Record Volcengine replay smoke readiness

[2002yy]

Scope

Synchronize the single project-status source after PR #89 merged.

Recorded facts

• replay-only Volcengine Ark adapter and bounded smoke/full workflow are in main
• PR Add Volcengine smoke and full Provider replay #89 merge SHA b1fac061f390cf224fe9288e7786474c3d1d1f6e
• final feature head 898cb0599e941830dd16e48185584719acceb7da
• CI #1696 passed all backend, RAG, static, security, mypy, frontend, browser and real-stack gates
• standard Ark /api/v3, exact Model/Endpoint ID and VOLCENGINE_API_KEY with ARK_API_KEY fallback are recorded
• Agent Plan and Coding Plan endpoints are explicitly rejected
• smoke runs one fixed answerable case; full runs all 10 cases plus offline AnswerClaim evaluation
• no real smoke or full run exists yet, so no real model-quality metric is claimed
• next operation is manual Volcengine smoke; only a green smoke unlocks two full runs
• production Volcengine chat Provider, claims, ChatTurn writes and executable-agent expansion remain frozen

Validation

Documentation-only synchronization. Keep Draft until the complete repository CI passes.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bc082c7231

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]

Do not claim smoke validates JSON and usage

When Ark returns malformed JSON or omits token usage, parse_provider_answer() records a parse_error fallback and run_provider_answer_replay() still reports status="completed"; incomplete usage is likewise represented only by usage.complete=false. The smoke workflow checks neither field before succeeding, so a green smoke can currently fail both guarantees documented here and still unlock the full runs described below. Either enforce parse_error == "" and complete usage in the smoke validation or describe these as manual inspection items rather than validated contract properties.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Describe the Ark endpoint as a default, not a restriction

When an operator supplies the workflow's optional base_url, VolcengineArkReplayProvider._validated_base_url() accepts any single-line URL except the two explicitly forbidden path substrings; it does not require the Ark host or /api/v3. Such a run is still labeled with provider_profile="volcengine", so the statement that this adapter uses only the standard endpoint overstates the recorded provenance boundary. Either enforce the claimed endpoint restriction or say that /api/v3 is merely the default.

Useful? React with 👍 / 👎.