
@@ -287,7 +286,7 @@ export function createCodexAttemptTurnWatchController(params: {
...timeout.details,
});
embeddedAgentLog.warn("codex app-server turn idle timed out waiting for progress", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs: timeout.timeoutMs,
@@ -332,7 +331,7 @@ export function createCodexAttemptTurnWatchController(params: {
params.onTimeout(timeout);
params.onMarkTimedOut();
params.onRecordEvent("turn.completion_idle_timeout", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs,
@@ -340,7 +339,7 @@ export function createCodexAttemptTurnWatchController(params: {
...timeout.details,
});
embeddedAgentLog.warn("codex app-server turn idle timed out waiting for completion", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs,
@@ -375,7 +374,7 @@ export function createCodexAttemptTurnWatchController(params: {
params.onTimeout(timeout);
params.onMarkTimedOut();
params.onRecordEvent("turn.terminal_idle_timeout", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs: timeout.timeoutMs,
@@ -383,7 +382,7 @@ export function createCodexAttemptTurnWatchController(params: {
...timeout.details,
});
embeddedAgentLog.warn("codex app-server turn idle timed out waiting for terminal event", {
threadId: currentThreadId(),
threadId: params.threadId,
turnId: params.getTurnId(),
idleMs,
timeoutMs: timeout.timeoutMs,
@@ -458,11 +457,9 @@ export function createCodexAttemptTurnWatchController(params: {
details?: Record<string, unknown>;
attemptProgress?: boolean;
attemptTimeoutMs?: number;
receivedAtMs?: number;
},
) => {
const now = Date.now();
completionLastActivityAt = Math.min(now, options?.receivedAtMs ?? now);
completionLastActivityAt = Date.now();
completionLastActivityReason = `notification:${method}`;
if (options?.details !== undefined) {
completionLastActivityDetails = options.details;

View File

@@ -8,56 +8,37 @@ import {
} from "openclaw/plugin-sdk/agent-harness";
import { AUTH_PROFILE_RUNTIME_CONTRACT } from "openclaw/plugin-sdk/agent-runtime-test-contracts";
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
import type { CodexAppServerClientFactory } from "./client-factory.js";
import { runCodexAppServerAttempt as runCodexAppServerAttemptImpl } from "./run-attempt.js";
import {
readCodexAppServerBinding,
registerCodexTestSessionIdentity,
resetCodexTestBindingStore,
testCodexAppServerBindingStore,
writeCodexAppServerBinding,
} from "./session-binding.test-helpers.js";
import type { CodexAppServerClientLeaseFactory } from "./shared-client.js";
import {
adaptCodexTestClientFactory,
createCodexTestModel,
type CodexTestAppServerClientFactory,
} from "./test-support.js";
import { readCodexAppServerBinding, writeCodexAppServerBinding } from "./session-binding.js";
import { createCodexTestModel } from "./test-support.js";
let codexAppServerClientLeaseFactoryForTest: CodexAppServerClientLeaseFactory | undefined;
let codexAppServerClientFactoryForTest: CodexAppServerClientFactory | undefined;
type RunCodexAppServerAttemptImplOptions = NonNullable<
type RunCodexAppServerAttemptOptions = NonNullable<
Parameters<typeof runCodexAppServerAttemptImpl>[1]
>;
type RunCodexAppServerAttemptOptions = Omit<RunCodexAppServerAttemptImplOptions, "bindingStore"> & {
bindingStore?: RunCodexAppServerAttemptImplOptions["bindingStore"];
};
function setCodexAppServerClientFactoryForTest(factory: CodexTestAppServerClientFactory): void {
codexAppServerClientLeaseFactoryForTest = adaptCodexTestClientFactory(factory);
function setCodexAppServerClientFactoryForTest(factory: CodexAppServerClientFactory): void {
codexAppServerClientFactoryForTest = factory;
}
function resetCodexAppServerClientFactoryForTest(): void {
codexAppServerClientLeaseFactoryForTest = undefined;
codexAppServerClientFactoryForTest = undefined;
}
function runCodexAppServerAttempt(
params: EmbeddedRunAttemptParams,
options: RunCodexAppServerAttemptOptions = {},
) {
const clientLeaseFactory = options.clientLeaseFactory ?? codexAppServerClientLeaseFactoryForTest;
return runCodexAppServerAttemptImpl(params, {
...options,
bindingStore: options.bindingStore ?? testCodexAppServerBindingStore,
...(clientLeaseFactory ? { clientLeaseFactory } : {}),
});
const clientFactory = options.clientFactory ?? codexAppServerClientFactoryForTest;
return runCodexAppServerAttemptImpl(
params,
clientFactory ? { ...options, clientFactory } : options,
);
}
function createParams(sessionFile: string, workspaceDir: string): EmbeddedRunAttemptParams {
registerCodexTestSessionIdentity(
sessionFile,
AUTH_PROFILE_RUNTIME_CONTRACT.sessionId,
AUTH_PROFILE_RUNTIME_CONTRACT.sessionKey,
);
return {
prompt: AUTH_PROFILE_RUNTIME_CONTRACT.workspacePrompt,
sessionId: AUTH_PROFILE_RUNTIME_CONTRACT.sessionId,
@@ -130,8 +111,7 @@ function createCodexAuthProfileHarness(params: { startMethod: "thread/start" | "
const seenAuthProfileIds: Array<string | undefined> = [];
const seenAgentDirs: Array<string | undefined> = [];
const requests: Array<{ method: string; params: unknown }> = [];
const notificationHandlers = new Set<(notification: unknown) => Promise<void> | void>();
const requestHandlers = new Set<(request: unknown) => unknown>();
let notify: (notification: unknown) => Promise<void> = async () => undefined;
setCodexAppServerClientFactoryForTest(async (_startOptions, authProfileId, agentDir) => {
seenAuthProfileIds.push(authProfileId);
seenAgentDirs.push(agentDir);
@@ -146,22 +126,13 @@ function createCodexAuthProfileHarness(params: { startMethod: "thread/start" | "
}
throw new Error(`unexpected method: ${method}`);
}),
addNotificationHandler: (handler: (notification: unknown) => Promise<void> | void) => {
notificationHandlers.add(handler);
return () => notificationHandlers.delete(handler);
addNotificationHandler: (handler: (notification: unknown) => Promise<void>) => {
notify = handler;
return () => undefined;
},
addRequestHandler: (handler: (request: unknown) => unknown) => {
requestHandlers.add(handler);
return () => requestHandlers.delete(handler);
},
addCloseHandler: () => () => undefined,
addRequestHandler: () => () => undefined,
} as never;
});
const notify = async (notification: unknown) => {
await Promise.all(
[...notificationHandlers].map((handler) => Promise.resolve(handler(notification))),
);
};
return {
seenAuthProfileIds,
seenAgentDirs,
@@ -187,7 +158,6 @@ describe("Auth profile runtime contract - Codex app-server adapter", () => {
let tmpDir: string;
beforeEach(async () => {
resetCodexTestBindingStore();
vi.useRealTimers();
tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-auth-contract-"));
});
@@ -223,7 +193,6 @@ describe("Auth profile runtime contract - Codex app-server adapter", () => {
it("reuses a bound OpenAI Codex auth profile when resume params omit authProfileId", async () => {
const harness = createCodexAuthProfileHarness({ startMethod: "thread/resume" });
const sessionFile = path.join(tmpDir, "session.jsonl");
const params = createParams(sessionFile, tmpDir);
await writeCodexAppServerBinding(sessionFile, {
threadId: "thread-auth-contract",
cwd: tmpDir,
@@ -231,6 +200,7 @@ describe("Auth profile runtime contract - Codex app-server adapter", () => {
dynamicToolsFingerprint: "[]",
});
// authProfileId is intentionally omitted to exercise the resume-bound profile path.
const params = createParams(sessionFile, tmpDir);
const run = runCodexAppServerAttempt(params);
await vi.waitFor(
@@ -248,13 +218,13 @@ describe("Auth profile runtime contract - Codex app-server adapter", () => {
it("prefers an explicit runtime auth profile over a stale persisted binding", async () => {
const harness = createCodexAuthProfileHarness({ startMethod: "thread/resume" });
const sessionFile = path.join(tmpDir, "session.jsonl");
const params = createParams(sessionFile, tmpDir);
await writeCodexAppServerBinding(sessionFile, {
threadId: "thread-auth-contract",
cwd: tmpDir,
authProfileId: "openai:stale",
dynamicToolsFingerprint: "[]",
});
const params = createParams(sessionFile, tmpDir);
params.authProfileId = AUTH_PROFILE_RUNTIME_CONTRACT.openAiCodexProfileId;
const run = runCodexAppServerAttempt(params);

View File

@@ -0,0 +1,67 @@
/**
* Lazy factories for shared and leased Codex app-server clients.
*/
import type { resolveCodexAppServerAuthProfileIdForAgent } from "./auth-bridge.js";
import type { CodexAppServerClient } from "./client.js";
import type { CodexAppServerStartOptions } from "./config.js";
type AuthProfileOrderConfig = Parameters<
typeof resolveCodexAppServerAuthProfileIdForAgent
>[0]["config"];
/** Factory signature used by Codex attempt startup to acquire a client. */
export type CodexAppServerClientFactory = (
startOptions?: CodexAppServerStartOptions,
authProfileId?: string,
agentDir?: string,
config?: AuthProfileOrderConfig,
options?: {
onStartedClient?: (client: CodexAppServerClient) => void;
abandonSignal?: AbortSignal;
},
) => Promise<CodexAppServerClient>;
let sharedClientModulePromise: Promise<typeof import("./shared-client.js")> | null = null;
const loadSharedClientModule = async () => {
sharedClientModulePromise ??= import("./shared-client.js");
return await sharedClientModulePromise;
};
/** Returns the process-shared app-server client for normal attempt reuse. */
export const defaultCodexAppServerClientFactory: CodexAppServerClientFactory = (
startOptions,
authProfileId,
agentDir,
config,
options,
) =>
loadSharedClientModule().then(({ getSharedCodexAppServerClient }) =>
getSharedCodexAppServerClient({
startOptions,
authProfileId,
agentDir,
config,
onStartedClient: options?.onStartedClient,
abandonSignal: options?.abandonSignal,
}),
);
/** Returns a leased shared client so startup can release ownership explicitly. */
export const defaultLeasedCodexAppServerClientFactory: CodexAppServerClientFactory = (
startOptions,
authProfileId,
agentDir,
config,
options,
) =>
loadSharedClientModule().then(({ getLeasedSharedCodexAppServerClient }) =>
getLeasedSharedCodexAppServerClient({
startOptions,
authProfileId,
agentDir,
config,
onStartedClient: options?.onStartedClient,
abandonSignal: options?.abandonSignal,
}),
);

View File

@@ -1,78 +0,0 @@
import { afterEach, describe, expect, it, vi } from "vitest";
import type { CodexAppServerClient } from "./client.js";
import { createClientHarness } from "./test-support.js";
const mocks = vi.hoisted(() => ({
refreshAuth: vi.fn(async () => ({ accessToken: "refreshed", chatgptAccountId: "account" })),
mergeRateLimitUpdate: vi.fn(),
}));
vi.mock("./auth-bridge.js", () => ({
refreshCodexAppServerAuthTokens: mocks.refreshAuth,
}));
vi.mock("./rate-limit-cache.js", () => ({
mergeCodexRateLimitsUpdate: mocks.mergeRateLimitUpdate,
}));
const { ensureCodexAppServerClientRuntime } = await import("./client-runtime.js");
describe("Codex app-server client runtime", () => {
const clients: CodexAppServerClient[] = [];
afterEach(() => {
for (const client of clients) {
client.close();
}
clients.length = 0;
mocks.refreshAuth.mockClear();
mocks.mergeRateLimitUpdate.mockClear();
});
it("installs shared handlers once per physical client", async () => {
const harness = createClientHarness();
clients.push(harness.client);
const context = {
agentDir: "/tmp/agent",
authProfileId: "openai:default",
config: {},
};
const updatedContext = {
...context,
authProfileStore: { version: 1 as const, profiles: {} },
config: { models: { mode: "merge" as const } },
};
const addNotificationHandler = vi.spyOn(harness.client, "addNotificationHandler");
const addRequestHandler = vi.spyOn(harness.client, "addRequestHandler");
const addCloseHandler = vi.spyOn(harness.client, "addCloseHandler");
ensureCodexAppServerClientRuntime(harness.client, context);
ensureCodexAppServerClientRuntime(harness.client, updatedContext);
expect(addNotificationHandler).toHaveBeenCalledTimes(1);
expect(addRequestHandler).toHaveBeenCalledTimes(1);
expect(addCloseHandler).not.toHaveBeenCalled();
harness.send({
method: "account/rateLimits/updated",
params: { rateLimits: { primary: { usedPercent: 12 } } },
});
harness.send({
id: "refresh-1",
method: "account/chatgptAuthTokens/refresh",
params: { reason: "expired" },
});
await vi.waitFor(() => expect(mocks.mergeRateLimitUpdate).toHaveBeenCalledTimes(1));
await vi.waitFor(() => expect(mocks.refreshAuth).toHaveBeenCalledTimes(1));
expect(mocks.refreshAuth).toHaveBeenCalledWith(updatedContext);
expect(mocks.mergeRateLimitUpdate).toHaveBeenCalledWith(harness.client, {
rateLimits: { primary: { usedPercent: 12 } },
});
await vi.waitFor(() =>
expect(harness.writes.map((line) => JSON.parse(line) as unknown)).toContainEqual({
id: "refresh-1",
result: { accessToken: "refreshed", chatgptAccountId: "account" },
}),
);
});
});

View File

@@ -1,50 +0,0 @@
/** Client-scoped Codex auth and account observers. */
import { refreshCodexAppServerAuthTokens } from "./auth-bridge.js";
import type { CodexAppServerClient } from "./client.js";
import type { JsonValue } from "./protocol.js";
import { mergeCodexRateLimitsUpdate } from "./rate-limit-cache.js";
import type { CodexAppServerAuthProfileLookup } from "./session-binding.js";
type ClientRuntimeContext = Omit<CodexAppServerAuthProfileLookup, "agentDir"> & {
agentDir: string;
};
type ClientRuntime = {
context: ClientRuntimeContext;
};
const configuredClients = new WeakMap<CodexAppServerClient, ClientRuntime>();
/** Installs one auth-refresh handler and one rate-limit observer per physical client. */
export function ensureCodexAppServerClientRuntime(
client: CodexAppServerClient,
context: ClientRuntimeContext,
): void {
const existing = configuredClients.get(client);
if (existing) {
// Shared-client keys already isolate agent/auth identity. Keep config fresh
// without installing another physical-client handler set.
existing.context = context;
return;
}
const runtime: ClientRuntime = { context };
configuredClients.set(client, runtime);
client.addRequestHandler(async (request) => {
if (request.method !== "account/chatgptAuthTokens/refresh") {
return undefined;
}
return (await refreshCodexAppServerAuthTokens({
agentDir: runtime.context.agentDir,
authProfileId: runtime.context.authProfileId,
...(runtime.context.authProfileStore
? { authProfileStore: runtime.context.authProfileStore }
: {}),
config: runtime.context.config,
})) as unknown as JsonValue;
});
client.addNotificationHandler((notification) => {
if (notification.method === "account/rateLimits/updated") {
mergeCodexRateLimitsUpdate(client, notification.params);
}
});
}

View File

@@ -50,78 +50,6 @@ describe("CodexAppServerClient", () => {
expect(outbound.method).toBe("model/list");
});
it("keeps a shared thread subscribed until every local owner releases it", async () => {
const harness = createClientHarness();
clients.push(harness.client);
const firstResume = harness.client.request("thread/resume", { threadId: "thread-1" });
const secondResume = harness.client.request("thread/resume", { threadId: "thread-1" });
const [firstRequest, secondRequest] = harness.writes.map((line) => JSON.parse(line)) as Array<{
id: number;
}>;
const resumeResult = {
thread: { id: "thread-1", cwd: "/tmp", status: { type: "idle" } },
model: "gpt-5.5",
};
harness.send({ id: firstRequest?.id, result: resumeResult });
harness.send({ id: secondRequest?.id, result: resumeResult });
await Promise.all([firstResume, secondResume]);
await expect(
harness.client.request("thread/unsubscribe", { threadId: "thread-1" }),
).resolves.toEqual({ status: "unsubscribed" });
expect(harness.writes).toHaveLength(2);
const finalRelease = harness.client.request("thread/unsubscribe", {
threadId: "thread-1",
});
const releaseRequest = JSON.parse(harness.writes[2] ?? "{}") as { id?: number };
harness.send({ id: releaseRequest.id, result: { status: "unsubscribed" } });
await expect(finalRelease).resolves.toEqual({ status: "unsubscribed" });
expect(harness.writes).toHaveLength(3);
});
it("pairs written resume failures without retaining pre-aborted requests", async () => {
const harness = createClientHarness();
clients.push(harness.client);
const firstResume = harness.client.request("thread/resume", { threadId: "thread-1" });
const firstRequest = JSON.parse(harness.writes[0] ?? "{}") as { id?: number };
harness.send({
id: firstRequest.id,
result: {
thread: { id: "thread-1", cwd: "/tmp", status: { type: "idle" } },
model: "gpt-5.5",
},
});
await firstResume;
const failedResume = harness.client.request("thread/resume", { threadId: "thread-1" });
const failedRequest = JSON.parse(harness.writes[1] ?? "{}") as { id?: number };
harness.send({ id: failedRequest.id, error: { code: -32000, message: "resume failed" } });
await expect(failedResume).rejects.toThrow("resume failed");
await expect(
harness.client.request("thread/unsubscribe", { threadId: "thread-1" }),
).resolves.toEqual({ status: "unsubscribed" });
expect(harness.writes).toHaveLength(2);
const controller = new AbortController();
controller.abort();
await expect(
harness.client.request(
"thread/resume",
{ threadId: "thread-1" },
{ signal: controller.signal },
),
).rejects.toThrow("thread/resume aborted");
const unsubscribe = harness.client.request("thread/unsubscribe", { threadId: "thread-1" });
expect(harness.writes).toHaveLength(3);
const unsubscribeRequest = JSON.parse(harness.writes[2] ?? "{}") as { id?: number };
harness.send({ id: unsubscribeRequest.id, result: { status: "unsubscribed" } });
await expect(unsubscribe).resolves.toEqual({ status: "unsubscribed" });
});
it("removes unpaired surrogate code units from outbound JSON-RPC strings", async () => {
const harness = createClientHarness();
clients.push(harness.client);
@@ -142,9 +70,9 @@ describe("CodexAppServerClient", () => {
expect(outbound.params?.nested).toEqual(["lowend", "emoji 🙈 ok"]);
harness.send({
id: JSON.parse(harness.writes[0] ?? "{}").id,
result: { thread: { id: "thread-1" } },
result: { threadId: "thread-1" },
});
await expect(request).resolves.toEqual({ thread: { id: "thread-1" } });
await expect(request).resolves.toEqual({ threadId: "thread-1" });
});
it("logs a redacted preview for malformed app-server messages", async () => {
@@ -212,30 +140,6 @@ describe("CodexAppServerClient", () => {
expect(warn).not.toHaveBeenCalled();
});
it("contains synchronous notification handler failures and continues fanout", async () => {
const warn = vi.spyOn(embeddedAgentLog, "warn").mockImplementation(() => undefined);
const harness = createClientHarness();
clients.push(harness.client);
const laterHandler = vi.fn();
harness.client.addNotificationHandler(() => {
throw new Error("handler exploded");
});
harness.client.addNotificationHandler(laterHandler);
expect(() =>
harness.send({
method: "item/commandExecution/outputDelta",
params: { delta: "still routed" },
}),
).not.toThrow();
await vi.waitFor(() => expect(laterHandler).toHaveBeenCalledTimes(1));
expect(warn).toHaveBeenCalledWith(
"codex app-server notification handler failed",
expect.objectContaining({ error: expect.any(Error) }),
);
});
it("preserves JSON-RPC error codes", async () => {
const harness = createClientHarness();
clients.push(harness.client);
@@ -316,95 +220,6 @@ describe("CodexAppServerClient", () => {
expect(harness.writes).toHaveLength(1);
});
it.each([
{
method: "thread/start" as const,
params: {},
abandonment: "timeout" as const,
expectedError: "thread/start timed out",
},
{
method: "thread/fork" as const,
params: { threadId: "parent-thread" },
abandonment: "abort" as const,
expectedError: "thread/fork aborted",
},
])("unsubscribes a late successful $method after local $abandonment", async (testCase) => {
vi.useFakeTimers();
const harness = createClientHarness();
clients.push(harness.client);
const controller = new AbortController();
const options =
testCase.abandonment === "timeout" ? { timeoutMs: 1 } : { signal: controller.signal };
const request = harness.client.request(testCase.method, testCase.params, options);
const outbound = JSON.parse(harness.writes[0] ?? "{}") as { id?: number };
const rejected = expect(request).rejects.toThrow(testCase.expectedError);
if (testCase.abandonment === "timeout") {
await vi.advanceTimersByTimeAsync(100);
} else {
controller.abort();
}
await rejected;
harness.send({ id: outbound.id, result: { thread: { id: "late-thread" } } });
expect(JSON.parse(harness.writes[1] ?? "{}")).toEqual({
id: expect.any(Number),
method: "thread/unsubscribe",
params: { threadId: "late-thread" },
});
});
it("closes when a late thread creation subscription cannot be released", async () => {
const harness = createClientHarness();
clients.push(harness.client);
const controller = new AbortController();
const request = harness.client.request("thread/start", {}, { signal: controller.signal });
const outbound = JSON.parse(harness.writes[0] ?? "{}") as { id?: number };
const rejected = expect(request).rejects.toThrow("thread/start aborted");
controller.abort();
await rejected;
harness.send({ id: outbound.id, result: { thread: { id: "late-thread" } } });
const unsubscribe = JSON.parse(harness.writes[1] ?? "{}") as { id?: number };
harness.send({
id: unsubscribe.id,
error: { code: -32_000, message: "unsubscribe failed" },
});
await vi.waitFor(() => expect(harness.stdinDestroyed).toBe(true));
});
it("does not unsubscribe a late rejected thread creation", async () => {
const harness = createClientHarness();
clients.push(harness.client);
const controller = new AbortController();
const request = harness.client.request("thread/start", {}, { signal: controller.signal });
const outbound = JSON.parse(harness.writes[0] ?? "{}") as { id?: number };
const rejected = expect(request).rejects.toThrow("thread/start aborted");
controller.abort();
await rejected;
harness.send({ id: outbound.id, error: { code: -32000, message: "start failed" } });
expect(harness.writes).toHaveLength(1);
});
it("closes after the bounded late-creation cleanup ledger fills", async () => {
const harness = createClientHarness();
clients.push(harness.client);
for (let index = 0; index < 129; index += 1) {
const controller = new AbortController();
const request = harness.client.request("thread/start", {}, { signal: controller.signal });
const rejected = expect(request).rejects.toThrow("thread/start aborted");
controller.abort();
await rejected;
}
expect(harness.stdinDestroyed).toBe(true);
});
it("initializes with the required client version", async () => {
const { harness, initializing, outbound } = startInitialize();
harness.send({
@@ -701,26 +516,6 @@ describe("CodexAppServerClient", () => {
});
});
it.each(["execCommandApproval", "applyPatchApproval"])(
"fails closed for unhandled legacy %s requests",
async (method) => {
const harness = createClientHarness();
clients.push(harness.client);
harness.send({
id: "legacy-approval-1",
method,
params: { conversationId: "thread-1" },
});
await vi.waitFor(() => expect(harness.writes.length).toBe(1));
expect(JSON.parse(harness.writes[0] ?? "{}")).toEqual({
id: "legacy-approval-1",
result: { decision: "denied" },
});
},
);
it("fails closed for unhandled native app-server approvals", async () => {
const harness = createClientHarness();
clients.push(harness.client);
@@ -738,41 +533,6 @@ describe("CodexAppServerClient", () => {
});
});
it.each([
[
"item/tool/call",
{
contentItems: [
{
type: "inputText",
text: "OpenClaw did not register a handler for this app-server tool call.",
},
],
success: false,
},
],
["item/permissions/requestApproval", { permissions: {}, scope: "turn" }],
["mcpServer/elicitation/request", { action: "decline" }],
[
"item/future/requestApproval",
{
decision: "decline",
reason: "OpenClaw codex app-server bridge does not grant unknown native approvals.",
},
],
])("fails closed for an unhandled %s request", async (method, expected) => {
const harness = createClientHarness();
clients.push(harness.client);
harness.send({ id: "unhandled-1", method, params: { threadId: "thread-1" } });
await vi.waitFor(() => expect(harness.writes.length).toBe(1));
expect(JSON.parse(harness.writes[0] ?? "{}")).toEqual({
id: "unhandled-1",
result: expected,
});
});
it("only treats known Codex app-server approval methods as approvals", () => {
expect(isCodexAppServerApprovalRequest("item/commandExecution/requestApproval")).toBe(true);
expect(isCodexAppServerApprovalRequest("item/fileChange/requestApproval")).toBe(true);

View File

@@ -12,7 +12,6 @@ import {
type CodexInitializeParams,
type CodexInitializeResponse,
isRpcResponse,
readCodexThreadCreationResponseId,
type CodexServerNotification,
type JsonValue,
type RpcMessage,
@@ -35,8 +34,6 @@ const CODEX_APP_SERVER_PARSE_BUFFER_MAX = 1_000_000;
const CODEX_APP_SERVER_PARSE_BUFFER_MAX_LINES = 1_000;
const CODEX_DYNAMIC_TOOL_SERVER_REQUEST_TIMEOUT_MS = 600_000;
const CODEX_APP_SERVER_STDERR_TAIL_MAX = 2_000;
const CODEX_APP_SERVER_ABANDONED_THREAD_CREATION_MAX = 128;
const CODEX_APP_SERVER_LATE_THREAD_CLEANUP_TIMEOUT_MS = 5_000;
const UNPAIRED_SURROGATE_RE =
/[\uD800-\uDBFF](?![\uDC00-\uDFFF])|(?<![\uD800-\uDBFF])[\uDC00-\uDFFF]/g;
@@ -114,10 +111,7 @@ export class CodexAppServerClient {
private readonly requestHandlers = new Set<CodexServerRequestHandler>();
private readonly notificationHandlers = new Set<CodexServerNotificationHandler>();
private readonly closeHandlers = new Set<(client: CodexAppServerClient) => void>();
private readonly threadSubscriptionOwners = new Map<string, number>();
// Codex may finish a locally abandoned create request. Remember its RPC id
// until response/close so the unknown thread subscription can be released.
private readonly abandonedThreadCreationRequestIds = new Set<number | string>();
private activeSharedLeaseCountProvider: (() => number | undefined) | undefined;
private nextId = 1;
private initialized = false;
private closed = false;
@@ -231,27 +225,11 @@ export class CodexAppServerClient {
if (options.signal?.aborted) {
return Promise.reject(new Error(`${method} aborted`));
}
const requestedThreadId = readRequestThreadId(params);
if (
method === "thread/unsubscribe" &&
requestedThreadId &&
this.releaseThreadSubscriptionOwner(requestedThreadId)
) {
// Codex subscriptions are connection-wide sets. A logical owner can
// release without silencing another turn on the same physical client.
return Promise.resolve({ status: "unsubscribed" } as unknown as T);
}
if (method === "thread/resume" && requestedThreadId) {
// Every resume attempt owns one release, even if the response times out
// or aborts: Codex may have subscribed before OpenClaw saw the outcome.
this.retainThreadSubscriptionOwner(requestedThreadId);
}
const id = this.nextId++;
const message: RpcRequest = { id, method, params: params as JsonValue | undefined };
return new Promise<T>((resolve, reject) => {
let timeout: ReturnType<typeof setTimeout> | undefined;
let cleanupAbort: (() => void) | undefined;
let requestWritten = false;
const cleanup = () => {
if (timeout) {
clearTimeout(timeout);
@@ -260,37 +238,23 @@ export class CodexAppServerClient {
cleanupAbort?.();
cleanupAbort = undefined;
};
const rejectPending = (error: Error, rememberLateThreadCreation = false) => {
const rejectPending = (error: Error) => {
if (!this.pending.has(id)) {
return;
}
this.pending.delete(id);
if (rememberLateThreadCreation && isThreadCreationRequest(method)) {
if (
this.abandonedThreadCreationRequestIds.size >=
CODEX_APP_SERVER_ABANDONED_THREAD_CREATION_MAX
) {
// Lost create responses can hide server subscriptions. Once the
// bounded cleanup ledger fills, closing is the only safe release.
this.closeWithError(
new Error("codex app-server abandoned thread creation limit exceeded"),
);
} else {
this.abandonedThreadCreationRequestIds.add(id);
}
}
cleanup();
reject(error);
};
if (options.timeoutMs && Number.isFinite(options.timeoutMs) && options.timeoutMs > 0) {
timeout = setTimeout(
() => rejectPending(new Error(`${method} timed out`), true),
() => rejectPending(new Error(`${method} timed out`)),
Math.max(100, options.timeoutMs),
);
timeout.unref?.();
}
if (options.signal) {
const abortListener = () => rejectPending(new Error(`${method} aborted`), requestWritten);
const abortListener = () => rejectPending(new Error(`${method} aborted`));
options.signal.addEventListener("abort", abortListener, { once: true });
cleanupAbort = () => options.signal?.removeEventListener("abort", abortListener);
}
@@ -298,12 +262,6 @@ export class CodexAppServerClient {
method,
resolve: (value) => {
cleanup();
if (method === "thread/start" || method === "thread/fork") {
const threadId = readCodexThreadCreationResponseId(value);
if (threadId) {
this.retainThreadSubscriptionOwner(threadId);
}
}
resolve(value as T);
},
reject: (error) => {
@@ -317,7 +275,6 @@ export class CodexAppServerClient {
return;
}
try {
requestWritten = true;
this.writeMessage(message, (error) => rejectPending(error));
} catch (error) {
rejectPending(error instanceof Error ? error : new Error(String(error)));
@@ -342,6 +299,18 @@ export class CodexAppServerClient {
return () => this.notificationHandlers.delete(handler);
}
/** Installs a lease-count provider used to route unscoped notifications. */
setActiveSharedLeaseCountProviderForUnscopedNotifications(
provider: (() => number | undefined) | undefined,
): void {
this.activeSharedLeaseCountProvider = provider;
}
/** Reads the active shared-client lease count when available. */
getActiveSharedLeaseCountForUnscopedNotifications(): number | undefined {
return this.activeSharedLeaseCountProvider?.();
}
/** Registers a close handler and returns its disposer. */
addCloseHandler(handler: (client: CodexAppServerClient) => void): () => void {
this.closeHandlers.add(handler);
@@ -460,15 +429,6 @@ export class CodexAppServerClient {
}
private handleResponse(response: RpcResponse): void {
if (this.abandonedThreadCreationRequestIds.delete(response.id)) {
if (!response.error) {
const threadId = readCodexThreadCreationResponseId(response.result);
if (threadId) {
this.unsubscribeLateThreadCreation(threadId);
}
}
return;
}
const pending = this.pending.get(response.id);
if (!pending) {
return;
@@ -546,14 +506,7 @@ export class CodexAppServerClient {
private handleNotification(notification: CodexServerNotification): void {
for (const handler of this.notificationHandlers) {
let result: Promise<void> | void;
try {
result = handler(notification);
} catch (error) {
embeddedAgentLog.warn("codex app-server notification handler failed", { error });
continue;
}
Promise.resolve(result).catch((error: unknown) => {
Promise.resolve(handler(notification)).catch((error: unknown) => {
embeddedAgentLog.warn("codex app-server notification handler failed", { error });
});
}
@@ -571,54 +524,11 @@ export class CodexAppServerClient {
}
this.closed = true;
this.closeError = error;
this.threadSubscriptionOwners.clear();
this.abandonedThreadCreationRequestIds.clear();
this.lines.close();
this.rejectPendingRequests(error);
return true;
}
private unsubscribeLateThreadCreation(threadId: string): void {
// This late response never registered a local owner. Track the wire
// release anyway; an unconfirmed cleanup makes this client unsafe to pool.
void this.request(
"thread/unsubscribe",
{ threadId },
{ timeoutMs: CODEX_APP_SERVER_LATE_THREAD_CLEANUP_TIMEOUT_MS },
).catch((error: unknown) => {
embeddedAgentLog.debug("codex app-server late thread unsubscribe failed", {
threadId,
error,
});
this.closeWithError(
new Error(`Codex late thread subscription could not be released: ${threadId}`, {
cause: error,
}),
);
});
}
private retainThreadSubscriptionOwner(threadId: string): void {
this.threadSubscriptionOwners.set(
threadId,
(this.threadSubscriptionOwners.get(threadId) ?? 0) + 1,
);
}
/** Returns true when another local owner still needs the wire subscription. */
private releaseThreadSubscriptionOwner(threadId: string): boolean {
const owners = this.threadSubscriptionOwners.get(threadId);
if (owners === undefined) {
return false;
}
if (owners > 1) {
this.threadSubscriptionOwners.set(threadId, owners - 1);
return true;
}
this.threadSubscriptionOwners.delete(threadId);
return false;
}
private rejectPendingRequests(error: Error): void {
for (const pending of this.pending.values()) {
pending.cleanup();
@@ -631,17 +541,6 @@ export class CodexAppServerClient {
}
}
function readRequestThreadId(value: unknown): string | undefined {
if (!isJsonObject(value) || typeof value.threadId !== "string") {
return undefined;
}
return value.threadId.trim() || undefined;
}
function isThreadCreationRequest(method: string): boolean {
return method === "thread/start" || method === "thread/fork";
}
function defaultServerRequestResponse(
request: Required<Pick<RpcRequest, "id" | "method">> & { params?: JsonValue },
): JsonValue {
@@ -656,9 +555,6 @@ function defaultServerRequestResponse(
success: false,
};
}
if (request.method === "execCommandApproval" || request.method === "applyPatchApproval") {
return { decision: "denied" };
}
if (
request.method === "item/commandExecution/requestApproval" ||
request.method === "item/fileChange/requestApproval"
@@ -674,12 +570,6 @@ function defaultServerRequestResponse(
reason: "OpenClaw codex app-server bridge does not grant native approvals yet.",
};
}
if (request.method.includes("requestApproval")) {
return {
decision: "decline",
reason: "OpenClaw codex app-server bridge does not grant unknown native approvals.",
};
}
if (request.method === "item/tool/requestUserInput") {
return {
answers: {},

File diff suppressed because it is too large Load Diff

View File

@@ -7,396 +7,145 @@ import {
type EmbeddedAgentCompactResult,
} from "openclaw/plugin-sdk/agent-harness-runtime";
import {
CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
isCodexAppServerUnsafeSubscriptionError,
settleCodexAppServerClientLease,
} from "./attempt-client-cleanup.js";
import { readCodexNotificationItem } from "./attempt-notifications.js";
import { resolveCodexTurnTerminalIdleTimeoutMs } from "./attempt-timeouts.js";
import { CodexAppServerRpcError } from "./client.js";
defaultLeasedCodexAppServerClientFactory,
type CodexAppServerClientFactory,
} from "./client-factory.js";
import { resolveCodexAppServerRuntimeOptions } from "./config.js";
import { isJsonObject, type JsonObject, type JsonValue } from "./protocol.js";
import type { JsonObject } from "./protocol.js";
import { resolveCodexNativeExecutionBlock } from "./sandbox-guard.js";
import {
CODEX_APP_SERVER_BINDING_GUARDED_REQUEST_TIMEOUT_MS,
sessionBindingIdentity,
type CodexAppServerBindingIdentity,
type CodexAppServerBindingStore,
readCodexAppServerBinding,
withCodexAppServerBindingLock,
writeCodexAppServerBinding,
type CodexAppServerThreadBinding,
} from "./session-binding.js";
import {
leaseSharedCodexAppServerClient,
type CodexAppServerClientLease,
type CodexAppServerClientLeaseFactory,
type CodexAppServerClientOptions,
} from "./shared-client.js";
import { resumeCodexAppServerThread } from "./thread-resume.js";
import { withTimeout } from "./timeout.js";
import {
getCodexAppServerTurnRouter,
isCodexTerminalTurnNotification,
type CodexNativeTurnCompletionWatch,
type CodexThreadRouteReservation,
} from "./turn-router.js";
import { releaseLeasedSharedCodexAppServerClient } from "./shared-client.js";
const warnedIgnoredCompactionOverrides = new Set<string>();
type CodexAppServerCompactOptions = {
bindingStore: CodexAppServerBindingStore;
pluginConfig?: unknown;
clientLeaseFactory?: CodexAppServerClientLeaseFactory;
clientFactory?: CodexAppServerClientFactory;
allowNonManualNativeRequest?: boolean;
};
class CodexNativeTurnBindingChangedError extends Error {}
type CodexNativeTurnRequest = {
bindingStore: CodexAppServerBindingStore;
bindingIdentity: CodexAppServerBindingIdentity;
expectedBinding: CodexAppServerThreadBinding;
pluginConfig?: unknown;
authProfileId?: string;
agentDir?: string;
config?: CodexAppServerClientOptions["config"];
abortSignal?: AbortSignal;
clientLeaseFactory?: CodexAppServerClientLeaseFactory;
};
export type CodexNativeTurnKind = "compact" | "review";
/** Starts one native Codex turn and retains its app-server owner through completion. */
export async function requestCodexNativeTurnForBinding(
params: CodexNativeTurnRequest,
kind: CodexNativeTurnKind,
): Promise<void> {
const isCompaction = kind === "compact";
const label = isCompaction ? "compaction" : "review";
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig: params.pluginConfig });
const requestTimeoutMs = Math.min(
appServer.requestTimeoutMs,
CODEX_APP_SERVER_BINDING_GUARDED_REQUEST_TIMEOUT_MS,
);
await params.bindingStore.withLease(params.bindingIdentity, async () => {
const currentBinding = await params.bindingStore.read(params.bindingIdentity);
if (!currentBinding || !isSameNativeTurnBinding(currentBinding, params.expectedBinding)) {
throw new CodexNativeTurnBindingChangedError(
`Codex thread binding changed before native ${label}`,
);
}
const clientLease = await (params.clientLeaseFactory ?? leaseSharedCodexAppServerClient)({
startOptions: appServer.start,
authProfileId: params.authProfileId ?? currentBinding.authProfileId,
agentDir: params.agentDir,
config: params.config,
abandonSignal: params.abortSignal,
timeoutMs: appServer.requestTimeoutMs,
});
const client = clientLease.client;
let subscribedThreadId: string | undefined;
let abandonClient = false;
let lifecycleTransferred = false;
let awaitingNativeTurnStart = false;
const terminalTurnsBeforeWatch = new Set<string>();
let route: CodexThreadRouteReservation | undefined;
let completionWatch: CodexNativeTurnCompletionWatch | undefined;
let observedContextCompaction = false;
let bindingInvalidated = false;
let resolveNativeTurnStarted!: () => void;
const nativeTurnStarted = new Promise<void>((resolve) => {
resolveNativeTurnStarted = resolve;
});
try {
const router = getCodexAppServerTurnRouter(client);
route = router.reserveThread({
threadId: currentBinding.threadId,
onNotificationReceived: (notification, scope) => {
const contextCompactionStarted =
isCompaction &&
Boolean(scope.turnId) &&
notification.method === "item/started" &&
readCodexNotificationItem(notification.params)?.type === "contextCompaction";
if (contextCompactionStarted) {
observedContextCompaction = true;
}
if (!awaitingNativeTurnStart || !scope.turnId) {
return;
}
if (isCodexTerminalTurnNotification(notification)) {
terminalTurnsBeforeWatch.add(scope.turnId);
}
if (contextCompactionStarted) {
completionWatch ??= router.watchNativeTurnCompletion({
threadId: currentBinding.threadId,
turnId: scope.turnId,
timeoutMs: resolveCodexTurnTerminalIdleTimeoutMs(undefined),
});
resolveNativeTurnStarted();
}
},
onNotification: () => undefined,
});
throwIfCodexNativeTurnAborted(params.abortSignal, kind);
let resumed;
try {
subscribedThreadId = currentBinding.threadId;
resumed = await resumeCodexAppServerThread({
client,
abandonClient: clientLease.abandon,
request: {
threadId: currentBinding.threadId,
excludeTurns: true,
persistExtendedHistory: true,
},
timeoutMs: requestTimeoutMs,
signal: params.abortSignal,
});
} catch (error) {
abandonClient = isCodexAppServerUnsafeSubscriptionError(error);
throw error;
}
const invalidateNativeContextBinding = async () => {
if (bindingInvalidated) {
return;
}
const invalidated = await params.bindingStore.mutate(params.bindingIdentity, {
kind: "invalidate-native-context",
threadId: currentBinding.threadId,
...(isCompaction ? { invalidateContextEngineProjection: true as const } : {}),
});
if (!invalidated) {
throw new CodexNativeTurnBindingChangedError(
`Codex thread binding changed before native ${label}`,
);
}
bindingInvalidated = true;
};
if (isCompaction && observedContextCompaction) {
await invalidateNativeContextBinding();
}
if (resumed.thread.status?.type === "active") {
throw new Error(
`Codex thread already has an active turn; retry ${label} after it finishes`,
);
}
throwIfCodexNativeTurnAborted(params.abortSignal, kind);
await invalidateNativeContextBinding();
awaitingNativeTurnStart = true;
let requestResult: JsonValue | undefined;
try {
requestResult = await client.request(
isCompaction ? "thread/compact/start" : "review/start",
isCompaction
? { threadId: currentBinding.threadId }
: { threadId: currentBinding.threadId, target: { type: "uncommittedChanges" } },
{ timeoutMs: requestTimeoutMs },
);
} catch (error) {
const requestRejected = error instanceof CodexAppServerRpcError;
if (requestRejected) {
// A structured rejection proves this request did not start a native
// turn. Preserve only compaction already observed on the same thread.
completionWatch?.cancel();
completionWatch = undefined;
if (!isCompaction || !observedContextCompaction) {
const restored = await params.bindingStore.mutate(params.bindingIdentity, {
kind: "set",
binding: currentBinding,
});
if (!restored) {
throw new Error(`Codex thread binding changed after native ${label} was rejected`, {
cause: error,
});
}
}
throw error;
}
if (completionWatch) {
embeddedAgentLog.debug(`codex app-server ${kind} request failed after startup`, {
threadId: currentBinding.threadId,
error,
});
} else {
abandonClient = true;
throw error;
}
}
if (!isCompaction) {
try {
const review = assertCodexReviewStartResponse(requestResult);
if (review.reviewThreadId !== currentBinding.threadId) {
throw new Error(
`Codex review/start returned ${review.reviewThreadId} for inline review on ${currentBinding.threadId}`,
);
}
completionWatch = terminalTurnsBeforeWatch.has(review.turnId)
? { completion: Promise.resolve(true), cancel: () => undefined }
: router.watchNativeTurnCompletion({
threadId: currentBinding.threadId,
turnId: review.turnId,
timeoutMs: resolveCodexTurnTerminalIdleTimeoutMs(undefined),
});
} catch (error) {
abandonClient = true;
throw error;
}
} else if (!completionWatch) {
try {
await waitForCodexNativeTurnStart({
started: nativeTurnStarted,
routeSignal: route.signal,
timeoutMs: requestTimeoutMs,
threadId: currentBinding.threadId,
kind,
});
} catch (error) {
// Codex accepted Op::Compact, so missing startup confirmation is
// ambiguous. Keep facts invalidated and retire this connection.
abandonClient = true;
throw error;
}
}
awaitingNativeTurnStart = false;
route.release();
route = undefined;
const transferredWatch = completionWatch;
if (!transferredWatch) {
abandonClient = true;
throw new Error(
`codex app-server ${kind} turn started without a turn id for thread ${currentBinding.threadId}`,
);
}
completionWatch = undefined;
lifecycleTransferred = true;
monitorCodexNativeTurn({
completionWatch: transferredWatch,
clientLease,
subscribedThreadId,
threadId: currentBinding.threadId,
kind,
});
} finally {
if (!lifecycleTransferred) {
completionWatch?.cancel();
route?.release();
await settleCodexAppServerClientLease(clientLease, {
threadId: subscribedThreadId,
timeoutMs: CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
abandon: abandonClient,
});
}
}
});
}
function assertCodexReviewStartResponse(value: JsonValue | undefined): {
turnId: string;
reviewThreadId: string;
} {
if (
!isJsonObject(value) ||
!isJsonObject(value.turn) ||
typeof value.turn.id !== "string" ||
!value.turn.id.trim() ||
typeof value.reviewThreadId !== "string" ||
!value.reviewThreadId.trim()
) {
throw new Error("invalid Codex review/start response");
}
return { turnId: value.turn.id, reviewThreadId: value.reviewThreadId };
}
function monitorCodexNativeTurn(params: {
completionWatch: CodexNativeTurnCompletionWatch;
clientLease: CodexAppServerClientLease;
subscribedThreadId?: string;
threadId: string;
kind: CodexNativeTurnKind;
}): void {
void (async () => {
const completed = await params.completionWatch.completion;
await settleCodexAppServerClientLease(params.clientLease, {
threadId: params.subscribedThreadId,
timeoutMs: CODEX_APP_SERVER_UNSUBSCRIBE_TIMEOUT_MS,
abandon: !completed,
});
if (!completed) {
embeddedAgentLog.warn(`codex app-server ${params.kind} turn lost terminal confirmation`, {
threadId: params.threadId,
});
}
})().catch(async (error: unknown) => {
await params.clientLease.abandon().catch(() => undefined);
embeddedAgentLog.warn(`codex app-server ${params.kind} turn cleanup failed`, {
threadId: params.threadId,
error,
});
});
}
function throwIfCodexNativeTurnAborted(
signal: AbortSignal | undefined,
kind: CodexNativeTurnKind,
): void {
if (!signal?.aborted) {
return;
}
if (signal.reason instanceof Error) {
throw signal.reason;
}
throw new Error(`codex app-server ${kind} aborted before native turn startup`, {
cause: signal.reason,
});
}
async function waitForCodexNativeTurnStart(params: {
started: Promise<void>;
routeSignal: AbortSignal;
timeoutMs: number;
threadId: string;
kind: CodexNativeTurnKind;
}): Promise<void> {
const signal = params.routeSignal;
let removeAbort: (() => void) | undefined;
const aborted = new Promise<never>((_resolve, reject) => {
const onAbort = () => reject(asNativeTurnAbortError(signal));
signal.addEventListener("abort", onAbort, { once: true });
removeAbort = () => signal.removeEventListener("abort", onAbort);
if (signal.aborted) {
onAbort();
}
});
try {
await withTimeout(
Promise.race([params.started, aborted]),
params.timeoutMs,
`codex app-server ${params.kind} turn did not start for thread ${params.threadId}`,
);
} finally {
removeAbort?.();
}
}
function asNativeTurnAbortError(signal: AbortSignal): Error {
return signal.reason instanceof Error
? signal.reason
: new Error("codex app-server native turn startup aborted", { cause: signal.reason });
}
/**
* Starts native Codex compaction for a manually requested bound session, or
* reports why Codex-owned automatic compaction should handle the trigger.
*/
export async function maybeCompactCodexAppServerSession(
params: CompactEmbeddedAgentSessionParams,
options: CodexAppServerCompactOptions,
options: CodexAppServerCompactOptions = {},
): Promise<EmbeddedAgentCompactResult | undefined> {
warnIfIgnoringOpenClawCompactionOverrides(params);
// Codex owns automatic context-pressure compaction for Codex runtime sessions.
// This entry point starts native Codex compaction for the bound thread and
// returns immediately; Codex applies the compaction inside its app-server.
return compactCodexNativeThread(params, options);
}
function warnIfIgnoringOpenClawCompactionOverrides(
params: CompactEmbeddedAgentSessionParams,
): void {
const ignoredConfig = readIgnoredCompactionOverridePaths(params);
if (ignoredConfig.length === 0) {
return;
}
const warningKey = ignoredConfig.join("\0");
if (warnedIgnoredCompactionOverrides.has(warningKey)) {
return;
}
warnedIgnoredCompactionOverrides.add(warningKey);
embeddedAgentLog.warn(
"ignoring OpenClaw compaction overrides for Codex app-server compaction; Codex uses native server-side compaction",
{
sessionId: params.sessionId,
sessionKey: params.sessionKey,
ignoredConfig,
},
);
}
function readIgnoredCompactionOverridePaths(params: CompactEmbeddedAgentSessionParams): string[] {
const ignored = new Set<string>();
for (const entry of readCompactionOverrideEntries(params)) {
const localProvider =
typeof entry.record.provider === "string" ? entry.record.provider.trim() : "";
const inheritedProvider =
!localProvider && typeof entry.inheritedRecord?.provider === "string"
? entry.inheritedRecord.provider.trim()
: "";
const providerPath = localProvider
? `${entry.path}.compaction.provider`
: inheritedProvider && entry.inheritedPath
? `${entry.inheritedPath}.compaction.provider`
: undefined;
if (typeof entry.record.model === "string" && entry.record.model.trim()) {
ignored.add(`${entry.path}.compaction.model`);
}
if (providerPath) {
ignored.add(providerPath);
}
}
return [...ignored];
}
function readCompactionOverrideEntries(params: CompactEmbeddedAgentSessionParams): Array<{
path: string;
record: Record<string, unknown>;
inheritedRecord?: Record<string, unknown>;
inheritedPath?: string;
}> {
const entries: Array<{
path: string;
record: Record<string, unknown>;
inheritedRecord?: Record<string, unknown>;
inheritedPath?: string;
}> = [];
const defaultCompaction = readRecord(readRecord(params.config?.agents)?.defaults)?.compaction;
const defaultRecord = readRecord(defaultCompaction);
if (defaultRecord) {
entries.push({ path: "agents.defaults", record: defaultRecord });
}
const agentId = readAgentIdFromSessionKey(params.sessionKey ?? params.sandboxSessionKey);
if (!agentId) {
return entries;
}
const agents = Array.isArray(params.config?.agents?.list) ? params.config.agents.list : [];
const activeAgent = agents.find((agent) => {
const id = typeof agent?.id === "string" ? agent.id.trim().toLowerCase() : "";
return id === agentId;
});
const agentCompaction = readRecord(activeAgent)?.compaction;
const agentRecord = readRecord(agentCompaction);
if (agentRecord) {
entries.push({
path: `agents.list.${agentId}`,
record: agentRecord,
inheritedRecord: defaultRecord,
inheritedPath: "agents.defaults",
});
}
return entries;
}
function readAgentIdFromSessionKey(sessionKey: string | undefined): string | undefined {
const parts = sessionKey?.trim().toLowerCase().split(":").filter(Boolean) ?? [];
if (parts.length < 3 || parts[0] !== "agent") {
return undefined;
}
return parts[1]?.trim() || undefined;
}
function readRecord(value: unknown): Record<string, unknown> | undefined {
return value && typeof value === "object" && !Array.isArray(value)
? (value as Record<string, unknown>)
: undefined;
}
async function compactCodexNativeThread(
params: CompactEmbeddedAgentSessionParams,
options: CodexAppServerCompactOptions,
options: CodexAppServerCompactOptions = {},
): Promise<EmbeddedAgentCompactResult | undefined> {
if (params.trigger !== "manual" && !options.allowNonManualNativeRequest) {
embeddedAgentLog.info("skipping codex app-server compaction for non-manual trigger", {
@@ -423,7 +172,6 @@ async function compactCodexNativeThread(
}
const nativeExecutionBlock = resolveCodexNativeExecutionBlock({
config: params.config,
agentId: params.agentId,
sessionKey: params.sandboxSessionKey ?? params.sessionKey,
sessionId: params.sessionId,
surface: "native compaction",
@@ -431,20 +179,17 @@ async function compactCodexNativeThread(
if (nativeExecutionBlock) {
return { ok: false, compacted: false, reason: nativeExecutionBlock };
}
const bindingIdentity: CodexAppServerBindingIdentity = sessionBindingIdentity({
sessionId: params.sessionId,
sessionKey: params.sessionKey,
agentId: params.agentId,
const appServer = resolveCodexAppServerRuntimeOptions({ pluginConfig: options.pluginConfig });
const initialBinding = await readCodexAppServerBinding(params.sessionFile, {
config: params.config,
});
const initialBinding = await options.bindingStore.read(bindingIdentity);
if (!initialBinding?.threadId) {
return failedCodexThreadBindingCompactionResult(params, {
reason: "no codex app-server thread binding",
recovery: "missing_thread_binding",
});
}
const binding = initialBinding;
let binding = initialBinding;
const requestedAuthProfileId = params.authProfileId?.trim() || undefined;
if (
requestedAuthProfileId &&
@@ -455,42 +200,85 @@ async function compactCodexNativeThread(
// with another profile risks operating on a different Codex account.
return { ok: false, compacted: false, reason: "auth profile mismatch for session binding" };
}
if (options.allowNonManualNativeRequest && params.abortSignal?.aborted) {
const currentBinding = await options.bindingStore.read(bindingIdentity);
return skippedCodexNativeCompactionResult(params, {
reason: "codex app-server compaction aborted before native compaction",
code: "aborted_before_native_compaction",
expectedThreadId: binding.threadId,
currentThreadId: currentBinding?.threadId,
});
}
const shouldReleaseDefaultLease = !options.clientFactory;
const clientFactory = options.clientFactory ?? defaultLeasedCodexAppServerClientFactory;
const client = await clientFactory(
appServer.start,
requestedAuthProfileId ?? binding.authProfileId,
params.agentDir,
params.config,
);
try {
await requestCodexNativeTurnForBinding(
{
bindingIdentity,
bindingStore: options.bindingStore,
expectedBinding: binding,
pluginConfig: options.pluginConfig,
authProfileId: requestedAuthProfileId,
agentDir: params.agentDir,
config: params.config,
abortSignal: params.abortSignal,
clientLeaseFactory: options.clientLeaseFactory,
},
"compact",
);
if (options.allowNonManualNativeRequest) {
const guardedResult = await withCodexAppServerBindingLock(params.sessionFile, async () => {
const currentBinding = await readCodexAppServerBinding(params.sessionFile, {
config: params.config,
});
if (params.abortSignal?.aborted) {
return {
started: false as const,
result: skippedCodexNativeCompactionResult(params, {
reason: "codex app-server compaction aborted before native compaction",
code: "aborted_before_native_compaction",
expectedThreadId: binding.threadId,
currentThreadId: currentBinding?.threadId,
}),
};
}
if (!currentBinding || !isSameNativeCompactionBinding(currentBinding, binding)) {
embeddedAgentLog.warn(
"skipping codex app-server compaction because the thread binding changed",
{
sessionId: params.sessionId,
sessionKey: params.sessionKey,
expectedThreadId: binding.threadId,
currentThreadId: currentBinding?.threadId,
},
);
return {
started: false as const,
result: skippedCodexNativeCompactionResult(params, {
reason: "codex app-server binding changed before native compaction",
code: "binding_changed_before_native_compaction",
expectedThreadId: binding.threadId,
currentThreadId: currentBinding?.threadId,
}),
};
}
binding = currentBinding;
await clearContextEngineProjectionBeforeNativeCompaction({
sessionId: params.sessionId,
sessionFile: params.sessionFile,
binding,
config: params.config,
});
await client.request(
"thread/compact/start",
{
threadId: binding.threadId,
},
{
timeoutMs: Math.min(
appServer.requestTimeoutMs,
CODEX_APP_SERVER_BINDING_GUARDED_REQUEST_TIMEOUT_MS,
),
},
);
return { started: true as const };
});
if (!guardedResult.started) {
return guardedResult.result;
}
} else {
await client.request("thread/compact/start", {
threadId: binding.threadId,
});
}
embeddedAgentLog.info("started codex app-server compaction", {
sessionId: params.sessionId,
threadId: binding.threadId,
});
} catch (error) {
if (
options.allowNonManualNativeRequest &&
error instanceof CodexNativeTurnBindingChangedError
) {
const latestBinding = await options.bindingStore.read(bindingIdentity);
return skippedBindingChangeResult(params, binding.threadId, latestBinding?.threadId);
}
if (isCodexThreadNotFoundError(error)) {
return failedCodexThreadBindingCompactionResult(params, {
threadId: binding.threadId,
@@ -509,6 +297,10 @@ async function compactCodexNativeThread(
compacted: false,
reason: formatCompactionError(error),
};
} finally {
if (shouldReleaseDefaultLease) {
releaseLeasedSharedCodexAppServerClient(client);
}
}
const resultDetails: JsonObject = {
backend: "codex-app-server",
@@ -534,25 +326,6 @@ async function compactCodexNativeThread(
};
}
function skippedBindingChangeResult(
params: CompactEmbeddedAgentSessionParams,
expectedThreadId: string,
currentThreadId: string | undefined,
): EmbeddedAgentCompactResult {
embeddedAgentLog.warn("skipping codex app-server compaction because the thread binding changed", {
sessionId: params.sessionId,
sessionKey: params.sessionKey,
expectedThreadId,
currentThreadId,
});
return skippedCodexNativeCompactionResult(params, {
reason: "codex app-server binding changed before native compaction",
code: "binding_changed_before_native_compaction",
expectedThreadId,
currentThreadId,
});
}
function skippedCodexNativeCompactionResult(
params: CompactEmbeddedAgentSessionParams,
skipped: {
@@ -609,7 +382,39 @@ function failedCodexThreadBindingCompactionResult(
};
}
function isSameNativeTurnBinding(
async function clearContextEngineProjectionBeforeNativeCompaction(params: {
sessionId: string;
sessionFile: string;
binding: CodexAppServerThreadBinding;
config: CompactEmbeddedAgentSessionParams["config"];
}): Promise<void> {
const contextEngineBinding = params.binding.contextEngine;
if (!contextEngineBinding?.projection) {
return;
}
// Native Codex compaction mutates the thread history outside the projection
// guard. Clear only the projection marker so the next turn reprojects context.
await writeCodexAppServerBinding(
params.sessionFile,
{
...params.binding,
contextEngine: {
...contextEngineBinding,
projection: undefined,
},
createdAt: params.binding.createdAt,
},
{ config: params.config },
);
embeddedAgentLog.info("cleared codex context-engine projection before native compaction", {
sessionId: params.sessionId,
threadId: params.binding.threadId,
previousEpoch: contextEngineBinding.projection.epoch,
previousFingerprint: contextEngineBinding.projection.fingerprint,
});
}
function isSameNativeCompactionBinding(
current: CodexAppServerThreadBinding,
expected: CodexAppServerThreadBinding,
): boolean {

View File

@@ -1,7 +1,5 @@
// Codex tests cover config plugin behavior.
import fs from "node:fs/promises";
import os from "node:os";
import path from "node:path";
import { MAX_TIMER_TIMEOUT_MS } from "openclaw/plugin-sdk/number-runtime";
import { describe, expect, it, vi } from "vitest";
import {
@@ -308,6 +306,7 @@ describe("Codex app-server config", () => {
const switchedLocalModel = resolveCodexModelBackedReviewerPolicyContext({
model: "lmstudio/local-model",
bindingModel: "gpt-5.5",
nativeAuthProfile: true,
});
expect(switchedLocalModel).toEqual({
modelProvider: "lmstudio",
@@ -494,39 +493,6 @@ describe("Codex app-server config", () => {
});
});
it("reloads Codex config.toml policy when Codex can reload it", async () => {
const agentDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-config-"));
const codexHome = path.join(agentDir, "codex-home");
const configPath = path.join(codexHome, "config.toml");
await fs.mkdir(codexHome);
try {
await fs.writeFile(configPath, 'openai_base_url = "http://localhost:8080/v1"\n');
const context = { modelProvider: "openai", model: "gpt-5.5", agentDir };
expect(canUseCodexModelBackedApprovalsReviewerForModel(context)).toBe(false);
await fs.writeFile(configPath, 'openai_base_url = "https://api.openai.com/v1"\n');
expect(canUseCodexModelBackedApprovalsReviewerForModel(context)).toBe(true);
} finally {
await fs.rm(agentDir, { recursive: true, force: true });
}
});
it("observes a Codex config.toml created after the first policy check", async () => {
const agentDir = await fs.mkdtemp(path.join(os.tmpdir(), "openclaw-codex-config-"));
const codexHome = path.join(agentDir, "codex-home");
const configPath = path.join(codexHome, "config.toml");
await fs.mkdir(codexHome);
try {
const context = { modelProvider: "openai", model: "gpt-5.5", agentDir };
expect(canUseCodexModelBackedApprovalsReviewerForModel(context)).toBe(true);
await fs.writeFile(configPath, 'openai_base_url = "http://localhost:8080/v1"\n');
expect(canUseCodexModelBackedApprovalsReviewerForModel(context)).toBe(false);
} finally {
await fs.rm(agentDir, { recursive: true, force: true });
}
});
it("forces prompting when explicit no-prompt config cannot use model-backed review", () => {
const runtime = resolveRuntimeForTest({
pluginConfig: {
@@ -724,8 +690,8 @@ allowed_sandbox_modes = ["read-only", "workspace-write"]
env: {},
modelProvider: "openai",
requirementsPath: "/custom/codex/requirements.toml",
readRequirementsFile: (requirementsPath) => {
readPaths.push(requirementsPath);
readRequirementsFile: (path) => {
readPaths.push(path);
return 'allowed_sandbox_modes = ["read-only", "workspace-write"]\n';
},
});
@@ -745,8 +711,8 @@ allowed_sandbox_modes = ["read-only", "workspace-write"]
env: { ProgramData: "D:\\ManagedData" },
modelProvider: "openai",
platform: "win32",
readRequirementsFile: (requirementsPath) => {
readPaths.push(requirementsPath);
readRequirementsFile: (path) => {
readPaths.push(path);
return 'allowed_sandbox_modes = ["read-only", "workspace-write"]\n';
},
});

View File

@@ -150,11 +150,6 @@ export type CodexAppServerRuntimeOptions = {
serviceTier?: CodexServiceTier;
};
export type CodexAppServerRuntimeResolution = {
appServer: CodexAppServerRuntimeOptions;
modelBackedReviewerAvailable: boolean;
};
export type CodexModelBackedReviewerContext = {
modelProvider?: string;
model?: string;
@@ -410,34 +405,25 @@ export function resolveCodexPluginsPolicy(pluginConfig?: unknown): ResolvedCodex
};
}
type CodexAppServerRuntimeParams = {
pluginConfig?: unknown;
execMode?: OpenClawExecMode;
execPolicy?: OpenClawExecPolicyForCodexAppServer;
modelProvider?: string;
model?: string;
config?: ProviderAuthAliasConfig;
env?: NodeJS.ProcessEnv;
agentDir?: string;
codexConfigToml?: string | null;
requirementsToml?: string | null;
requirementsPath?: string;
readRequirementsFile?: (path: string) => string | undefined;
platform?: NodeJS.Platform;
hostName?: string;
openClawSandboxActive?: boolean;
};
export function resolveCodexAppServerRuntimeOptions(
params: CodexAppServerRuntimeParams = {},
params: {
pluginConfig?: unknown;
execMode?: OpenClawExecMode;
execPolicy?: OpenClawExecPolicyForCodexAppServer;
modelProvider?: string;
model?: string;
config?: ProviderAuthAliasConfig;
env?: NodeJS.ProcessEnv;
agentDir?: string;
codexConfigToml?: string | null;
requirementsToml?: string | null;
requirementsPath?: string;
readRequirementsFile?: (path: string) => string | undefined;
platform?: NodeJS.Platform;
hostName?: string;
openClawSandboxActive?: boolean;
} = {},
): CodexAppServerRuntimeOptions {
return resolveCodexAppServerRuntime(params).appServer;
}
/** Resolves runtime options and the model-policy fact computed with them. */
export function resolveCodexAppServerRuntime(
params: CodexAppServerRuntimeParams = {},
): CodexAppServerRuntimeResolution {
const env = params.env ?? process.env;
const config = readCodexPluginConfig(params.pluginConfig).appServer ?? {};
const transport = resolveTransport(config.transport);
@@ -561,46 +547,43 @@ export function resolveCodexAppServerRuntime(
: "implicit";
return {
modelBackedReviewerAvailable: canUseModelBackedReviewer,
appServer: {
start: {
transport,
command,
commandSource,
args: args.length > 0 ? args : ["app-server", "--listen", "stdio://"],
...(url ? { url } : {}),
...(authToken ? { authToken } : {}),
headers,
...(transport === "stdio" && clearEnv.length > 0 ? { clearEnv } : {}),
},
codeModeOnly: config.codeModeOnly === true,
requestTimeoutMs: normalizePositiveNumber(config.requestTimeoutMs, 60_000),
turnCompletionIdleTimeoutMs: normalizePositiveNumber(
config.turnCompletionIdleTimeoutMs,
60_000,
),
...(config.postToolRawAssistantCompletionIdleTimeoutMs !== undefined
? {
postToolRawAssistantCompletionIdleTimeoutMs: normalizePositiveNumber(
config.postToolRawAssistantCompletionIdleTimeoutMs,
60_000,
),
}
: {}),
approvalPolicy: forcedPolicy?.approvalPolicy ?? approvalPolicy,
approvalPolicySource,
sandbox:
forcedPolicy?.sandbox ??
configuredSandbox ??
defaultPolicy?.sandbox ??
(policyMode === "guardian" ? "workspace-write" : "danger-full-access"),
approvalsReviewer:
forcedPolicy?.approvalsReviewer ??
explicitApprovalsReviewer ??
defaultPolicy?.approvalsReviewer ??
(policyMode === "guardian" ? "auto_review" : "user"),
...(serviceTier ? { serviceTier } : {}),
start: {
transport,
command,
commandSource,
args: args.length > 0 ? args : ["app-server", "--listen", "stdio://"],
...(url ? { url } : {}),
...(authToken ? { authToken } : {}),
headers,
...(transport === "stdio" && clearEnv.length > 0 ? { clearEnv } : {}),
},
codeModeOnly: config.codeModeOnly === true,
requestTimeoutMs: normalizePositiveNumber(config.requestTimeoutMs, 60_000),
turnCompletionIdleTimeoutMs: normalizePositiveNumber(
config.turnCompletionIdleTimeoutMs,
60_000,
),
...(config.postToolRawAssistantCompletionIdleTimeoutMs !== undefined
? {
postToolRawAssistantCompletionIdleTimeoutMs: normalizePositiveNumber(
config.postToolRawAssistantCompletionIdleTimeoutMs,
60_000,
),
}
: {}),
approvalPolicy: forcedPolicy?.approvalPolicy ?? approvalPolicy,
approvalPolicySource,
sandbox:
forcedPolicy?.sandbox ??
configuredSandbox ??
defaultPolicy?.sandbox ??
(policyMode === "guardian" ? "workspace-write" : "danger-full-access"),
approvalsReviewer:
forcedPolicy?.approvalsReviewer ??
explicitApprovalsReviewer ??
defaultPolicy?.approvalsReviewer ??
(policyMode === "guardian" ? "auto_review" : "user"),
...(serviceTier ? { serviceTier } : {}),
};
}
@@ -672,6 +655,7 @@ export function resolveCodexModelBackedReviewerPolicyContext(params: {
model?: string;
bindingModelProvider?: string;
bindingModel?: string;
nativeAuthProfile?: boolean;
}): CodexModelBackedReviewerContext {
const provider = params.provider?.trim();
if (provider && provider.toLowerCase() !== "codex") {
@@ -703,7 +687,7 @@ export function resolveCodexModelBackedReviewerPolicyContext(params: {
};
}
return {
modelProvider: undefined,
modelProvider: params.nativeAuthProfile === true ? "openai" : undefined,
model: params.model ?? params.bindingModel,
};
}
@@ -770,7 +754,6 @@ export function codexAppServerStartOptionsKey(
options: CodexAppServerStartOptions,
params: {
authProfileId?: string;
authAccountCacheKey?: string;
agentDir?: string;
fallbackApiKeyCacheKey?: string;
} = {},
@@ -790,7 +773,6 @@ export function codexAppServerStartOptionsKey(
.map(([key, value]) => [key, hashSecretForKey(value, `env:${key}`)]),
clearEnv: [...(options.clearEnv ?? [])].toSorted(),
authProfileId: params.authProfileId ?? null,
authAccountCacheKey: params.authAccountCacheKey ?? null,
agentDir: params.agentDir ?? null,
fallbackApiKeyCacheKey: params.fallbackApiKeyCacheKey ?? null,
});

View File

@@ -9,17 +9,15 @@ import {
import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
import {
addSandboxShellDynamicToolsIfAvailable,
buildDynamicTools,
filterCodexDynamicToolsForAllowlist,
hasWildcardCodexToolsAllow,
includeForcedCodexDynamicToolAllow,
prepareDynamicToolCatalog,
resetOpenClawCodingToolsFactoryForTests,
resolveOpenClawCodingToolsSessionKeys,
resolveCodexMessageToolProvider,
setOpenClawCodingToolsFactoryForTests,
shouldEnableCodexAppServerNativeToolSurface,
shouldForceMessageTool,
type OpenClawCodingToolsFactory,
} from "./dynamic-tool-build.js";
import {
filterCodexDynamicTools,
@@ -101,13 +99,13 @@ function createRuntimeDynamicTool(name: string): RuntimeDynamicToolForTest {
async function buildDynamicToolsForTest(
params: EmbeddedRunAttemptParams,
workspaceDir: string,
options: Partial<Parameters<typeof prepareDynamicToolCatalog>[0]> = {},
options: Partial<Parameters<typeof buildDynamicTools>[0]> = {},
) {
const sandboxSessionKey = params.sessionKey;
if (!sandboxSessionKey) {
throw new Error("createParams must provide a sessionKey for Codex dynamic tool tests.");
}
const catalog = await prepareDynamicToolCatalog({
return buildDynamicTools({
params,
resolvedWorkspace: workspaceDir,
effectiveWorkspace: workspaceDir,
@@ -120,7 +118,6 @@ async function buildDynamicToolsForTest(
onYieldDetected: () => undefined,
...options,
});
return catalog.tools;
}
describe("Codex app-server dynamic tool build", () => {
@@ -135,15 +132,6 @@ describe("Codex app-server dynamic tool build", () => {
await fs.rm(tempDir, { recursive: true, force: true });
});
it("uses the message tool channel before a differing ingress provider", () => {
expect(
resolveCodexMessageToolProvider({
messageChannel: "discord",
messageProvider: "discord-voice",
}),
).toBe("discord");
});
it("filters Codex-native dynamic tools from app-server tool exposure", () => {
const tools = [
"read",
@@ -171,53 +159,6 @@ describe("Codex app-server dynamic tool build", () => {
]);
});
it("prepares runtime and durable tool views from one OpenClaw catalog", async () => {
const messageTool = createRuntimeDynamicTool("message");
const webSearchTool = createRuntimeDynamicTool("web_search");
const heartbeatTool = createRuntimeDynamicTool("heartbeat_respond");
const factory = vi.fn<OpenClawCodingToolsFactory>((options) => [
messageTool,
webSearchTool,
...(options?.enableHeartbeatTool ? [heartbeatTool] : []),
]);
setOpenClawCodingToolsFactoryForTests(factory);
const sessionFile = path.join(tempDir, "session.jsonl");
const workspaceDir = path.join(tempDir, "workspace");
const params = createParams(sessionFile, workspaceDir);
params.disableTools = false;
const runtimePlan = createCodexRuntimePlanFixture();
params.runtimePlan = {
...runtimePlan,
tools: {
normalize: (tools: Array<{ name: string }>) =>
tools.filter((tool) => tool.name === "message"),
logDiagnostics: () => undefined,
},
} as unknown as NonNullable<EmbeddedRunAttemptParams["runtimePlan"]>;
const catalog = await prepareDynamicToolCatalog({
params,
resolvedWorkspace: workspaceDir,
effectiveWorkspace: workspaceDir,
sandboxSessionKey: params.sessionKey ?? "agent:main:session-1",
sandbox: { enabled: false, backendId: "docker" } as never,
nativeToolSurfaceEnabled: true,
runAbortController: new AbortController(),
sessionAgentId: "main",
pluginConfig: {},
onYieldDetected: () => undefined,
});
expect(factory).toHaveBeenCalledTimes(1);
expect(factory.mock.calls[0]?.[0]?.enableHeartbeatTool).toBe(true);
expect(catalog.tools.map((tool) => tool.name)).toEqual(["message"]);
expect(catalog.registeredTools.map((tool) => tool.name)).toEqual([
"message",
"web_search",
"heartbeat_respond",
]);
});
it("applies additional Codex dynamic tool excludes without exposing Codex-native tools", () => {
const tools = ["read", "exec", "message", "custom_tool"].map((name) => ({ name }));
@@ -608,28 +549,6 @@ describe("Codex app-server dynamic tool build", () => {
);
});
it("passes native and routable channel targets into Codex dynamic tools", async () => {
const sessionFile = path.join(tempDir, "session.jsonl");
const workspaceDir = path.join(tempDir, "workspace");
const params = createParams(sessionFile, workspaceDir);
params.disableTools = false;
params.currentChannelId = "D123";
params.currentMessagingTarget = "user:U123";
params.runtimePlan = createCodexRuntimePlanFixture();
const factoryOptions: unknown[] = [];
setOpenClawCodingToolsFactoryForTests((options) => {
factoryOptions.push(options);
return [];
});
await buildDynamicToolsForTest(params, workspaceDir, { sandbox: null as never });
expect(factoryOptions[0]).toMatchObject({
currentChannelId: "D123",
currentMessagingTarget: "user:U123",
});
});
it("passes runtime config into Codex exec dynamic tool construction", async () => {
const sessionFile = path.join(tempDir, "session.jsonl");
const workspaceDir = path.join(tempDir, "workspace");

View File

@@ -38,9 +38,6 @@ type OpenClawCodingToolsOptions = NonNullable<
export type OpenClawCodingToolsFactory =
(typeof import("openclaw/plugin-sdk/agent-harness"))["createOpenClawCodingTools"];
type OpenClawDynamicTool = ReturnType<OpenClawCodingToolsFactory>[number];
type OpenClawDynamicToolProjection = ReturnType<
typeof filterProviderNormalizableTools<OpenClawDynamicTool>
>;
type OpenClawSandboxContext = Awaited<ReturnType<typeof resolveSandboxContext>>;
type CodexDynamicToolBuildEvent = Parameters<
NonNullable<EmbeddedRunAttemptParams["onAgentEvent"]>
@@ -55,7 +52,6 @@ const CODEX_NATIVE_SANDBOX_TOOL_REQUIREMENTS = [
"apply_patch",
] as const;
const CODEX_MEMORY_FLUSH_DYNAMIC_TOOL_ALLOW = new Set(["read", "write"]);
const CODEX_HEARTBEAT_DYNAMIC_TOOL_NAME = "heartbeat_respond";
/** Runtime inputs needed to derive the exact Codex dynamic tool surface for a turn. */
export type DynamicToolBuildParams = {
@@ -70,6 +66,8 @@ export type DynamicToolBuildParams = {
sessionAgentId: string;
pluginConfig: CodexPluginConfig;
profilerEnabled?: boolean;
forceHeartbeatTool?: boolean;
ignoreRuntimePlan?: boolean;
onYieldDetected: () => void;
onCodexAppServerEvent?: (event: CodexDynamicToolBuildEvent) => void;
};
@@ -98,13 +96,6 @@ export function resolveOpenClawCodingToolsSessionKeys(
};
}
/** Returns the canonical channel used for Codex message routing and receipts. */
export function resolveCodexMessageToolProvider(
params: Pick<EmbeddedRunAttemptParams, "messageChannel" | "messageProvider">,
): string | undefined {
return params.messageChannel ?? params.messageProvider;
}
/** Resolves the channel id that hook events should target for this Codex app-server turn. */
export function resolveCodexAppServerHookChannelId(
params: EmbeddedRunAttemptParams,
@@ -130,11 +121,6 @@ type CodexDynamicToolBuildStageSummary = {
stages: CodexDynamicToolBuildStageTiming[];
};
type CodexDynamicToolBuildStageTracker = {
mark: (name: string) => void;
snapshot: () => CodexDynamicToolBuildStageSummary;
};
const CODEX_DYNAMIC_TOOL_BUILD_WARN_TOTAL_MS = 1_000;
const CODEX_DYNAMIC_TOOL_BUILD_WARN_STAGE_MS = 500;
@@ -196,42 +182,17 @@ export function formatCodexDynamicToolBuildStageSummary(
: "none";
}
/** Builds the turn-visible and durable registration views from one OpenClaw tool catalog. */
export async function prepareDynamicToolCatalog(input: DynamicToolBuildParams): Promise<{
tools: OpenClawDynamicTool[];
registeredTools: OpenClawDynamicTool[];
}> {
/** Builds, filters, and normalizes Codex-compatible runtime tools for a single turn. */
export async function buildDynamicTools(input: DynamicToolBuildParams) {
const { params } = input;
if (params.disableTools || !supportsModelTools(params.model)) {
return { tools: [], registeredTools: [] };
return [];
}
// Dynamic tool construction is on the reply hot path, so per-stage
// Date.now/span bookkeeping runs only when the Codex profiler flag is set.
const toolBuildStages = createCodexDynamicToolBuildStageTracker({
enabled: input.profilerEnabled,
});
// The durable schema must include heartbeat_respond across normal and heartbeat
// turns. Build that superset once, then hide it only from normal turn exposure.
const allTools = await buildOpenClawDynamicToolSource(input, toolBuildStages);
const readableTools = filterProviderNormalizableTools(allTools);
toolBuildStages.mark("provider-normalization");
const tools = projectDynamicTools(input, readableTools, toolBuildStages, {
excludeHeartbeatTool: params.trigger !== "heartbeat",
phase: "runtime-tools",
stagePrefix: "runtime",
});
const registeredTools = projectDynamicTools(input, readableTools, toolBuildStages, {
ignoreRuntimePlan: true,
phase: "registered-tools",
reportDiagnostics: false,
stagePrefix: "registered",
});
return { tools, registeredTools };
}
async function buildOpenClawDynamicToolSource(
input: DynamicToolBuildParams,
toolBuildStages: CodexDynamicToolBuildStageTracker,
): Promise<OpenClawDynamicTool[]> {
const { params } = input;
const modelHasVision = params.model.input?.includes("image") ?? false;
const agentDir = params.agentDir ?? resolveAgentDir(params.config ?? {}, input.sessionAgentId);
const createOpenClawCodingTools =
@@ -248,8 +209,7 @@ async function buildOpenClawDynamicToolSource(
elevated: params.bashElevated,
},
sandbox: input.sandbox,
messageProvider: resolveCodexMessageToolProvider(params),
toolPolicyMessageProvider: params.messageProvider ?? params.messageChannel,
messageProvider: params.messageChannel ?? params.messageProvider,
agentAccountId: params.agentAccountId,
messageTo: params.messageTo,
messageThreadId: params.messageThreadId,
@@ -298,7 +258,6 @@ async function buildOpenClawDynamicToolSource(
),
suppressManagedWebSearch: false,
currentChannelId: params.currentChannelId,
currentMessagingTarget: params.currentMessagingTarget,
hookChannelId: resolveCodexAppServerHookChannelId(params, input.sandboxSessionKey),
currentThreadTs: params.currentThreadTs,
currentMessageId: params.currentMessageId,
@@ -310,8 +269,8 @@ async function buildOpenClawDynamicToolSource(
sourceReplyDeliveryMode: params.sourceReplyDeliveryMode,
disableMessageTool: params.disableMessageTool,
forceMessageTool: shouldForceMessageTool(params),
enableHeartbeatTool: true,
forceHeartbeatTool: true,
enableHeartbeatTool: params.trigger === "heartbeat" || input.forceHeartbeatTool === true,
forceHeartbeatTool: params.trigger === "heartbeat" || input.forceHeartbeatTool === true,
onYield: (message) => {
input.onYieldDetected();
input.onCodexAppServerEvent?.({
@@ -324,30 +283,10 @@ async function buildOpenClawDynamicToolSource(
},
});
toolBuildStages.mark("create-openclaw-coding-tools");
return allTools;
}
function projectDynamicTools(
input: DynamicToolBuildParams,
source: OpenClawDynamicToolProjection,
toolBuildStages: CodexDynamicToolBuildStageTracker,
options: {
excludeHeartbeatTool?: boolean;
ignoreRuntimePlan?: boolean;
phase?: "runtime-tools" | "registered-tools";
reportDiagnostics?: boolean;
stagePrefix?: string;
} = {},
): OpenClawDynamicTool[] {
const { params } = input;
const markStage = (name: string) =>
toolBuildStages.mark(options.stagePrefix ? `${options.stagePrefix}-${name}` : name);
const preNormalizationDiagnostics: RuntimeToolSchemaDiagnostic[] = [...source.diagnostics];
const readableAllTools = [...source.tools].filter(
(tool) =>
!options.excludeHeartbeatTool ||
normalizeCodexDynamicToolName(tool.name) !== CODEX_HEARTBEAT_DYNAMIC_TOOL_NAME,
);
const preNormalizationDiagnostics: RuntimeToolSchemaDiagnostic[] = [];
const readableAllToolProjection = filterProviderNormalizableTools(allTools);
preNormalizationDiagnostics.push(...readableAllToolProjection.diagnostics);
const readableAllTools = [...readableAllToolProjection.tools];
const codexFilteredTools = addNodeShellDynamicToolsIfNeeded(
addSandboxShellDynamicToolsIfAvailable(
isCodexMemoryFlushRun(params)
@@ -359,18 +298,17 @@ function projectDynamicTools(
readableAllTools,
input,
);
markStage("codex-filtering");
const modelHasVision = params.model.input?.includes("image") ?? false;
toolBuildStages.mark("codex-filtering");
const visionFilteredTools = filterToolsForVisionInputs(codexFilteredTools, {
modelHasVision,
hasInboundImages: (params.images?.length ?? 0) > 0,
});
markStage("vision-filtering");
toolBuildStages.mark("vision-filtering");
const toolsAllow = includeForcedCodexDynamicToolAllow(params.toolsAllow, params);
const filteredTools = filterCodexDynamicToolsForAllowlist(visionFilteredTools, toolsAllow);
markStage("allowlist-filter");
toolBuildStages.mark("allowlist-filter");
const normalizedTools = normalizeAgentRuntimeTools({
runtimePlan: options.ignoreRuntimePlan ? undefined : params.runtimePlan,
runtimePlan: input.ignoreRuntimePlan ? undefined : params.runtimePlan,
tools: filteredTools,
provider: params.provider,
config: params.config,
@@ -379,14 +317,11 @@ function projectDynamicTools(
modelId: params.modelId,
modelApi: params.model.api,
model: params.model,
// Registration is a projection of the already-prepared catalog. Never
// activate another provider runtime while constructing its durable schema.
allowProviderRuntimePluginLoad: options.ignoreRuntimePlan ? false : undefined,
onPreNormalizationSchemaDiagnostics: (diagnostics) =>
preNormalizationDiagnostics.push(...diagnostics),
});
markStage("runtime-normalization");
if (options.reportDiagnostics !== false && preNormalizationDiagnostics.length > 0) {
toolBuildStages.mark("runtime-normalization");
if (preNormalizationDiagnostics.length > 0) {
embeddedAgentLog.warn(
`codex app-server quarantined ${preNormalizationDiagnostics.length} unsupported runtime tool schema${preNormalizationDiagnostics.length === 1 ? "" : "s"} before dynamic tool registration`,
{
@@ -403,7 +338,7 @@ function projectDynamicTools(
}
const summary = toolBuildStages.snapshot();
if (shouldWarnCodexDynamicToolBuildStageSummary(summary)) {
const phase = options.phase ?? "runtime-tools";
const phase = input.forceHeartbeatTool ? "registered-tools" : "runtime-tools";
embeddedAgentLog.warn(
`codex app-server dynamic tool build timings runId=${params.runId} sessionId=${params.sessionId} phase=${phase} totalMs=${summary.totalMs} stages=${formatCodexDynamicToolBuildStageSummary(summary)}`,
{
@@ -417,7 +352,8 @@ function projectDynamicTools(
visionFilteredToolCount: visionFilteredTools.length,
filteredToolCount: filteredTools.length,
normalizedToolCount: normalizedTools.length,
ignoreRuntimePlan: options.ignoreRuntimePlan === true,
forceHeartbeatTool: input.forceHeartbeatTool === true,
ignoreRuntimePlan: input.ignoreRuntimePlan === true,
nativeToolSurfaceEnabled: input.nativeToolSurfaceEnabled === true,
},
);

View File

@@ -18,7 +18,6 @@ import {
import {
createEmptyPluginRegistry,
createMockPluginRegistry,
createTestRegistry,
setActivePluginRegistry,
} from "openclaw/plugin-sdk/plugin-test-runtime";
import { afterEach, describe, expect, it, vi } from "vitest";
@@ -799,163 +798,6 @@ describe("createCodexDynamicToolBridge", () => {
]);
});
it("records the current provider and transport thread for implicit message sends", async () => {
const hasRepliedRef = { value: false };
setActivePluginRegistry(
createTestRegistry([
{
pluginId: "slack",
plugin: {
id: "slack",
messaging: { normalizeTarget: (raw: string) => raw.trim().toLowerCase() },
threading: {
resolveAutoThreadId: ({
to,
toolContext,
}: {
to: string;
toolContext?: {
currentChannelId?: string;
currentMessagingTarget?: string;
currentThreadTs?: string;
replyToMode?: "off" | "first" | "all" | "batched";
hasRepliedRef?: { value: boolean };
};
}) => {
if (
to !== toolContext?.currentMessagingTarget &&
to !== toolContext?.currentChannelId
) {
return undefined;
}
if (
(toolContext?.replyToMode === "first" ||
toolContext?.replyToMode === "batched") &&
!toolContext.hasRepliedRef?.value
) {
return toolContext.currentThreadTs;
}
return undefined;
},
},
},
source: "test",
},
]),
);
const bridge = createCodexDynamicToolBridge({
tools: [
createTool({
name: "message",
execute: vi.fn(async () => {
hasRepliedRef.value = true;
return textToolResult("Sent.");
}),
}),
],
signal: new AbortController().signal,
hookContext: {
currentChannelProvider: "slack",
currentChannelId: "D1",
currentMessagingTarget: "user:u1",
currentThreadId: "171.222",
replyToMode: "first",
hasRepliedRef,
},
});
await handleMessageToolCall(bridge, {
action: "send",
to: "user:U1",
text: "hello from Codex",
});
expect(bridge.telemetry.messagingToolSentTargets).toEqual([
{
tool: "message",
provider: "slack",
to: "user:u1",
threadId: "171.222",
threadImplicit: true,
text: "hello from Codex",
},
]);
});
it("records the provider-confirmed route for successful message sends", async () => {
const registry = createTestRegistry([
{
pluginId: "mattermost",
plugin: {
id: "mattermost",
messaging: { normalizeTarget: (raw: string) => raw.trim().toLowerCase() },
actions: {
extractToolSend: ({ args }: { args: Record<string, unknown> }) =>
args.action === "send" && typeof args.to === "string"
? { to: args.to, threadImplicit: true }
: null,
extractToolSendResult: ({ result }: { result: unknown }) => {
const details = requireRecord(
requireRecord(result, "message result").details,
"message details",
);
const toolSend = requireRecord(details.toolSend, "tool send details");
return {
to: String(toolSend.to),
threadId: String(toolSend.threadId),
};
},
},
},
source: "test",
},
]);
const middleware = vi.fn(async (event: { result: AgentToolResult<unknown> }) => {
const details = requireRecord(event.result.details, "middleware details");
const toolSend = requireRecord(details.toolSend, "middleware tool send");
toolSend.to = "channel:corrupted";
toolSend.threadId = "corrupted-root";
return undefined;
});
registry.agentToolResultMiddlewares.push({
pluginId: "route-details-stripper",
pluginName: "Route details stripper",
rawHandler: middleware,
handler: middleware,
runtimes: ["codex"],
source: "test",
});
setActivePluginRegistry(registry);
const bridge = createBridgeWithToolResult(
"message",
textToolResult("Sent.", {
toolSend: {
to: "channel:resolved-id",
threadId: "root-post-id",
},
}),
);
await handleMessageToolCall(bridge, {
action: "send",
provider: "mattermost",
to: "town-square",
text: "hello from Codex",
});
expect(bridge.telemetry.messagingToolSentTargets).toEqual([
{
tool: "message",
provider: "mattermost",
to: "channel:resolved-id",
threadId: "root-post-id",
threadImplicit: undefined,
threadSuppressed: undefined,
text: "hello from Codex",
},
]);
});
it("records message tool media attachment aliases as delivery evidence", async () => {
const toolResult = {
content: [{ type: "text", text: "Sent." }],

View File

@@ -6,8 +6,6 @@ import type { AgentToolResult } from "openclaw/plugin-sdk/agent-core";
import {
createAgentToolResultMiddlewareRunner,
createCodexAppServerToolResultExtensionRunner,
extractMessagingToolSend,
extractMessagingToolSendResult,
extractToolResultMediaArtifact,
filterToolResultMediaUrls,
HEARTBEAT_RESPONSE_TOOL_NAME,
@@ -53,22 +51,10 @@ type CodexDynamicToolHookContext = {
sessionKey?: string;
runId?: string;
channelId?: string;
currentChannelProvider?: string;
currentChannelId?: string;
currentMessagingTarget?: string;
currentThreadId?: string;
replyToMode?: "off" | "first" | "all" | "batched";
hasRepliedRef?: { value: boolean };
};
type CodexToolResultHookContext = Omit<CodexDynamicToolHookContext, "config">;
type AgentToolResultObserver = (event: {
toolName: string;
result: unknown;
isError: boolean;
}) => void;
type ProjectedCodexDynamicTool = {
tool: AnyAgentTool;
name: string;
@@ -81,22 +67,6 @@ type CodexDynamicToolSchemaQuarantine = {
violations: readonly string[];
};
function applyCurrentMessageProvider(
toolName: string,
args: Record<string, unknown>,
currentProvider: string | undefined,
): Record<string, unknown> {
const hasProvider =
typeof args.provider === "string" && args.provider.trim().length > 0
? true
: typeof args.channel === "string" && args.channel.trim().length > 0;
const provider = currentProvider?.trim();
if (toolName !== "message" || hasProvider || !provider) {
return args;
}
return { ...args, provider };
}
/** Runtime bridge returned to Codex app-server attempt code. */
export type CodexDynamicToolBridge = {
availableSpecs: CodexDynamicToolSpec[];
@@ -105,7 +75,7 @@ export type CodexDynamicToolBridge = {
params: CodexDynamicToolCallParams,
options?: {
signal?: AbortSignal;
onAgentToolResult?: AgentToolResultObserver;
onAgentToolResult?: EmbeddedRunAttemptParams["onAgentToolResult"];
},
) => Promise<CodexDynamicToolCallResponse>;
telemetry: {
@@ -243,30 +213,9 @@ export function createCodexDynamicToolBridge(params: {
// Prepare before marking side-effect evidence; argument preparation can
// fail without the target tool actually starting.
const preparedArgs = tool.prepareArguments ? tool.prepareArguments(args) : args;
const telemetryArgs = isRecord(preparedArgs) ? preparedArgs : args;
const messagingTelemetryArgs = applyCurrentMessageProvider(
toolName,
telemetryArgs,
params.hookContext?.currentChannelProvider,
);
const messagingTarget =
isMessagingTool(toolName) && isMessagingToolSendAction(toolName, telemetryArgs)
? extractMessagingToolSend(toolName, messagingTelemetryArgs, {
config: params.hookContext?.config,
currentChannelId: params.hookContext?.currentChannelId,
currentMessagingTarget: params.hookContext?.currentMessagingTarget,
currentThreadId: params.hookContext?.currentThreadId,
replyToMode: params.hookContext?.replyToMode,
hasRepliedRef: params.hookContext?.hasRepliedRef,
})
: undefined;
didStartExecution = true;
const rawResult = await tool.execute(call.callId, preparedArgs, signal);
const rawIsError = isCodexToolResultError(rawResult);
const confirmedMessagingTarget =
!rawIsError && messagingTarget
? extractMessagingToolSendResult(messagingTarget, rawResult)
: messagingTarget;
const middlewareResult = await middlewareRunner.applyToolResultMiddleware({
threadId: call.threadId,
turnId: call.turnId,
@@ -288,12 +237,11 @@ export function createCodexDynamicToolBridge(params: {
notifyAgentToolResult(options?.onAgentToolResult, toolName, result, resultIsError);
collectToolTelemetry({
toolName,
args: telemetryArgs,
args,
result,
mediaTrustResult: rawResult,
telemetry,
isError: resultIsError,
messagingTarget: confirmedMessagingTarget,
});
void runAgentHarnessAfterToolCallHook({
toolName,
@@ -375,7 +323,7 @@ export function createCodexDynamicToolBridge(params: {
}
function notifyAgentToolResult(
observer: AgentToolResultObserver | undefined,
observer: EmbeddedRunAttemptParams["onAgentToolResult"] | undefined,
toolName: string,
result: unknown,
isError: boolean,
@@ -683,7 +631,6 @@ function collectToolTelemetry(params: {
mediaTrustResult?: AgentToolResult<unknown>;
telemetry: CodexDynamicToolBridge["telemetry"];
isError: boolean;
messagingTarget?: MessagingToolSend;
}): void {
if (params.isError) {
return;
@@ -736,13 +683,11 @@ function collectToolTelemetry(params: {
const mediaUrls = collectMediaUrls(params.args);
params.telemetry.messagingToolSentMediaUrls.push(...mediaUrls);
params.telemetry.messagingToolSentTargets.push({
...(params.messagingTarget ?? {
tool: params.toolName,
provider: readFirstString(params.args, ["provider", "channel"]) ?? params.toolName,
accountId: readFirstString(params.args, ["accountId", "account_id"]),
to: readFirstString(params.args, ["to", "target", "recipient"]),
threadId: readFirstString(params.args, ["threadId", "thread_id", "messageThreadId"]),
}),
tool: params.toolName,
provider: readFirstString(params.args, ["provider", "channel"]) ?? params.toolName,
accountId: readFirstString(params.args, ["accountId", "account_id"]),
to: readFirstString(params.args, ["to", "target", "recipient"]),
threadId: readFirstString(params.args, ["threadId", "thread_id", "messageThreadId"]),
...(text ? { text } : {}),
...(mediaUrls.length > 0 ? { mediaUrls } : {}),
});

View File

@@ -24,6 +24,7 @@ import {
type CodexAppServerEventProjectorOptions,
type CodexAppServerToolTelemetry,
} from "./event-projector.js";
import { rememberCodexRateLimits, resetCodexRateLimitCacheForTests } from "./rate-limit-cache.js";
import { createCodexTestModel } from "./test-support.js";
const THREAD_ID = "thread-1";
@@ -107,6 +108,7 @@ afterEach(async () => {
resetAgentEventsForTest();
resetDiagnosticEventsForTest();
resetGlobalHookRunner();
resetCodexRateLimitCacheForTests();
vi.restoreAllMocks();
vi.unstubAllEnvs();
for (const tempDir of tempDirs) {
@@ -859,11 +861,10 @@ describe("CodexAppServerEventProjector", () => {
});
it("uses Codex rate-limit resets for usage-limit app-server errors", async () => {
const projector = await createProjector();
const resetsAt = Math.ceil(Date.now() / 1000) + 120;
const projector = await createProjector(undefined, {
readRecentRateLimits: () => rateLimitsUpdated(resetsAt).params,
});
await projector.handleNotification(rateLimitsUpdated(resetsAt));
await projector.handleNotification(
forCurrentTurn("error", {
error: {
@@ -884,11 +885,10 @@ describe("CodexAppServerEventProjector", () => {
});
it("uses Codex rate-limit resets for failed turns", async () => {
const projector = await createProjector();
const resetsAt = Math.ceil(Date.now() / 1000) + 120;
const projector = await createProjector(undefined, {
readRecentRateLimits: () => rateLimitsUpdated(resetsAt).params,
});
await projector.handleNotification(rateLimitsUpdated(resetsAt));
await projector.handleNotification(
forCurrentTurn("turn/completed", {
turn: {
@@ -912,8 +912,9 @@ describe("CodexAppServerEventProjector", () => {
});
it("uses a recent Codex rate-limit snapshot when failed turns omit reset details", async () => {
const projector = await createProjector();
const resetsAt = Math.ceil(Date.now() / 1000) + 120;
const rateLimits = {
rememberCodexRateLimits({
rateLimits: {
limitId: "codex",
limitName: "Codex",
@@ -924,9 +925,6 @@ describe("CodexAppServerEventProjector", () => {
rateLimitReachedType: "rate_limit_reached",
},
rateLimitsByLimitId: null,
};
const projector = await createProjector(undefined, {
readRecentRateLimits: () => rateLimits,
});
await projector.handleNotification(
@@ -978,19 +976,19 @@ describe("CodexAppServerEventProjector", () => {
expect(result.promptErrorSource).toBe("prompt");
});
it("normalizes current app-server token usage", async () => {
it("normalizes snake_case current token usage fields", async () => {
const projector = await createProjector();
await projector.handleNotification(agentMessageDelta("done"));
await projector.handleNotification(
forCurrentTurn("thread/tokenUsage/updated", {
tokenUsage: {
total: { totalTokens: 1_000_000 },
last: {
totalTokens: 17,
inputTokens: 8,
cachedInputTokens: 3,
outputTokens: 9,
total: { total_tokens: 1_000_000 },
last_token_usage: {
total_tokens: 17,
input_tokens: 8,
cached_input_tokens: 3,
output_tokens: 9,
},
},
}),

View File

@@ -26,7 +26,10 @@ import type { AssistantMessage, Usage } from "openclaw/plugin-sdk/llm";
import { saveMediaBuffer } from "openclaw/plugin-sdk/media-store";
import { asDateTimestampMs } from "openclaw/plugin-sdk/number-runtime";
import { resolveCodexLocalRuntimeAttribution } from "./local-runtime-attribution.js";
import { isCodexNotificationForTurn } from "./notification-correlation.js";
import {
readCodexNotificationThreadId,
readCodexNotificationTurnId,
} from "./notification-correlation.js";
import { readCodexTurn } from "./protocol-validators.js";
import {
isJsonObject,
@@ -37,6 +40,7 @@ import {
type JsonObject,
type JsonValue,
} from "./protocol.js";
import { readRecentCodexRateLimits, rememberCodexRateLimits } from "./rate-limit-cache.js";
import { formatCodexUsageLimitErrorMessage } from "./rate-limits.js";
import { readCodexMirroredSessionHistoryMessages } from "./session-history.js";
import {
@@ -61,7 +65,6 @@ export type CodexAppServerToolTelemetry = {
export type CodexAppServerEventProjectorOptions = {
nativePostToolUseRelayEnabled?: boolean;
readRecentRateLimits?: () => JsonValue | undefined;
trajectoryRecorder?: CodexTrajectoryRecorder | null;
};
@@ -89,6 +92,22 @@ const ZERO_USAGE: Usage = {
},
};
const CURRENT_TOKEN_USAGE_KEYS = [
"last",
"current",
"lastCall",
"lastCallUsage",
"lastTokenUsage",
"last_token_usage",
] as const;
const CODEX_PROMPT_TOTAL_INPUT_KEYS = [
"inputTokens",
"input_tokens",
"promptTokens",
"prompt_tokens",
] as const;
const MAX_TOOL_OUTPUT_DELTA_MESSAGES_PER_ITEM = 20;
const TOOL_TRANSCRIPT_OUTPUT_MAX_CHARS = 12_000;
const MISSING_TOOL_RESULT_ERROR =
@@ -179,6 +198,8 @@ export class CodexAppServerEventProjector {
private tokenUsage: ReturnType<typeof normalizeUsage>;
private guardianReviewCount = 0;
private completedCompactionCount = 0;
private latestRateLimits: JsonValue | undefined;
constructor(
private readonly params: EmbeddedRunAttemptParams,
private readonly threadId: string,
@@ -200,6 +221,11 @@ export class CodexAppServerEventProjector {
if (!params) {
return;
}
if (notification.method === "account/rateLimits/updated") {
this.latestRateLimits = params;
rememberCodexRateLimits(params);
return;
}
if (isHookNotificationMethod(notification.method)) {
if (!this.isHookNotificationForCurrentThread(params)) {
return;
@@ -252,7 +278,7 @@ export class CodexAppServerEventProjector {
await this.handleRawResponseItemCompleted(params);
break;
case "error":
if (params.willRetry === true) {
if (readBooleanAlias(params, ["willRetry", "will_retry"]) === true) {
break;
}
this.promptError = this.formatCodexErrorMessage(params) ?? "codex app-server error";
@@ -645,7 +671,9 @@ export class CodexAppServerEventProjector {
private handleTokenUsage(params: JsonObject): void {
const tokenUsage = isJsonObject(params.tokenUsage) ? params.tokenUsage : undefined;
const current = tokenUsage && isJsonObject(tokenUsage.last) ? tokenUsage.last : undefined;
const current =
(tokenUsage ? readFirstJsonObject(tokenUsage, CURRENT_TOKEN_USAGE_KEYS) : undefined) ??
readFirstJsonObject(params, CURRENT_TOKEN_USAGE_KEYS);
if (!current) {
return;
}
@@ -716,7 +744,7 @@ export class CodexAppServerEventProjector {
formatCodexUsageLimitErrorMessage({
message: turn.error?.message,
codexErrorInfo: turn.error?.codexErrorInfo as JsonValue | null | undefined,
rateLimits: this.options.readRecentRateLimits?.(),
rateLimits: this.latestRateLimits ?? readRecentCodexRateLimits(),
}) ??
turn.error?.message ??
"codex app-server turn failed";
@@ -1583,7 +1611,7 @@ export class CodexAppServerEventProjector {
formatCodexUsageLimitErrorMessage({
message: error ? readString(error, "message") : undefined,
codexErrorInfo: error?.codexErrorInfo,
rateLimits: this.options.readRecentRateLimits?.(),
rateLimits: this.latestRateLimits ?? readRecentCodexRateLimits(),
}) ?? readCodexErrorNotificationMessage(params)
);
}
@@ -1758,7 +1786,9 @@ export class CodexAppServerEventProjector {
}
private isNotificationForTurn(params: JsonObject): boolean {
return isCodexNotificationForTurn(params, this.threadId, this.turnId);
const threadId = readCodexNotificationThreadId(params);
const turnId = readNotificationTurnId(params);
return threadId === this.threadId && turnId === this.turnId;
}
private isHookNotificationForCurrentThread(params: JsonObject): boolean {
@@ -1772,6 +1802,10 @@ function isHookNotificationMethod(method: string): method is "hook/started" | "h
return method === "hook/started" || method === "hook/completed";
}
function readNotificationTurnId(record: JsonObject): string | undefined {
return readCodexNotificationTurnId(record);
}
function readString(record: JsonObject, key: string): string | undefined {
const value = record[key];
return typeof value === "string" ? value : undefined;
@@ -1861,6 +1895,21 @@ function readNonNegativeInteger(record: JsonObject, key: string): number | undef
return value !== undefined && Number.isInteger(value) && value >= 0 ? value : undefined;
}
function readBoolean(record: JsonObject, key: string): boolean | undefined {
const value = record[key];
return typeof value === "boolean" ? value : undefined;
}
function readBooleanAlias(record: JsonObject, keys: readonly string[]): boolean | undefined {
for (const key of keys) {
const value = readBoolean(record, key);
if (value !== undefined) {
return value;
}
}
return undefined;
}
function readCodexErrorNotificationMessage(record: JsonObject): string | undefined {
const error = record.error;
if (isJsonObject(error)) {
@@ -1888,19 +1937,52 @@ function readHookOutputEntries(
});
}
function readFirstJsonObject(record: JsonObject, keys: readonly string[]): JsonObject | undefined {
for (const key of keys) {
const value = record[key];
if (isJsonObject(value)) {
return value;
}
}
return undefined;
}
function readNumberAlias(record: JsonObject, keys: readonly string[]): number | undefined {
for (const key of keys) {
const value = readNumber(record, key);
if (value !== undefined) {
return value;
}
}
return undefined;
}
function normalizeCodexTokenUsage(record: JsonObject): ReturnType<typeof normalizeUsage> {
const promptTotalInput = readNumber(record, "inputTokens");
const cacheRead = readNumber(record, "cachedInputTokens");
const promptTotalInput = readNumberAlias(record, CODEX_PROMPT_TOTAL_INPUT_KEYS);
const cacheRead = readNumberAlias(record, [
"cachedInputTokens",
"cached_input_tokens",
"cacheRead",
"cache_read",
"cache_read_input_tokens",
"cached_tokens",
]);
const input =
promptTotalInput !== undefined && cacheRead !== undefined
? Math.max(0, promptTotalInput - cacheRead)
: promptTotalInput;
: (promptTotalInput ?? readNumber(record, "input"));
return normalizeUsage({
input,
output: readNumber(record, "outputTokens"),
output: readNumberAlias(record, ["outputTokens", "output_tokens", "output"]),
cacheRead,
total: readNumber(record, "totalTokens"),
cacheWrite: readNumberAlias(record, [
"cacheWrite",
"cache_write",
"cacheCreationInputTokens",
"cache_creation_input_tokens",
]),
total: readNumberAlias(record, ["totalTokens", "total_tokens", "total"]),
});
}

View File

@@ -8,10 +8,6 @@ import type { CodexAppServerClient } from "./client.js";
import type { CodexAppServerStartOptions } from "./config.js";
import { readCodexModelListResponse } from "./protocol-validators.js";
import type { CodexModel, CodexReasoningEffortOption } from "./protocol.js";
import {
createIsolatedCodexAppServerClient,
leaseSharedCodexAppServerClient,
} from "./shared-client.js";
/** Normalized model metadata returned by the Codex app-server model listing helper. */
export type CodexAppServerModel = {
@@ -40,11 +36,10 @@ export type CodexAppServerListModelsOptions = {
includeHidden?: boolean;
timeoutMs?: number;
startOptions?: CodexAppServerStartOptions;
authProfileId?: string | null;
authProfileId?: string;
agentDir?: string;
config?: Parameters<typeof resolveCodexAppServerAuthProfileIdForAgent>[0]["config"];
sharedClient?: boolean;
signal?: AbortSignal;
};
/** Lists one Codex app-server model page using the configured auth/client options. */
@@ -59,37 +54,27 @@ export async function listCodexAppServerModels(
/** Walks Codex app-server model pages until exhaustion or the max-page guard. */
export async function listAllCodexAppServerModels(
options: CodexAppServerListModelsOptions & { maxPages?: number } = {},
): Promise<CodexAppServerModelListResult> {
return await withCodexAppServerModelClient(options, async ({ client, timeoutMs }) =>
listAllCodexAppServerModelsWithClient(client, { ...options, timeoutMs }),
);
}
/** Walks all model pages on an already-owned physical app-server client. */
export async function listAllCodexAppServerModelsWithClient(
client: CodexAppServerClient,
options: CodexAppServerListModelsOptions & { maxPages?: number } = {},
): Promise<CodexAppServerModelListResult> {
const maxPages = normalizeMaxPages(options.maxPages);
const timeoutMs = options.timeoutMs ?? 2500;
const models: CodexAppServerModel[] = [];
let cursor = options.cursor;
let nextCursor: string | undefined;
for (let page = 0; page < maxPages; page += 1) {
options.signal?.throwIfAborted();
const result = await requestModelListPage(client, {
...options,
timeoutMs,
cursor,
});
models.push(...result.models);
nextCursor = result.nextCursor;
if (!nextCursor) {
return { models };
return await withCodexAppServerModelClient(options, async ({ client, timeoutMs }) => {
const models: CodexAppServerModel[] = [];
let cursor = options.cursor;
let nextCursor: string | undefined;
for (let page = 0; page < maxPages; page += 1) {
const result = await requestModelListPage(client, {
...options,
timeoutMs,
cursor,
});
models.push(...result.models);
nextCursor = result.nextCursor;
if (!nextCursor) {
return { models };
}
cursor = nextCursor;
}
cursor = nextCursor;
}
return { models, nextCursor, truncated: true };
return { models, nextCursor, truncated: true };
});
}
async function withCodexAppServerModelClient<T>(
@@ -98,32 +83,33 @@ async function withCodexAppServerModelClient<T>(
): Promise<T> {
const timeoutMs = options.timeoutMs ?? 2500;
const useSharedClient = options.sharedClient !== false;
const clientLease = useSharedClient
? await leaseSharedCodexAppServerClient({
const {
createIsolatedCodexAppServerClient,
getLeasedSharedCodexAppServerClient,
releaseLeasedSharedCodexAppServerClient,
} = await import("./shared-client.js");
const client = useSharedClient
? await getLeasedSharedCodexAppServerClient({
startOptions: options.startOptions,
timeoutMs,
authProfileId: options.authProfileId,
agentDir: options.agentDir,
config: options.config,
abandonSignal: options.signal,
})
: undefined;
const client =
clientLease?.client ??
(await createIsolatedCodexAppServerClient({
startOptions: options.startOptions,
timeoutMs,
authProfileId: options.authProfileId,
agentDir: options.agentDir,
config: options.config,
}));
: await createIsolatedCodexAppServerClient({
startOptions: options.startOptions,
timeoutMs,
authProfileId: options.authProfileId,
agentDir: options.agentDir,
config: options.config,
});
try {
return await run({ client, timeoutMs });
} finally {
if (useSharedClient) {
clientLease?.release();
releaseLeasedSharedCodexAppServerClient(client);
} else {
await client.closeAndWait({ exitTimeoutMs: 2_000, forceKillDelayMs: 250 });
client.close();
}
}
}
@@ -139,7 +125,7 @@ async function requestModelListPage(
cursor: options.cursor ?? null,
includeHidden: options.includeHidden ?? null,
},
{ timeoutMs: options.timeoutMs, signal: options.signal },
{ timeoutMs: options.timeoutMs },
);
return readModelListResult(response);
}

View File

@@ -4,12 +4,7 @@
*/
import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
import { resolveSandboxRuntimeStatus } from "openclaw/plugin-sdk/sandbox";
import {
loadSessionStore,
resolveSessionStoreEntry,
resolveStorePath,
type SessionEntry,
} from "openclaw/plugin-sdk/session-store-runtime";
import { getSessionEntry, type SessionEntry } from "openclaw/plugin-sdk/session-store-runtime";
type ExecHost = "sandbox" | "gateway" | "node";
type ExecTarget = "auto" | ExecHost;
@@ -49,21 +44,20 @@ export function resolveCodexNativeExecutionPolicy(params: {
}): CodexNativeExecutionPolicy {
const config = params.config ?? {};
const sessionKey = params.sessionKey?.trim() || params.sessionId?.trim() || undefined;
const agentId = resolvePolicyAgentId({ config, sessionKey, agentId: params.agentId });
const sessionEntry =
params.sessionEntry ??
(params.readRuntimeSessionEntry && sessionKey
? readRuntimeSessionEntryBestEffort(config, sessionKey, agentId)
? readRuntimeSessionEntryBestEffort(sessionKey)
: undefined);
const sandboxAvailable =
params.sandboxAvailable ??
(sessionKey
? resolveSandboxRuntimeStatus({
cfg: config,
agentId,
sessionKey,
}).sandboxed
: false);
const agentId = resolvePolicyAgentId({ config, sessionKey, agentId: params.agentId });
const agentExec = resolvePolicyAgentExec({ config, agentId });
const globalExec = config.tools?.exec;
const requestedExecHost =
@@ -200,17 +194,9 @@ function resolveEffectiveExecHost(params: {
return params.requestedExecHost;
}
function readRuntimeSessionEntryBestEffort(
config: OpenClawConfig,
sessionKey: string,
agentId: string,
): SessionEntry | undefined {
function readRuntimeSessionEntryBestEffort(sessionKey: string): SessionEntry | undefined {
try {
const storePath = resolveStorePath(config.session?.store, { agentId });
return resolveSessionStoreEntry({
store: loadSessionStore(storePath, { skipCache: true }),
sessionKey,
}).existing;
return getSessionEntry({ sessionKey, hydrateSkillPromptRefs: false });
} catch {
return undefined;
}

View File

@@ -13,6 +13,7 @@ import {
addTimerTimeoutGraceMs,
finiteSecondsToTimerSafeMilliseconds,
} from "openclaw/plugin-sdk/number-runtime";
import type { CodexAppServerRuntimeOptions } from "./config.js";
import type { JsonObject, JsonValue } from "./protocol.js";
/** Codex hook events that can be registered through OpenClaw's native relay. */
@@ -23,6 +24,8 @@ export const CODEX_NATIVE_HOOK_RELAY_EVENTS: readonly NativeHookRelayEvent[] = [
"before_agent_finalize",
] as const;
const CODEX_NATIVE_HOOK_RELAY_EVENTS_WITH_APP_SERVER_APPROVALS =
CODEX_NATIVE_HOOK_RELAY_EVENTS.filter((event) => event !== "permission_request");
const CODEX_NATIVE_HOOK_RELAY_MIN_TTL_MS = 30 * 60_000;
/** Extra relay lifetime after the expected turn budget, preventing late hook drops. */
export const CODEX_NATIVE_HOOK_RELAY_TTL_GRACE_MS = 5 * 60_000;
@@ -146,8 +149,9 @@ export function createCodexNativeHookRelay(params: {
allowedEvents: params.events,
ttlMs: resolveCodexNativeHookRelayTtlMs({
explicitTtlMs: params.options?.ttlMs,
operationBudgetMs:
params.attemptTimeoutMs + params.startupTimeoutMs + params.turnStartTimeoutMs,
attemptTimeoutMs: params.attemptTimeoutMs,
startupTimeoutMs: params.startupTimeoutMs,
turnStartTimeoutMs: params.turnStartTimeoutMs,
}),
signal: params.signal,
command: {
@@ -159,27 +163,38 @@ export function createCodexNativeHookRelay(params: {
});
}
/** Selects the native hook events Codex should install for this thread. */
/** Selects the native hook events Codex should install for the current approval mode. */
export function resolveCodexNativeHookRelayEvents(params: {
configuredEvents?: readonly NativeHookRelayEvent[];
appServer: Pick<CodexAppServerRuntimeOptions, "approvalPolicy">;
}): readonly NativeHookRelayEvent[] {
if (params.configuredEvents?.length) {
return params.configuredEvents;
}
// Thread config is fixed before Codex reports the authoritative provider.
// Install the stable superset; the relay defers permission prompts from guarded turns.
return CODEX_NATIVE_HOOK_RELAY_EVENTS;
// Codex emits PermissionRequest before the app-server approval reviewer has
// resolved the command. In native approval modes, let Codex's app-server
// approval bridge own the real escalation instead of surfacing a stale
// pre-guardian OpenClaw plugin approval prompt.
return params.appServer.approvalPolicy === "never"
? CODEX_NATIVE_HOOK_RELAY_EVENTS
: CODEX_NATIVE_HOOK_RELAY_EVENTS_WITH_APP_SERVER_APPROVALS;
}
/** Derives the native hook relay TTL from the turn budget unless explicitly configured. */
export function resolveCodexNativeHookRelayTtlMs(params: {
explicitTtlMs: number | undefined;
operationBudgetMs: number;
attemptTimeoutMs: number;
startupTimeoutMs: number;
turnStartTimeoutMs: number;
}): number {
if (params.explicitTtlMs !== undefined) {
return params.explicitTtlMs;
}
const relayBudgetMs = params.operationBudgetMs + CODEX_NATIVE_HOOK_RELAY_TTL_GRACE_MS;
const relayBudgetMs =
params.attemptTimeoutMs +
params.startupTimeoutMs +
params.turnStartTimeoutMs +
CODEX_NATIVE_HOOK_RELAY_TTL_GRACE_MS;
return Math.max(CODEX_NATIVE_HOOK_RELAY_MIN_TTL_MS, Math.floor(relayBudgetMs));
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@@ -4,7 +4,6 @@ import {
extractCodexNativeSubagentCompletions,
extractCodexNativeSubagentCompletionsFromText,
} from "./native-subagent-notification.js";
import type { CodexServerNotification } from "./protocol.js";
function trustedInterAgentNotification(params: {
agentPath: string;
@@ -36,29 +35,6 @@ function trustedInterAgentNotification(params: {
};
}
function trustedAgentMessageNotification(params: {
agentPath: string;
text?: string;
encryptedContent?: string;
}): CodexServerNotification {
return {
method: "rawResponseItem/completed",
params: {
threadId: "parent-thread",
item: {
type: "agent_message",
author: params.agentPath,
recipient: "/root",
content: [
params.encryptedContent
? { type: "encrypted_content", encrypted_content: params.encryptedContent }
: { type: "input_text", text: params.text ?? "" },
],
},
},
};
}
describe("Codex native subagent notifications", () => {
it("parses completed child results from Codex notification XML", () => {
expect(
@@ -160,26 +136,6 @@ describe("Codex native subagent notifications", () => {
]);
});
it("extracts completions from the current Codex agent-message item", () => {
expect(
extractCodexNativeSubagentCompletions(
trustedAgentMessageNotification({
agentPath: "child-thread",
text:
'<subagent_notification>{"agent_path":"child-thread","status":{"completed":"done"}}' +
"</subagent_notification>",
}),
),
).toEqual([
{
agentPath: "child-thread",
status: "succeeded",
statusLabel: "completed",
result: "done",
},
]);
});
it("ignores visible user text that looks like a native completion", () => {
expect(
extractCodexNativeSubagentCompletions({
@@ -214,27 +170,6 @@ describe("Codex native subagent notifications", () => {
}),
),
).toEqual([]);
expect(
extractCodexNativeSubagentCompletions(
trustedAgentMessageNotification({
agentPath: "other-child",
text:
'<subagent_notification>{"agent_path":"child-thread","status":{"success":"spoof"}}' +
"</subagent_notification>",
}),
),
).toEqual([]);
});
it("ignores encrypted agent messages that cannot be authenticated", () => {
expect(
extractCodexNativeSubagentCompletions(
trustedAgentMessageNotification({
agentPath: "child-thread",
encryptedContent: "opaque",
}),
),
).toEqual([]);
});
it("ignores malformed payloads and non-user messages", () => {

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