fix(telemetry): export bounded request cohort identity

[seonghobae]

Scope

Issue #1110 needs comparable request cohorts before any observed-workload latency claim. The retained admin export had a validated decision duration and policy hash but omitted its known endpoint path and initial selection kind. This change exports only allowlisted values and keeps unrecognized stored strings null.

The selection kind is a hook identity; it does not yet distinguish the caller's route/conduct choice. The export remains local, incomplete, and unreconciled against ingress. This PR does not claim a p95 improvement, customer accuracy, or a released metric.

Evidence

• RED: projection regression failed with KeyError: 'endpoint_path' before the change.
• Native extension built from this worktree in its local .venv.
• uv run --no-sync python -m pytest -q -W error tests/test_paginated_decision_provenance.py tests/test_request_outcome_export.py: 63 passed, exit 0 at ff1242cc.
• Restarted HTTP export regression confirms the exact endpoint and selection kind; malformed stored text remains null.
• git diff --check: clean.

Follow-up boundary

An observed pilot must still reconcile all ingress, capture the caller's validated route/conduct choice, declare the workload and observation window, independently judge delivered answers, and calculate uncertainty before the #1110 or customer KPI acceptance claim. No routing, credential, security gate, or timeout behavior changes here.

Hosted review and queue evidence — 2026-09-28 05:18 UTC

At exact head ff1242cc4db1cd64ccb8d4603011bd401179ac67, the Strix run 36373616120 completed its scanner process with exit code 0; its run.json records status=completed, scan_completed=true, success=true, and SARIF has 0 results. The trusted central gate still returned failure because the report log contained strix.tools.web_search.tool: web_search invoked without PERPLEXITY_API_KEY configured. That optional-tool warning is not evidence that orchestrator/free exhausted. The owned narrow sanitizer repair already exists in ContextualWisdomLab/.github#2117; it is still Draft and has not been adopted by a terminal current-head run. Scanner completion and zero findings do not override this gate failure.

The Security “Tests and package quality” job 108774861880 remains queued with runner_id=0; 11 required jobs on this head are queued. The reusable central scan-pr-queue job 108774863086 is already labeled self-hosted, linux, x64 but has no assigned runner. No current-head hosted quality or independent review verdict follows from this queue observation. Preserve this head until a confirmed fix or review action calls for a push.

2026-09-28 Strix and queue RCA

Strix run 36373616120 ended with scanner rc=0, run.json completed/success, and SARIF 2.1.0 with zero results. Its exact optional web-search warning plus repeated PTY session-count threshold notices caused the central gate to label the completed scan STRIX_PROVIDER_UNAVAILABLE. The optional warning sanitizer was already proposed in central #2117; new commit fb4b3338 narrows the PTY threshold notice as well. Replaying this run's report and console through the current proposed classifier changes the false console signal from 1 to 0; 49 adjacent Strix tests pass locally. Central #2117 remains Draft with current-head checks queued and independent review outstanding, so this parent PR's hosted Strix failure is not resolved yet.

At 2026-09-28 05:47 UTC, the org API listed six self-hosted Linux/X64 runners online and busy. This PR's queued scan-pr-queue job requires self-hosted,linux,x64, while queued Tests and package quality requires ubuntu-24.04; both still had runner_id=0. These observations show backlog and no assigned runner for those jobs, not a source failure or proof of the sole scheduling cause. No parent HEAD change or gate bypass follows.

[coderabbitai]

Warning

Review limit reached

Next included review available in 51 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: f1c5dd09-6363-41dd-8689-384d774a20b8

📥 Commits

Reviewing files that changed from the base of the PR and between 8e1f1a8 and ff1242c.

📒 Files selected for processing (4)

• contextual_orchestrator/orchestrator.py
• docs/doctoring/request_outcome_export_validation.md
• tests/test_paginated_decision_provenance.py
• tests/test_request_outcome_export.py

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

Hosted Strix RCA for run 36373616120, artifact 10951314362: the sidecar preflight completed (chat/completions HTTP 200, 4 ready candidates). Strix itself exited 0 and wrote run.json with status=completed, scan_completed=true, success=true, and 466 model requests. The only report-log warning was the optional web_search call without PERPLEXITY_API_KEY; console also printed PTY session-count threshold notices. The central gate matched these broad WARNING strings and labeled the run STRIX_PROVIDER_UNAVAILABLE: contextual-orchestrator/orchestrator/free exhausted, without evidence of gateway exhaustion in this completed scan.

Central fix: ContextualWisdomLab/.github#2486 recognizes only those exact known notices, preserving unknown/provider warnings as failures. This does not retroactively turn this run into a passing protected check. A new current-head hosted verdict is still required after the central fix lands.

[opencode-agent]

Scheduled review-feedback autofix for this PR head.

• Head SHA: ff1242cc4db1cd64ccb8d4603011bd401179ac67

[seonghobae]

Exact-head admission correction — ff1242cc4db1cd64ccb8d4603011bd401179ac67

Ready is review admission only. Fresh audit against base 8e1f1a8bf3e96e56dc8fcc90ec777883a1d56ce6 found:

• latest terminal workflow blockers: Security and Quality 36373616066=cancelled, CodeQL PR 36373616019=failure

This PR is moved to Draft/Proposed until the causal owner repair is present on a successor exact head and re-audited. Queued/pending work is neither an additional blocker nor passing evidence. No Close, force push, destructive rebase, manual rerun, synthetic status/approval, merge, auto-merge, or bypass was performed.

[opencode-agent]

Scheduled review-feedback autofix for this PR head.

• Head SHA: ff1242cc4db1cd64ccb8d4603011bd401179ac67