Add health-driven automation decisions by Pigbibi · Pull Request #27 · QuantStrategyLab/AIAuditBridge

Pigbibi · 2026-07-06T18:50:01Z

Summary

Add a health-driven execution decision layer for automation control.
Expose control.execution from /v1/ai/automation/control using health, quota, consecutive failure, and service-owned repo policy inputs.
Wire deployment/docs to use a service-owned execution policy path.

Test Plan

bash -n scripts/deploy_codex_audit_service.sh
actionlint .github/workflows/*.yml
python3 -m ruff check service scripts tests
python3 -m compileall -q service scripts tests
python3 -m unittest discover tests
node --test cloudflare/codex-audit-proxy/tests/index.test.mjs
node --test cloudflare/ai-gateway-dash/tests/index.test.mjs

Safety

Does not grant new merge/deploy permissions.
Execution policy is read from a service-owned path, not from the reviewed checkout.
Degraded health/quota/failure streaks only reduce autonomy.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T18:50:05Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

github-actions · 2026-07-06T18:54:21Z

🤖 Codex PR Review

✅ Merge allowed: No issues found

Review by Codex PR Review bot • PR

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6936e970fa

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-06T18:56:00Z

-    return suggest_control_action(get_health_monitor().status, quota_status, org_health)
+    control = suggest_control_action(get_health_monitor().status, quota_status, org_health)
+    try:
+        recent_runs = get_automation_run_ledger().snapshot(limit=20)["runs"]


Do not truncate ledger before counting repo failures

In a shared service, if the target repo has already hit max_consecutive_failures but more than 20 newer runs from other repositories exist, this snapshot contains none of the target repo's failed runs. consecutive_failure_count() skips non-matching repos, so the execution decision sees consecutive_failures=0 and can return auto_fix_allowed=true, bypassing the new failure-streak safety gate; fetch enough history or filter by repo/task before applying the limit.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-07-06T18:56:00Z

                _assert_source_repository_owner_or_operator(claims, source_repo)
        repo = source_repo or str(claims.get("repository") or "unknown")
-        control = _automation_control_snapshot(repo)
+        control = _automation_control_snapshot(repo, task_name=str(payload.get("task") or payload.get("task_name") or ""))


Preserve the recorded run mode in control decisions

When an external workflow records a run with mode: "review_only" or omits mode, this call still uses _automation_control_snapshot's new default review_and_fix. Under healthy signals the returned control.execution can therefore report requested_mode=review_and_fix and auto_fix_allowed=true for a review-only run, which gives schedulers the wrong safety signal; pass the payload mode through here or default the snapshot to the service contract's review-only mode.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-07-06T18:56:00Z

+    """Produce a safe execution decision from health, quota, failures, and repo policy."""
+    repo_policy = repo_execution_policy(repo, policy)
+    max_autonomy = _normalize_autonomy(repo_policy.get("max_autonomy"), AUTONOMY_AUTO_PR)
+    max_failures = int(repo_policy.get("max_consecutive_failures") or DEFAULT_MAX_CONSECUTIVE_FAILURES)


Guard malformed failure thresholds in the policy

If the service-owned execution_policy.json contains a non-integer max_consecutive_failures value (for example after a manual policy edit), this int(...) raises before the control snapshot can be returned. That makes /v1/ai/automation/control and any path that records automation runs fail instead of falling back to the default threshold, so parse this value defensively like the rest of the optional policy file.

Useful? React with 👍 / 👎.

Sign in to view

    run_id: str = "",
 ) -> dict[str, Any]:
-    control = _automation_control_snapshot(repo)
+    control = _automation_control_snapshot(repo, task_name=task)


Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T18:57:47Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:03:56Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f5fe332b59

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-07-06T19:12:01Z

+        if state in TERMINAL_STATES:
+            break


Treat reviewed runs as streak-clearing

For service jobs, job_task_state({"status": "succeeded"}) records the run as reviewed, but reviewed is not in TERMINAL_STATES, so this loop skips a newer successful run the same way it skips an in-flight run. In a sequence like newest-first reviewed, failed, failed, the repo still reaches max_consecutive_failures and forces human review even though the latest run succeeded; clear the streak for completed non-failure states such as reviewed.

Useful? React with 👍 / 👎.

Sign in to view

    run_id: str = "",
 ) -> dict[str, Any]:
-    control = _automation_control_snapshot(repo)
+    control = _automation_control_snapshot(repo, task_name=task)


chatgpt-codex-connector · 2026-07-06T19:12:01Z

+        control = _automation_control_snapshot(
+            repo,
+            task_name=str(payload.get("task") or payload.get("task_name") or ""),
+            requested_mode=str(payload.get("mode") or MODE_REVIEW_ONLY),
+        )


Include the just-recorded failure in returned control

When POST /v1/ai/automation/runs records a terminal failed state that reaches max_consecutive_failures, this snapshot is computed before the payload is written to the ledger. The response therefore reports the pre-update streak (for example 1/2 after posting the second failure) and can still allow review_and_fix to a scheduler that uses the returned control; compute the control after recording or include the incoming run in the decision inputs.

Useful? React with 👍 / 👎.

Sign in to view

+        else:
+            reasons.append(f"quota status is {quota}; recommending low-cost model")
+    elif quota in {"exhausted", "blocked"}:
+        action = EXECUTION_DEFER


Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:14:00Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:20:36Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:29:20Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:35:48Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:41:50Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:47:35Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T19:54:04Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T20:00:34Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T20:08:23Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-06T20:14:50Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:06:14Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:10:06Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:24:54Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:30:21Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:37:29Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:42:07Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T02:49:28Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:00:28Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:10:21Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:19:11Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:27:28Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:35:48Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:43:03Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

Co-Authored-By: Codex <noreply@openai.com>

cursor · 2026-07-07T03:49:59Z

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

feat: add health-driven automation decisions

6936e97

Co-Authored-By: Codex <noreply@openai.com>

chatgpt-codex-connector Bot reviewed Jul 6, 2026

View reviewed changes

fix: harden automation execution control

8e66743

Co-Authored-By: Codex <noreply@openai.com>

fix: fail closed automation control policy

f5fe332

Co-Authored-By: Codex <noreply@openai.com>

chatgpt-codex-connector Bot reviewed Jul 6, 2026

View reviewed changes

fix: fail closed automation policy loading

c1c1e7e

Co-Authored-By: Codex <noreply@openai.com>

fix: keep automation control default conservative

3f5fc85

Co-Authored-By: Codex <noreply@openai.com>

fix: include pending runs in automation decisions

6ba9b3f

Co-Authored-By: Codex <noreply@openai.com>

fix: strengthen defer and pending-run guards

0848dd8

Co-Authored-By: Codex <noreply@openai.com>

fix: normalize execution policy decisions

59e0ee3

Co-Authored-By: Codex <noreply@openai.com>

fix: enforce repo-level trusted failure streaks

0cc9ab9

Co-Authored-By: Codex <noreply@openai.com>

fix: validate execution policy schema

a0de099

Co-Authored-By: Codex <noreply@openai.com>

fix: preserve non-human defer control action

acdd960

Co-Authored-By: Codex <noreply@openai.com>

fix: keep automation control legacy actions compatible

2afcb30

Co-Authored-By: Codex <noreply@openai.com>

fix: align defer with legacy pause control

9b84ef2

Co-Authored-By: Codex <noreply@openai.com>

fix: validate control mode and failure streak gaps

a349cf6

Co-Authored-By: Codex <noreply@openai.com>

fix: preserve low quota automation decisions

de0784d

Co-Authored-By: Codex <noreply@openai.com>

test: keep health decision review under gate

d9f0f7f

Co-Authored-By: Codex <noreply@openai.com>

fix: base automation control on recorded runs

09ca009

Co-Authored-By: Codex <noreply@openai.com>

fix: align retained failure boundaries

be99788

Co-Authored-By: Codex <noreply@openai.com>

fix: count authenticated workflow failures

863956d

Co-Authored-By: Codex <noreply@openai.com>

fix: align trusted workflow boundaries

a82d700

Co-Authored-By: Codex <noreply@openai.com>

fix: allow only known success boundaries

871a0fd

Co-Authored-By: Codex <noreply@openai.com>

fix: preserve runtime action compatibility

47998f1

Co-Authored-By: Codex <noreply@openai.com>

fix: tighten control action compatibility

b3730af

Co-Authored-By: Codex <noreply@openai.com>

fix: trust only service-owned safety boundaries

5abdbd1

Co-Authored-By: Codex <noreply@openai.com>

fix: validate explicit blank automation modes

dc47e3d

Co-Authored-By: Codex <noreply@openai.com>

fix: preserve stricter automation action

e74113f

Co-Authored-By: Codex <noreply@openai.com>

fix: keep automation control fields consistent

55f95f5

Co-Authored-By: Codex <noreply@openai.com>

fix: recover unknown automation history after boundary

531390f

Co-Authored-By: Codex <noreply@openai.com>

Pigbibi merged commit b85a0c3 into main Jul 7, 2026
4 checks passed

Pigbibi deleted the feat/health-driven-decision-p3 branch July 7, 2026 03:54

Uh oh!

Conversation

Pigbibi commented Jul 6, 2026

Summary

Test Plan

Safety

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

github-actions Bot commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🤖 Codex PR Review

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

This comment was marked as off-topic.

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

This comment was marked as off-topic.

Uh oh!

chatgpt-codex-connector Bot Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

This comment was marked as off-topic.

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 6, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

cursor Bot commented Jul 7, 2026

Uh oh!

github-actions Bot commented Jul 6, 2026 •

edited

Loading