[vincentkoc]

Summary

Native Fusion benchmark failures lacked request correlation for provider investigation. Persist a unique OpenAI client request ID before dispatch and retain a bounded provider request ID before parsing the response, including HTTP and JSON failures. No prompts, completions, arbitrary headers or raw error bodies are added to receipts.

Verification

• Focused Node tests after rebase: 96 passed; 38 optional Docker integration cases skipped.
• Success, HTTP error, invalid JSON and transport failure fixtures verify saved IDs, unresolved reservations and no follow-on calls after failure.
• Provider tests cover unique IDs, unchanged request bodies, excluded malformed IDs and unchanged Anthropic requests.
• Independent P0–P2 review: no actionable findings; git diff --check passed.

Deployment and risk

Benchmark scripts only. No deployment, default routing, retry, scoring or budget-policy changes. Provider IDs outside the documented receipt allowlist are omitted. No paid model calls were made for this change.

Rebased onto the landed probe-qualified handoff and fresh Azure task. Both sets of regression tests are retained; the final rebased patch passed independent P0–P2 review. Fresh CI applies to the new head.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review complete

ClawSweeper finished reviewing this revision. The review result is being finalized.

View the workflow run.

[clawsweeper]

Codex review: needs maintainer review before merge.

What this changes

This PR records OpenAI client and provider request IDs in native pair and workload benchmark receipts.

Example: An OpenAI benchmark request returns HTTP 400 with x-request-id=req_test-123.

• Before: The receipt retains the failure and unresolved reservation without request correlation IDs.
• After: The receipt also retains its generated clientRequestId and providerRequestId=req_test-123.

Review scores

Measure	Result	What it means
Overall readiness	🦐 gold shrimp (3/6)	A focused, useful patch with no identified defects; confidence in execution rests on fixture coverage and successful script CI rather than live provider evidence.
Proof confidence	🦐 gold shrimp (3/6)	This benchmark-only diagnostic change has fixture coverage through the real pair runner with injected transport responses; it does not require shipped application proof, and no live-provider result is claimed.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Product

Kind: Feature · Worth it: Yes · Fix scope: Complete
User problem: Benchmark investigators cannot correlate failed native pair and workload calls with OpenAI provider records.
Reason: Bounded correlation metadata offers useful diagnostic value at low implementation cost, and the maintainer author already owns this feature area and its direction.

Merge readiness

✅ Ready for maintainer review

Keep open: this maintainer-authored PR adds useful benchmark diagnostics absent from current main, with no actionable correctness or security findings.

Priority: P3
Reviewed head: 08415c399976b2df50e98d0444f3e9d47596a6a9

Before merge

None.

Findings

None.

Agent review details

How this fits together

Opt-in benchmark runners transform model requests, reserve spending, call native provider APIs, and write private JSON receipts containing accounting and bounded diagnostic metadata.

flowchart LR
  A[Benchmark runner] --> B[Native request adapter]
  B --> C[Reserve and save client ID]
  C --> D[Provider API]
  D --> E[Capture bounded response ID]
  E --> F[Settle or retain reservation]
  F --> G[Private JSON receipt]

Loading

Technical review

Best possible solution:

Keep correlation in the existing native adapter and receipt ledgers while preserving current privacy, accounting, and stopping rules.

Do we have a high-confidence way to reproduce the issue?

Current-main source shows that native pair and workload calls neither send client correlation IDs nor retain provider response IDs; no target code was executed during this read-only review.

Is this the best way to solve the issue?

The implementation is focused: two shared helpers serve the existing dispatch owners, leave serialized bodies unchanged, and retain correlation without introducing another transport or accounting path.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning medium; reviewed against 43d978e80dfa.

Provenance checked

• Native provider diagnostics keeps the original intent (fix(fusion): retain safe structured provider diagnostics #938: Preserve useful structured diagnostic labels while excluding raw provider messages and arbitrary error fields.)
• Pre-dispatch pair receipts keeps the original intent (fix(fusion): retain failed request fingerprints #990: Retain exact request identity on HTTP and transport failures without persisting request content or weakening uncertain-charge accounting.)
• Workload reservations and terminal failures keeps the original intent (fix(fusion): retain bounded provider failure diagnostics #878: Retain bounded failure diagnostics while stopping campaigns and preserving complete unresolved reservations.)

Testing

Proof path: in-process harness.

Security

None.

Evidence

What I checked:

• Pinned introduced change: The introduced diff adds shared correlation helpers, wires both repository benchmark runners, and adds focused receipt and privacy coverage. (scripts/fusion-native-provider.mjs:43, 08415c399976)
• Current main lacks this capability: The fetched default branch has native request construction but no client-request-ID attachment or provider-response-ID extraction. (scripts/fusion-native-provider.mjs:21, 43d978e80dfa)
• Failure accounting remains intact: The pair runner saves the client ID before dispatch, captures response metadata before JSON parsing, and preserves its existing unknown-cost reservation and campaign stop on failure. (scripts/fusion-pair-benchmark.mjs:132, 08415c399976)
• Additive receipt compatibility: The persisted additions are optional clientRequestId and providerRequestId fields on pair.json calls and provenance.json attempts; runners create fresh output directories, and the existing pair analyzer validates accounting fields without rejecting additional call properties. (scripts/fusion-pair-analysis.mjs:86, 43d978e80dfa)
• Prior diagnostic contract: fix(fusion): retain failed request fingerprints #990 established pre-dispatch request fingerprints and fix(fusion): retain safe structured provider diagnostics #938 established bounded diagnostics; this PR preserves their request-body privacy and uncertain-charge behavior. (scripts/fusion-pair-benchmark.mjs:133, 1d13a6739afa)
• Existing gateway correlation is a separate boundary: feat: add canonical request correlation #92 and fix(proxy): preserve native error envelopes and request IDs #301 implement Worker ingress and proxy correlation; these benchmark runners call provider APIs directly, so that existing capability does not populate their receipts. (scripts/fusion-native-provider.mjs:25, 08415c399976)

Preflight source

Source unknown. Source comparison at the observed PR snapshot only. Completion is self-reported; body, policy, tests, execution, and report bytes are not verified.

Review metrics

Metric	Value	Why it matters
Change scope	Benchmark code +24/-2; tests +77/-1; documentation +11	The small implementation reuses existing owners; the larger test portion covers distinct success, HTTP, parsing, transport, and privacy cases.

Labels

Label changes:

• add P3: This improves investigation of opt-in benchmark failures without changing production routing or user workflows.
• add rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• add status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR.

Label justifications:

• P3: This improves investigation of opt-in benchmark failures without changing production routing or user workflows.
• rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🦐 gold shrimp and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR.

Rating scale

6/6 🦀 challenger crab · 5/6 🦞 diamond lobster · 4/6 🐚 platinum hermit · 3/6 🦐 gold shrimp · 2/6 🦪 silver shellfish · 1/6 🧂 unranked krab. Overall follows the weaker of proof and patch quality; ✨ marks media proof (a screenshot, video, or linked artifact) that directly shows the changed behavior.

Workflow

ClawSweeper edits this one comment on every review. Comment @clawsweeper re-review for a fresh review only. Repair and merge execution lanes are retired.

Reviewed November 13, 2026, 1:34 PM ET / 18:34 UTC.