docs(adr): define TEPP topic-lineage consumption boundary (TRSL-TM + CHRONOS/TDT) - #417
docs(adr): define TEPP topic-lineage consumption boundary (TRSL-TM + CHRONOS/TDT)#417seonghobae wants to merge 802 commits into
Conversation
…lineage' into codex/fix-oidc-adr-title
# Conflicts: # CHANGELOG.md # uv.lock
…codex/fix-image-adr-title
…odex/fix-project-action-parser # Conflicts: # docs/product-technical-gap-baseline.md
…lineage' into codex/fix-oidc-adr-title
…codex/fix-image-adr-title
…odex/fix-project-action-parser
…s' into HEAD # Conflicts: # frontend/src/PostBody.test.tsx
…ty' into codex/review-pr311
…-311' into codex/post-structure-case-fixes
… into codex/normalize-source-indent-semantics # Conflicts: # docs/product-technical-gap-baseline.md
…lineage' into codex/fix-oidc-adr-title
…fix/pr307-restack
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
|
Reviewed the remaining exact-head info thread. |
![opencode-agent[bot]](https://avatars.githubusercontent.com/in/1549082?s=60&v=4){kind=small}
There was a problem hiding this comment.
Pull request overview
OpenCode cannot approve yet because required coverage evidence did not pass.
Review outcome
1. HIGH .github/workflows/opencode-review.yml:1 - Coverage evidence did not prove required test/docstring evidence
-
Problem: The required coverage-evidence job result was
failure, so OpenCode cannot establish approval sufficiency for this head. -
Root cause: Automated approval is only valid when the same-head coverage-evidence job proves supported repository test suites passed and configured docstring gates passed or were advisory, or reports not applicable because no supported source files or package manifests exist. Missing, failed, skipped, unavailable, or unsupported-tooling test evidence is a blocker.
-
Fix: Install or configure the repository test/docstring evidence tooling when source files or package manifests exist, rerun the current-head coverage-evidence job, and approve only after it reports
successwith required evidence or explicit no-source not-applicable evidence. -
Regression test: Keep the approval branch checking
needs.coverage-evidence.result == successbefore posting APPROVE, and publish REQUEST_CHANGES when coverage-evidence blocker states such as cancelled, skipped, failed, unsupported-tooling, or below-100 evidence are present. -
Result: REQUEST_CHANGES
-
Reason: coverage-evidence result was
failure, so required test/docstring evidence was not proven for current head91b7e1cd914442e61e322620f194d1ba9da48567. -
Head SHA:
91b7e1cd914442e61e322620f194d1ba9da48567 -
Workflow run: 32628753453
-
Workflow attempt: 1
Coverage evidence
Coverage Decision
- Result: FAIL
- Test evidence: not proven passing
- Docstring evidence: not proven passing when configured
- Failure count: 4
Changed-File Evidence Map
flowchart LR
Evidence["OpenCode evidence"] --> Review["Current PR review path"]
Review --> Verify["Required checks"]
OpenCode Review Overview
Pull request overviewOpenCode cannot approve yet because required coverage evidence did not pass. Review outcome1. HIGH .github/workflows/opencode-review.yml:1 - Coverage evidence did not prove required test/docstring evidence
Coverage evidenceCoverage Decision
Changed-File Evidence Mapflowchart LR
Evidence["OpenCode evidence"] --> Review["Current PR review path"]
Review --> Verify["Required checks"]
|
ADR 0132 decision 5 commits the Event Lineage DAG's topic-thread overlay to a "distinct visual channel, not color alone" per TEPP ADR 0016's evidence / inference / prediction status. Implements that primitive now, ahead of the DAG wiring, so it is reviewable and Storybook-covered while TEPP issue #156's completed-result contract remains open: - EvidenceStatusMark: distinguishes status by label text and glyph shape (solid dot / diamond / triangle) in addition to color, satisfying WCAG 1.4.1 with redundant, testable channels. Presentational only -- never infers or invents a status; every call site must source it from a real TEPP envelope once one exists. - evidenceStatusI18n.ts: en/ko/zh/ja/vi copy, mirrors lineageDagI18n.ts. - Three new badge-status design tokens (light + dark), extending the existing ADR 0099 badge-token family and its dark-mode-completeness test rather than inventing a parallel pattern. - Unit tests assert the three statuses are never text-indistinguishable and that a prediction's copy never reads as confirmed fact. Storybook stories cover each status plus a side-by-side comparison. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Mk2aU7vk6Pn4pTnzcKT3Ws
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Stand up the Playwright harness (frontend/playwright.config.ts, npm run test:e2e; vitest now excludes e2e/) and ship ADR 0132 decision 5's fail-closed spec ahead of the TEPP wiring: log in through the compose stack's Keycloak realm as demo.admin, expand Advanced review tools, and assert the seeded Failed topic-lineage row shows the ADR 0014 'kind · status · entity' caption, the connect-a-topic-lineage-transport next action, the machine failure code on detail history only, and zero evidence/inference/prediction marks — a status mark without a TEPP envelope would be an invented status (TEPP ADR 0016). On a deployment holding real imported source data the spec skips itself with the reason named: ADR 0001/0042 deliberately hide the synthetic Demo Corp runs there (verified live: /api/analysis-runs returns [] for demo.admin on this machine's real-corpus DB). It runs fully against the seed-only stack make seed builds. The golden-path e2e half still waits on TEPP issue #156 publishing the topic-identity envelope. Validation: oxlint clean; tsc -b + vite build clean; vitest 205/205 (e2e excluded); playwright test runs the login + navigation walk live and self-skips at the documented boundary on this real-data stack. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Mk2aU7vk6Pn4pTnzcKT3Ws
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
The Unreleased Fixed entry said Global Ask "remains stateless" per ADR 0090, but this same PR ships persisted conversation history (migration 0105, backend/app/global_ask_history.py, ADR 0126, which supersedes that ADR 0090 consequence) -- the release notes stated the opposite of what the code does. Add the missing Added entry for the actual feature and narrow the Fixed entry to the true, unrelated fact it was describing (Customer Master fixtures don't need that migration for their own setup). Addresses Devin Review finding on CHANGELOG.md:70-72 (PR #417). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XKohNQEbNH1mpNEkvsuhKv
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
parse_summary_response returned the model's korean_summary verbatim while the plain-text path ran _formalize_korean_summary, so the same post could render informal -함/-음 endings depending on which transport the provider used. Apply the formalizer in the JSON path as well and extend it with generic 함->합니다 / 음->습니다 rules (specific 없음/있음/ 않음 rules still run first). Regression test covers both sentences through parse_summary_response.
| def chat_completion_content(body: object) -> str: | ||
| """Extract text from a provider chat-completion envelope safely. | ||
|
|
||
| Provider error bodies and malformed success bodies must never be echoed by | ||
| a consumer through ``KeyError`` or a repr of the response. The caller | ||
| receives only a stable validation error and can translate it at its own | ||
| product boundary. | ||
| """ | ||
| if not isinstance(body, dict): | ||
| raise TypeError("provider response was not an object") | ||
| choices = body.get("choices") | ||
| if not isinstance(choices, list) or not choices: | ||
| raise ValueError("provider response did not contain a choice") | ||
| first_choice = choices[0] | ||
| if not isinstance(first_choice, dict): | ||
| raise TypeError("provider response choice was not an object") | ||
| message = first_choice.get("message") | ||
| if not isinstance(message, dict): | ||
| raise TypeError("provider response message was not an object") | ||
| content = message.get("content") | ||
| if not isinstance(content, str) or not content.strip(): | ||
| raise TypeError("provider response did not contain text content") | ||
| return content |
There was a problem hiding this comment.
📝 Info: Empty provider content now raises instead of scoring zero
chat_completion_content in http_client.py raises TypeError on empty/whitespace content. Consumers such as adjudication_client.py previously read the field directly and returned 0.0 on an empty string via a regex miss. Now an empty completion fails closed at the provider boundary. All callers wrap this in try/except, so impact is limited, but it is a semantic change worth confirming against providers that can return empty content.
Was this helpful? React with 👍 or 👎 to provide feedback.
| def _topic_lineage_envelope_is_valid(envelope: dict[str, Any]) -> bool: | ||
| """Require TEPP's versioned topic-identity/CHRONOS-status contract (ADR 0132). | ||
|
|
||
| ``_tepp_submission`` only checks that ``result`` is *a* dict -- a | ||
| ``completed`` envelope carrying the calibrated-measurement shape (or any | ||
| other unrelated payload) would pass it too, since both requests share the | ||
| same wire contract and differ only in ``model_contract_version`` / | ||
| ``output_profile``. This additionally requires TRSL-TM topic identity and | ||
| CHRONOS/TDT status, keyed by envelope version. | ||
| """ | ||
| result = envelope.get("result") | ||
| if not isinstance(result, dict): | ||
| return False | ||
| if not isinstance(result.get("envelope_version"), int): | ||
| return False | ||
| topic_identity = result.get("topic_identity") | ||
| if not isinstance(topic_identity, (list, dict)) or not topic_identity: | ||
| return False | ||
| chronos_status = result.get("chronos_status") | ||
| if not isinstance(chronos_status, (list, dict, str)) or not chronos_status: | ||
| return False | ||
| return True | ||
|
|
||
|
|
||
| def topic_lineage_submit_outcome( | ||
| client: TeppClient, | ||
| request: AnalysisRunRequest, | ||
| ) -> tuple[str, str, dict[str, Any] | None]: | ||
| """Submit through ``tepp_client`` and require the topic-lineage contract. | ||
|
|
||
| Mirrors :func:`tepp_submit_outcome`, but a syntactically ``completed`` | ||
| envelope that omits the versioned topic-identity/CHRONOS-status contract | ||
| is also Failed (``tepp_topic_contract_unavailable``, ADR 0132 Decision | ||
| item 3), not silently persisted as a topic-lineage result. | ||
| """ | ||
| status_code, failure_code, envelope = _tepp_submission(client, request) | ||
| if status_code == _SUCCEEDED and not ( | ||
| envelope is not None and _topic_lineage_envelope_is_valid(envelope) | ||
| ): | ||
| return _FAILED, "tepp_topic_contract_unavailable", None | ||
| return status_code, failure_code, envelope |
There was a problem hiding this comment.
📝 Info: Valid topic envelope without a remote run id fails at persist
_topic_lineage_envelope_is_valid in analysis_run_start.py requires envelope_version/topic_identity/chronos_status but not analysis_run_id/run_id. Such an envelope is reported Succeeded by topic_lineage_submit_outcome, then _persist_topic_lineage_result rejects it for a missing remote id and the run becomes Failed/tepp_result_not_persisted. Behavior is fail-closed and safe; noted so the validation split is understood.
Was this helpful? React with 👍 or 👎 to provide feedback.
2 participants
Summary
zcrht811_export_rowsboard posts.analysis_run_topic_lineagekind (migration 0131) and result envelope table (migration 0132), wired throughtepp_clientexactly like the existing TEPP measurement path (ADR 0022) — same fail-closedtepp_not_available/tepp_result_not_persistedbehavior until TEPP ships the versioned result contract (tracked as TEPP issue fix(ui): never dump raw post bodies for unmatched images (v0.86.3) #156).make seednow also writes a Demo Corp topic-lineage run; verified end-to-end against the live dev Postgres (Pending → Running → Failed/tepp_not_available).AnalysisRunKindCodegains the new variant; every exhaustive TypeScriptnever-check switch and the start/retry UI copy is updated per kind.docs/product-technical-gap-baseline.md,CLAUDE.md, andCHANGELOG.mdto reflect this increment.Remaining for this feature: DAG topic-thread overlay UI, Storybook states per CHRONOS status, and Playwright e2e coverage — tracked in ADR 0132 and the gap baseline. Those depend partly on TEPP publishing its topic-identity/CHRONOS envelope shape.
Test plan
uv run pytest -q— 852 passed, 17 skipped (full backend suite, no regressions)pnpm test— 200 passed (full frontend suite, no regressions)pnpm run build— TypeScript build clean (exhaustiveness checks pass)pnpm run lint— clean_seed_demo_topic_lineage_runexercised directly against the live dev DB; confirmed Pending → Running → Failed/tepp_not_availablelifecycle and outbox rows🤖 Generated with Claude Code
https://claude.ai/code/session_01Mk2aU7vk6Pn4pTnzcKT3Ws