歸檔 canonical batch 並新增 enumeration 優化切片 - #35
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThis PR adds per-fixture timeouts (multiprocessing) and timed_out reporting to batch verification, records converter per-phase timings/progress, exposes a review-viewer handoff in the UI, archives canonical batch evidence, and introduces an OpenSpec change to optimize source entity enumeration. ChangesTimeout-aware batch verification with phase tracking
Archived evidence and roadmap alignment
Sequence Diagram(s)sequenceDiagram
participant CLI
participant run_storage_batch_verification
participant _run_single_fixture_with_timeout
participant Process
participant _run_single_fixture
participant Converter
participant job_phase_progress_path
CLI->>run_storage_batch_verification: invoke (timeout_seconds)
run_storage_batch_verification->>_run_single_fixture_with_timeout: start fixture
_run_single_fixture_with_timeout->>Process: spawn (_run_single_fixture_process)
Process->>_run_single_fixture: execute intake + conversion
_run_single_fixture->>Converter: adapter.convert(converter_job)
Converter->>job_phase_progress_path: write phase progress JSON
_run_single_fixture->>_run_single_fixture_with_timeout: publish result via queue
_run_single_fixture_with_timeout->>run_storage_batch_verification: return result (passed/failed/timed_out)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
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. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (6)
openspec/changes/optimize-worker-source-entity-enumeration/specs/worker-artifact-pipeline/spec.md (1)
1-32:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winConsider translating spec content to Traditional Chinese.
The coding guidelines state: "All OpenSpec artifacts must use Traditional Chinese (繁體中文); preserve original text for API paths, schema fields, CLI flags, status enums, logs/errors, external product names, and OpenSpec parser required headers."
While technical identifiers (
source_entity_enumeration,ifc_entity_key, etc.) are correctly preserved in English, the descriptive requirement text and scenario prose should use Traditional Chinese for consistency with other OpenSpec artifacts in this repository. As per coding guidelines, OpenSpec artifacts should use Traditional Chinese (繁體中文).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@openspec/changes/optimize-worker-source-entity-enumeration/specs/worker-artifact-pipeline/spec.md` around lines 1 - 32, The spec's descriptive and scenario prose must be translated from English into Traditional Chinese while preserving all technical identifiers and API/schema/CLI tokens (e.g., source_entity_enumeration, ifc_entity_key, ifc_entity_id, ifc_class, ifc_guid, name, coverage_denominator, source_ifc_entity_count, minimum_coverage_locked) and keeping parser-required headers and any exact flags/enums/log text unchanged; update the body text of each Requirement and Scenario to 繁體中文, ensuring diagnostics/field names remain in English where specified and that the resulting document remains semantically identical.openspec/changes/optimize-worker-source-entity-enumeration/README.md (1)
1-4:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winConsider translating README content to Traditional Chinese.
As per coding guidelines, "All OpenSpec artifacts must use Traditional Chinese (繁體中文); preserve original text for API paths, schema fields, CLI flags, status enums, logs/errors, external product names, and OpenSpec parser required headers."
This README should use Traditional Chinese for the descriptive content while preserving technical identifiers like
_worker,IfcOpenShell,source_entity_enumerationin English.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@openspec/changes/optimize-worker-source-entity-enumeration/README.md` around lines 1 - 4, The README currently uses English for descriptive content but must be translated to Traditional Chinese (繁體中文) per guidelines while preserving technical identifiers; update the README.md text (but not the identifiers `_worker`, `IfcOpenShell`, `source_entity_enumeration`, "canonical 89MB IFC", API/CLI/schema names, logs/errors) by replacing the English description with a clear Traditional Chinese translation that retains those technical terms verbatim._worker/app/converters.py (1)
160-167:⚠️ Potential issue | 🟠 Major | ⚡ Quick winAttach GitNexus impact/detect-changes evidence for modified converter symbols before merge.
This PR modifies
IfcOpenShellUsdConverter.convertand related flow, but the PR notes indicate detect-changes was not refreshed. Please include the final GitNexus impact + detect-changes output for these symbols in this PR.As per coding guidelines, "
**/*.{ts,tsx,js,jsx,py}: Before modifying a function, class, or method, run GitNexus impact analysis... MUST run gitnexus_detect_changes() before committing."🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@_worker/app/converters.py` around lines 160 - 167, The PR changed the converter flow (notably IfcOpenShellUsdConverter.convert and its related converter symbols) but did not include the required GitNexus impact and detect-changes evidence; run gitnexus_detect_changes() locally (or via your CI step) targeting the modified symbols (IfcOpenShellUsdConverter.convert and any helper methods/classes you modified), capture the full GitNexus impact + detect-changes output, and attach those results to this PR (update the PR description or add a file) so the change log and impact analysis are recorded before merge.openspec/changes/optimize-worker-source-entity-enumeration/specs/runtime-verification-evidence/spec.md (1)
1-26:⚠️ Potential issue | 🟠 Major | ⚡ Quick winOpenSpec delta spec 內容語言需改為繁體中文。
目前新增規格主體為英文，請轉為繁體中文撰寫；僅保留規範允許維持原文的 API/欄位/CLI/status 等 token。
As per coding guidelines, "
openspec/**/*.md: All OpenSpec artifacts must use Traditional Chinese (繁體中文); preserve original text for API paths, schema fields, CLI flags, status enums, logs/errors, external product names, and OpenSpec parser required headers."🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@openspec/changes/optimize-worker-source-entity-enumeration/specs/runtime-verification-evidence/spec.md` around lines 1 - 26, 將此新增的 OpenSpec 規格主體（包括「Requirement: Source entity enumeration optimization evidence」段落及其所有情境描述）從英文完整翻譯為繁體中文，僅保留原文的 API/欄位/CLI/status 等 token 不翻譯，例如保留 source_entity_enumeration、canonical fixture identity、conversion_job_id、artifact_group_id、USDC、mapping artifact ID、minimum_coverage_locked、_worker、timed_out、blocked 及 CLI 範例 `--limit 1 --timeout-seconds 600` 等文字；請在翻譯中保留原有段落結構、標題層級和專有符號（如 `Requirement:`、`Scenario:`）不變，並確保所有描述性文字改為繁體中文以符合 openspec/ 檔案語言規範。_worker/app/ui.py (1)
434-459:⚠️ Potential issue | 🟠 Major | ⚡ Quick winReset review handoff state to avoid stale links across runs.
Line 507 only updates links in the
usdc_urlpath. If a previous conversion succeeded, a later failed/partial run can keep oldnextStep/reviewPreviewtargets enabled.Suggested patch
async function startConversion() { if (!selected) return; clearTimeout(pollTimer); + nextStep.href = "http://127.0.0.1:8004"; + nextStep.setAttribute("aria-disabled", "true"); + reviewPreview.href = "http://127.0.0.1:8004"; + reviewPreview.setAttribute("aria-disabled", "true"); + usdcUrlEl.textContent = "—"; + reviewHandoffEl.textContent = "—"; setStatus(jobStatus, "warn", "建立 job"); try { @@ function configureReviewHandoff(result, artifactGroupId) { @@ if (result.usdc_url) { nextStep.href = target; nextStep.setAttribute("aria-disabled", "false"); reviewPreview.href = target; reviewPreview.setAttribute("aria-disabled", "false"); + } else { + nextStep.href = "http://127.0.0.1:8004"; + nextStep.setAttribute("aria-disabled", "true"); + reviewPreview.href = "http://127.0.0.1:8004"; + reviewPreview.setAttribute("aria-disabled", "true"); } }Also applies to: 493-513
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@_worker/app/ui.py` around lines 434 - 459, Reset the previous run's handoff/review links and IDs at the start (and on error) of a new conversion so stale nextStep/reviewPreview targets are not left enabled; specifically, in startConversion() clear or disable the review link elements and reset jobIdEl and artifactGroupEl to placeholders before the fetch, and also ensure the catch block clears/disables those same review/nextStep UI targets when a conversion fails; reference the startConversion, jobIdEl, artifactGroupEl, and pollConversion call sites to locate where to clear/reset the review handoff state.openspec/changes/optimize-worker-source-entity-enumeration/design.md (1)
1-51:⚠️ Potential issue | 🟠 Major | ⚡ Quick win請將此 OpenSpec 設計文件改為繁體中文（保留規定英文 token）
這份新增
design.md目前大部分為英文敘述，與 OpenSpec 文件語言規範不一致，後續維運與審核可追溯性會受影響。As per coding guidelines, "All OpenSpec artifacts must use Traditional Chinese (繁體中文); preserve original text for API paths, schema fields, CLI flags, status enums, logs/errors, external product names, and OpenSpec parser required headers."
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@openspec/changes/optimize-worker-source-entity-enumeration/design.md` around lines 1 - 51, 將此 OpenSpec 設計文件從英文翻譯為繁體中文（保留規定英文 token 不變）: 翻譯所有段落標題與敘述（例如 "Context", "Goals / Non-Goals", "Decisions", 各條列內容與 "Risks / Trade-offs" 等）與細節為繁體中文，但嚴格保留並不翻譯 API paths, schema fields, CLI flags, status enums, logs/errors, external product names, 以及 OpenSpec parser required headers（保持其原始英文字串不變）；同時確保術語一致且原意不變、維持原有段落結構與所有英文 token 的精確大小寫，並在完成後執行快速對照檢查以確認沒有意外改動到保留的英文標記（例如保留 "source_entity_enumeration", "_worker", "--limit 1", "minimum_coverage_locked=true" 等字串）。
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@_worker/app/batch_verification.py`:
- Around line 141-154: The timeout cleanup currently calls process.terminate()
and process.join(5) but doesn't hard-kill if the child ignores SIGTERM; add a
fallback that checks process.is_alive() after process.join(5) and, if still
alive, call process.kill() (SIGKILL) and then process.join() again (with a short
timeout) to ensure the subprocess is reaped before building the _timeout_record
and calling _mark_job_timed_out; keep this logic inside the same timeout branch
where process.terminate() is invoked and handle any exceptions from kill/join so
the function still returns the record.
In `@_worker/app/converters.py`:
- Around line 135-155: The _record_phase_progress function currently writes
diagnostics via _write_json and will raise if the progress_path is invalid or
unwritable; make this persistence best-effort by wrapping the call to
_write_json (using the progress_path variable) in a try/except that catches
OSError/IOError and Exception from serialization, and suppresses the exception
(optionally logging the error) so that failures to write phase progress do not
propagate and fail the conversion job; ensure the function still returns None on
write failure and does not re-raise.
In `@_worker/scripts/verify_storage_batch.py`:
- Around line 20-33: Validate that the --timeout-seconds value is strictly
positive before calling run_storage_batch_verification: update the
parser.add_argument for "--timeout-seconds" or add a post-parse check that
inspects args.timeout_seconds and raises/prints an error and exits if the value
is <= 0; ensure the check runs before invoking
run_storage_batch_verification(Settings.from_env(), ...) so that invalid inputs
are rejected at parse time with a clear message.
In `@openspec/changes/optimize-worker-source-entity-enumeration/proposal.md`:
- Around line 1-31: 將整個 proposal.md 內容改寫為繁體中文（包含段落說明與敘述），但保留所有不能翻譯的原文 token：如
`_worker`、`worker-artifact-pipeline`、`runtime-verification-evidence`、`_worker/app/converters.py`、`_worker/app/batch_verification.py`、`_worker/tests/*`、`scripts/verify_storage_batch.py`、CLI
flag、schema 欄位、API 路徑、狀態 enum、logs/errors、外部產品名稱及 OpenSpec 必要的 header 不變；確保標題（例如
"Why", "What Changes", "Capabilities", "Impact"）或其翻譯仍符合 OpenSpec parser
要求並保留任何必需的原文標記與示例，以免破壞機器解析或識別欄位。
---
Outside diff comments:
In `@_worker/app/converters.py`:
- Around line 160-167: The PR changed the converter flow (notably
IfcOpenShellUsdConverter.convert and its related converter symbols) but did not
include the required GitNexus impact and detect-changes evidence; run
gitnexus_detect_changes() locally (or via your CI step) targeting the modified
symbols (IfcOpenShellUsdConverter.convert and any helper methods/classes you
modified), capture the full GitNexus impact + detect-changes output, and attach
those results to this PR (update the PR description or add a file) so the change
log and impact analysis are recorded before merge.
In `@_worker/app/ui.py`:
- Around line 434-459: Reset the previous run's handoff/review links and IDs at
the start (and on error) of a new conversion so stale nextStep/reviewPreview
targets are not left enabled; specifically, in startConversion() clear or
disable the review link elements and reset jobIdEl and artifactGroupEl to
placeholders before the fetch, and also ensure the catch block clears/disables
those same review/nextStep UI targets when a conversion fails; reference the
startConversion, jobIdEl, artifactGroupEl, and pollConversion call sites to
locate where to clear/reset the review handoff state.
In `@openspec/changes/optimize-worker-source-entity-enumeration/design.md`:
- Around line 1-51: 將此 OpenSpec 設計文件從英文翻譯為繁體中文（保留規定英文 token 不變）: 翻譯所有段落標題與敘述（例如
"Context", "Goals / Non-Goals", "Decisions", 各條列內容與 "Risks / Trade-offs"
等）與細節為繁體中文，但嚴格保留並不翻譯 API paths, schema fields, CLI flags, status enums,
logs/errors, external product names, 以及 OpenSpec parser required
headers（保持其原始英文字串不變）；同時確保術語一致且原意不變、維持原有段落結構與所有英文 token
的精確大小寫，並在完成後執行快速對照檢查以確認沒有意外改動到保留的英文標記（例如保留 "source_entity_enumeration",
"_worker", "--limit 1", "minimum_coverage_locked=true" 等字串）。
In `@openspec/changes/optimize-worker-source-entity-enumeration/README.md`:
- Around line 1-4: The README currently uses English for descriptive content but
must be translated to Traditional Chinese (繁體中文) per guidelines while preserving
technical identifiers; update the README.md text (but not the identifiers
`_worker`, `IfcOpenShell`, `source_entity_enumeration`, "canonical 89MB IFC",
API/CLI/schema names, logs/errors) by replacing the English description with a
clear Traditional Chinese translation that retains those technical terms
verbatim.
In
`@openspec/changes/optimize-worker-source-entity-enumeration/specs/runtime-verification-evidence/spec.md`:
- Around line 1-26: 將此新增的 OpenSpec 規格主體（包括「Requirement: Source entity
enumeration optimization evidence」段落及其所有情境描述）從英文完整翻譯為繁體中文，僅保留原文的
API/欄位/CLI/status 等 token 不翻譯，例如保留 source_entity_enumeration、canonical fixture
identity、conversion_job_id、artifact_group_id、USDC、mapping artifact
ID、minimum_coverage_locked、_worker、timed_out、blocked 及 CLI 範例 `--limit 1
--timeout-seconds 600` 等文字；請在翻譯中保留原有段落結構、標題層級和專有符號（如
`Requirement:`、`Scenario:`）不變，並確保所有描述性文字改為繁體中文以符合 openspec/ 檔案語言規範。
In
`@openspec/changes/optimize-worker-source-entity-enumeration/specs/worker-artifact-pipeline/spec.md`:
- Around line 1-32: The spec's descriptive and scenario prose must be translated
from English into Traditional Chinese while preserving all technical identifiers
and API/schema/CLI tokens (e.g., source_entity_enumeration, ifc_entity_key,
ifc_entity_id, ifc_class, ifc_guid, name, coverage_denominator,
source_ifc_entity_count, minimum_coverage_locked) and keeping parser-required
headers and any exact flags/enums/log text unchanged; update the body text of
each Requirement and Scenario to 繁體中文, ensuring diagnostics/field names remain
in English where specified and that the resulting document remains semantically
identical.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: b4fd1315-2509-461b-a564-5aba80e6c0a7
📒 Files selected for processing (29)
_worker/app/batch_verification.py_worker/app/converters.py_worker/app/store.py_worker/app/ui.py_worker/scripts/verify_storage_batch.py_worker/tests/test_worker_api.py_worker/tests/test_worker_batch_verification.py_worker/tests/test_worker_converters.py_worker/tests/test_worker_store.pydocs/plans/AI-BIM-governance-saas-roadmap-2026-05.htmldocs/plans/AI-BIM-governance-saas-roadmap-2026-05.mddocs/verification/2026-05-12-worker-canonical-storage-batch-baseline.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/.openspec.yamlopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/design.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/proposal.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/runtime-verification-evidence/spec.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/worker-artifact-pipeline/spec.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/worker-demo-upload-convert-ui/spec.mdopenspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/tasks.mdopenspec/changes/optimize-worker-source-entity-enumeration/.openspec.yamlopenspec/changes/optimize-worker-source-entity-enumeration/README.mdopenspec/changes/optimize-worker-source-entity-enumeration/design.mdopenspec/changes/optimize-worker-source-entity-enumeration/proposal.mdopenspec/changes/optimize-worker-source-entity-enumeration/specs/runtime-verification-evidence/spec.mdopenspec/changes/optimize-worker-source-entity-enumeration/specs/worker-artifact-pipeline/spec.mdopenspec/changes/optimize-worker-source-entity-enumeration/tasks.mdopenspec/specs/runtime-verification-evidence/spec.mdopenspec/specs/worker-artifact-pipeline/spec.mdopenspec/specs/worker-demo-upload-convert-ui/spec.md
| process.terminate() | ||
| process.join(5) | ||
| elapsed = perf_counter() - started | ||
| partial = _last_progress(messages) | ||
| phase_progress = _read_phase_progress(settings, partial) | ||
| record = _timeout_record( | ||
| source_item, | ||
| timeout_seconds=timeout_seconds, | ||
| elapsed_seconds=elapsed, | ||
| partial=partial, | ||
| phase_progress=phase_progress, | ||
| ) | ||
| _mark_job_timed_out(settings, record) | ||
| return record |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify timeout cleanup path around terminate/join lacks hard-kill fallback.
rg -n -C4 'process\.terminate\(\)|process\.join\(5\)|process\.is_alive\(\)' _worker/app/batch_verification.pyRepository: monkey1sai/AI-BIM-governance
Length of output: 498
🏁 Script executed:
#!/bin/bash
# Get complete context of the function containing the timeout cleanup
sed -n '130,160p' _worker/app/batch_verification.pyRepository: monkey1sai/AI-BIM-governance
Length of output: 1028
🏁 Script executed:
#!/bin/bash
# Check if there are any tests for the timeout path
find . -name "*.py" -path "*/test*" -o -name "*test*.py" | head -20Repository: monkey1sai/AI-BIM-governance
Length of output: 1745
🌐 Web query:
Python multiprocessing Process.kill() method documentation and usage
💡 Result:
In Python’s multiprocessing, multiprocessing.Process.kill immediately terminates the underlying child process by sending SIGKILL on POSIX (and an equivalent hard-kill on non-POSIX via the platform’s termination API). It is documented as “Same as terminate but using the SIGKILL signal on POSIX.” [1][2] Signature / where it lives - Method: multiprocessing.Process.kill(self) - Also present in the CPython implementation as Process.kill, which delegates to the underlying _popen.kill. [3] How to use it (typical pattern) 1) Create and start a multiprocessing.Process. 2) When you need to stop it urgently, call p.kill. [1][2] 3) Usually follow with p.join so the parent waits for the process to fully exit (and to avoid leaving a zombie on Unix). (join is the documented way to wait for the child process to terminate.) [4][3] Important behavioral notes / warnings - kill is “terminate with SIGKILL” (hard kill). [1][2] - If you kill a process while it is trying to use a multiprocessing.Queue, the queue data may become corrupted, potentially causing exceptions in other processes that later use the queue. [1][2] Relationship to terminate and join - terminate: a “graceful-ish” termination request compared to kill; on POSIX it uses SIGTERM and on Windows uses TerminateProcess. [5] - kill: like terminate but escalates to SIGKILL on POSIX (hard kill). [1][2] - join: does not kill; it blocks the calling thread until the process terminates (or a timeout occurs). [4] Example import multiprocessing as mp import time def worker: while True: time.sleep(1) if name == "main": p = mp.Process(target=worker) p.start time.sleep(2) p.kill # hard kill (SIGKILL on POSIX) p.join # wait for it to exit
Citations:
- 1: https://docs.python.org/3/library/multiprocessing.html?highlight=current_processing
- 2: https://docs.python.org/3/library/multiprocessing.html
- 3: https://github.com/python/cpython/blob/main/Lib/multiprocessing/process.py
- 4: https://stackoverflow.com/questions/19699802/difference-between-join-and-terminate
- 5: https://docs.python.org/3.12/library/multiprocessing.html
🏁 Script executed:
#!/bin/bash
# Check the timeout test coverage in test_worker_batch_verification.py
cat ./_worker/tests/test_worker_batch_verification.py | grep -n -A 20 "timeout\|is_alive\|terminate"Repository: monkey1sai/AI-BIM-governance
Length of output: 2693
🏁 Script executed:
#!/bin/bash
# Check if there are other process management patterns in the codebase
rg -n 'process\.(terminate|kill|is_alive)' _worker/ --type pyRepository: monkey1sai/AI-BIM-governance
Length of output: 203
Add hard-kill fallback after timeout terminate to prevent orphan processes
The timeout cleanup path calls process.terminate() then process.join(5), but lacks a fallback if the process doesn't respond to SIGTERM. If the subprocess ignores the graceful termination signal, it remains alive after the 5-second timeout, leaving orphaned processes that accumulate across batch runs and compromise stability.
Add an is_alive() check after the join timeout with a kill() fallback:
Proposed fix
if process.is_alive():
process.terminate()
process.join(5)
+ if process.is_alive():
+ process.kill()
+ process.join(5)
elapsed = perf_counter() - startedThe queue is already drained before this block, so the hard kill is safe from queue corruption. This follows Python multiprocessing best practices: graceful termination (SIGTERM) → timeout → hard kill (SIGKILL) → await cleanup.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@_worker/app/batch_verification.py` around lines 141 - 154, The timeout
cleanup currently calls process.terminate() and process.join(5) but doesn't
hard-kill if the child ignores SIGTERM; add a fallback that checks
process.is_alive() after process.join(5) and, if still alive, call
process.kill() (SIGKILL) and then process.join() again (with a short timeout) to
ensure the subprocess is reaped before building the _timeout_record and calling
_mark_job_timed_out; keep this logic inside the same timeout branch where
process.terminate() is invoked and handle any exceptions from kill/join so the
function still returns the record.
| def _record_phase_progress( | ||
| job: Mapping[str, Any], | ||
| phase: str, | ||
| phase_timings: dict[str, dict[str, Any]], | ||
| *, | ||
| status: str = "running", | ||
| ) -> None: | ||
| progress_path = job.get("phase_progress_path") | ||
| if not progress_path: | ||
| return | ||
| _write_json( | ||
| Path(str(progress_path)), | ||
| { | ||
| "conversion_job_id": job.get("conversion_job_id"), | ||
| "source_artifact_id": job.get("source_artifact_id"), | ||
| "current_phase": phase, | ||
| "status": status, | ||
| "phase_timings": phase_timings, | ||
| }, | ||
| ) | ||
|
|
There was a problem hiding this comment.
Make phase-progress persistence best-effort to avoid false conversion failures.
If phase_progress_path is invalid/unwritable, Line 145 currently raises and fails the whole conversion. Diagnostics I/O should not become a hard conversion gate.
Suggested patch
def _record_phase_progress(
@@
) -> None:
@@
- _write_json(
- Path(str(progress_path)),
- {
- "conversion_job_id": job.get("conversion_job_id"),
- "source_artifact_id": job.get("source_artifact_id"),
- "current_phase": phase,
- "status": status,
- "phase_timings": phase_timings,
- },
- )
+ try:
+ _write_json(
+ Path(str(progress_path)),
+ {
+ "conversion_job_id": job.get("conversion_job_id"),
+ "source_artifact_id": job.get("source_artifact_id"),
+ "current_phase": phase,
+ "status": status,
+ "phase_timings": phase_timings,
+ },
+ )
+ except Exception:
+ # best-effort diagnostics; must not fail conversion path
+ return🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@_worker/app/converters.py` around lines 135 - 155, The _record_phase_progress
function currently writes diagnostics via _write_json and will raise if the
progress_path is invalid or unwritable; make this persistence best-effort by
wrapping the call to _write_json (using the progress_path variable) in a
try/except that catches OSError/IOError and Exception from serialization, and
suppresses the exception (optionally logging the error) so that failures to
write phase progress do not propagate and fail the conversion job; ensure the
function still returns None on write failure and does not re-raise.
| parser.add_argument( | ||
| "--timeout-seconds", | ||
| type=float, | ||
| default=600.0, | ||
| help="Per-fixture timeout for real conversion runs.", | ||
| ) | ||
| args = parser.parse_args() | ||
|
|
||
| payload = run_storage_batch_verification(Settings.from_env(), limit=args.limit, dry_run=args.dry_run) | ||
| payload = run_storage_batch_verification( | ||
| Settings.from_env(), | ||
| limit=args.limit, | ||
| dry_run=args.dry_run, | ||
| timeout_seconds=None if args.dry_run else args.timeout_seconds, | ||
| ) |
There was a problem hiding this comment.
Validate --timeout-seconds is strictly positive.
Line 20 currently accepts 0/negative values, which can lead to immediate timeout behavior and misleading diagnostics. Reject invalid values at parse time.
Suggested patch
args = parser.parse_args()
+ if not args.dry_run and args.timeout_seconds <= 0:
+ parser.error("--timeout-seconds must be > 0 for real conversion runs.")
payload = run_storage_batch_verification(
Settings.from_env(),🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@_worker/scripts/verify_storage_batch.py` around lines 20 - 33, Validate that
the --timeout-seconds value is strictly positive before calling
run_storage_batch_verification: update the parser.add_argument for
"--timeout-seconds" or add a post-parse check that inspects args.timeout_seconds
and raises/prints an error and exits if the value is <= 0; ensure the check runs
before invoking run_storage_batch_verification(Settings.from_env(), ...) so that
invalid inputs are rejected at parse time with a clear message.
| ## Why | ||
|
|
||
| Canonical `storage/*.ifc` batch verification currently times out on the first 89MB fixture after `ifc_open`, with the last known phase stuck at `source_entity_enumeration`. This blocks single-fixture evidence, visual preview, and the full 13-file baseline lock even though timeout diagnostics are now recorded. | ||
|
|
||
| ## What Changes | ||
|
|
||
| - Optimize `_worker` IFC source entity enumeration so canonical fixtures can progress past the current bottleneck without weakening all-IFC-entity coverage semantics. | ||
| - Add measurable before/after timing evidence for `source_entity_enumeration`, including entity counts, elapsed duration, and timeout/budget diagnostics. | ||
| - Preserve existing artifact intake, conversion job, lineage, mapping, and review viewer handoff contracts. | ||
| - Keep `minimum_coverage_locked=false` unless the full canonical batch still satisfies the archived baseline requirements. | ||
| - Do not introduce viewer, coordinator, Kit runtime, WebRTC, GPU, auth, session lifecycle, or production batch-job responsibilities into `_worker`. | ||
|
|
||
| ## Capabilities | ||
|
|
||
| ### New Capabilities | ||
|
|
||
| - None. | ||
|
|
||
| ### Modified Capabilities | ||
|
|
||
| - `worker-artifact-pipeline`: require `_worker` source entity enumeration to be profiled and optimized for canonical IFC fixtures while preserving all-entity coverage and stable IFC-to-USD traceability. | ||
| - `runtime-verification-evidence`: require optimization evidence to record before/after source entity enumeration timing and the canonical single-fixture rerun result. | ||
|
|
||
| ## Impact | ||
|
|
||
| - Owner: `_worker`. | ||
| - Likely code paths: `_worker/app/converters.py`, `_worker/app/batch_verification.py`, and focused `_worker/tests/*`. | ||
| - Data structures: conversion quality metrics and batch verification evidence may gain source entity enumeration diagnostics; existing fields must remain backward-compatible. | ||
| - CLI: `scripts/verify_storage_batch.py` may be used for validation; no required CLI breaking change is expected. | ||
| - Dependencies: no new production dependency unless a measured standard-library or existing IfcOpenShell/USD path cannot solve the bottleneck. | ||
| - Runtime boundary: visual preview remains outside `_worker` and must continue through `bim-review-coordinator`, `web-viewer-sample`, and `bim-streaming-server` after conversion succeeds. |
There was a problem hiding this comment.
OpenSpec proposal內容語言需改為繁體中文。
目前新檔案主體為英文，與 OpenSpec markdown 語言規範不一致；請改為繁體中文，僅保留規範允許的 API/欄位/CLI 等原文 token。
As per coding guidelines, "openspec/**/*.md: All OpenSpec artifacts must use Traditional Chinese (繁體中文); preserve original text for API paths, schema fields, CLI flags, status enums, logs/errors, external product names, and OpenSpec parser required headers."
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@openspec/changes/optimize-worker-source-entity-enumeration/proposal.md`
around lines 1 - 31, 將整個 proposal.md 內容改寫為繁體中文（包含段落說明與敘述），但保留所有不能翻譯的原文 token：如
`_worker`、`worker-artifact-pipeline`、`runtime-verification-evidence`、`_worker/app/converters.py`、`_worker/app/batch_verification.py`、`_worker/tests/*`、`scripts/verify_storage_batch.py`、CLI
flag、schema 欄位、API 路徑、狀態 enum、logs/errors、外部產品名稱及 OpenSpec 必要的 header 不變；確保標題（例如
"Why", "What Changes", "Capabilities", "Impact"）或其翻譯仍符合 OpenSpec parser
要求並保留任何必需的原文標記與示例，以免破壞機器解析或識別欄位。
There was a problem hiding this comment.
Pull request overview
This PR archives the previously active canonical storage batch baseline OpenSpec change (while preserving its “runtime still blocked” status), folds the accepted requirements into the current specs, and introduces a new focused OpenSpec change to optimize _worker’s source_entity_enumeration phase (the current timeout blocker on the first 89MB canonical IFC fixture). In parallel, it updates _worker’s batch verification helper and UI handoff to capture richer timeout/phase diagnostics and enable review-viewer handoff without rendering USDC inside _worker.
Changes:
- Archive
worker-canonical-storage-batch-baseline, sync its key requirements into current OpenSpec specs, and add a detailed verification report + roadmap alignment indicating the canonical run remains blocked atsource_entity_enumeration. - Add a new OpenSpec change
optimize-worker-source-entity-enumerationwith proposal/design/spec deltas + task plan focused on profiling/optimizing enumeration without weakening all-entity denominator semantics. - Enhance
_workerbatch verification with per-fixture timeout classification (via multiprocessing), phase progress persistence, additional phase timings (includingartifact_publish), and UI review-viewer handoff parameters.
Reviewed changes
Copilot reviewed 22 out of 29 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| openspec/specs/worker-demo-upload-convert-ui/spec.md | Updates worker demo UI boundary + lineage/quality + review viewer handoff requirements. |
| openspec/specs/worker-artifact-pipeline/spec.md | Tightens canonical storage batch verification semantics (statuses, timings, single-fixture gate, handoff). |
| openspec/specs/runtime-verification-evidence/spec.md | Clarifies evidence tiers and canonical gates (single fixture, visual preview, batch status classification). |
| openspec/changes/optimize-worker-source-entity-enumeration/tasks.md | New task breakdown for profiling/optimizing enumeration + evidence and tests. |
| openspec/changes/optimize-worker-source-entity-enumeration/specs/worker-artifact-pipeline/spec.md | Delta requirements for enumeration optimization + additive diagnostics. |
| openspec/changes/optimize-worker-source-entity-enumeration/specs/runtime-verification-evidence/spec.md | Delta requirements for before/after timing evidence and baseline-unlocked semantics. |
| openspec/changes/optimize-worker-source-entity-enumeration/README.md | Change README describing the optimization intent. |
| openspec/changes/optimize-worker-source-entity-enumeration/proposal.md | Motivation/scope/impact statement for the optimization slice. |
| openspec/changes/optimize-worker-source-entity-enumeration/design.md | Design decisions (profile-first, minimal identity scan, additive progress, staged validation). |
| openspec/changes/optimize-worker-source-entity-enumeration/.openspec.yaml | New change metadata. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/tasks.md | Archived change tasks updated to reflect completion/blocked status narrative. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/worker-demo-upload-convert-ui/spec.md | Archived delta spec snapshot for the UI handoff boundary. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/worker-artifact-pipeline/spec.md | Archived delta spec snapshot for batch semantics/timings/gates. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/specs/runtime-verification-evidence/spec.md | Archived delta spec snapshot for evidence acceptance semantics. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/proposal.md | Archived proposal snapshot. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/design.md | Archived design snapshot. |
| openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/.openspec.yaml | Archived change metadata. |
| docs/verification/2026-05-12-worker-canonical-storage-batch-baseline.md | New verification report documenting timeout evidence and blockers. |
| docs/plans/AI-BIM-governance-saas-roadmap-2026-05.md | Roadmap updated to show archive + blocked status and point to enumeration optimization as next active slice. |
| _worker/tests/test_worker_store.py | Asserts artifact_publish phase timing is recorded in stored quality metrics. |
| _worker/tests/test_worker_converters.py | Adds assertions for converter phase_timings completion fields. |
| _worker/tests/test_worker_batch_verification.py | Adds coverage for dry-run/subset/timeout/failed/passed semantics + handoff payload. |
| _worker/tests/test_worker_api.py | Validates UI exposes review handoff without USD rendering strings. |
| _worker/scripts/verify_storage_batch.py | Adds --timeout-seconds and exits non-zero on timed out runs. |
| _worker/app/ui.py | Adds UI fields/actions for USDC URL + review viewer handoff link. |
| _worker/app/store.py | Adds phase progress path injection + artifact_publish timing injection. |
| _worker/app/converters.py | Adds conversion phase timings + on-disk phase progress writes for timeout diagnostics. |
| _worker/app/batch_verification.py | Adds multiprocessing timeout wrapper, richer per-fixture records, phase timing merging, and baseline lock semantics. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| <dt>USDC</dt><dd id="usdcUrl">—</dd> | ||
| <dt>handoff</dt><dd id="reviewHandoff">—</dd> | ||
| </dl> | ||
| <div class="demo-actions"> | ||
| <a id="nextStep" class="demo-btn" href="http://127.0.0.1:8004" aria-disabled="true">前往建立會議</a> | ||
| <a id="reviewPreview" class="demo-btn demo-btn--secondary" href="http://127.0.0.1:8004" aria-disabled="true">開啟 USDC Review</a> | ||
| </div> | ||
| </aside> |
| ctx = mp.get_context("spawn") | ||
| queue = ctx.Queue() | ||
| process = ctx.Process( | ||
| target=_run_single_fixture_process, | ||
| args=(settings, source_id, converter, queue), | ||
| ) | ||
| started = perf_counter() | ||
| process.start() | ||
| process.join(timeout_seconds) | ||
| messages = _drain_queue(queue) | ||
| if process.is_alive(): | ||
| process.terminate() | ||
| process.join(5) | ||
| elapsed = perf_counter() - started | ||
| partial = _last_progress(messages) | ||
| phase_progress = _read_phase_progress(settings, partial) | ||
| record = _timeout_record( | ||
| source_item, | ||
| timeout_seconds=timeout_seconds, | ||
| elapsed_seconds=elapsed, | ||
| partial=partial, | ||
| phase_progress=phase_progress, | ||
| ) | ||
| _mark_job_timed_out(settings, record) | ||
| return record |
| def _timeout_phase_timings( | ||
| last_phase: str, | ||
| progress_timings: Any = None, | ||
| *, | ||
| intake_completed: bool = False, | ||
| ) -> dict[str, dict[str, Any]]: | ||
| timings = _merge_phase_timings(_pending_phase_timings(), progress_timings) | ||
| if intake_completed: | ||
| timings["source_read"] = _completed_unmeasured_timing("completed_before_timeout") | ||
| timings["artifact_intake"] = _completed_unmeasured_timing("completed_before_timeout") | ||
| if last_phase in timings: | ||
| timings[last_phase] = { | ||
| "status": "timed_out", | ||
| "duration_seconds": None, | ||
| "diagnostic": "fixture_timed_out_during_phase", | ||
| } | ||
| else: | ||
| timings["conversion_total"] = { | ||
| "status": "timed_out", | ||
| "duration_seconds": None, | ||
| "diagnostic": f"fixture_timed_out_after_{last_phase}", | ||
| } | ||
| return timings |
| - **THEN** evidence 記錄 `minimum_coverage_locked=true`、`minimum_coverage_ratio=1.0`、`coverage_denominator=source_ifc_entity_count`、per-fixture metrics，並將 batch status 設為 `passed` | ||
|
|
||
| #### Scenario: Storage fixture batch is incomplete | ||
| #### Scenario: Storage fixture batch incomplete |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b7d5e25c7e
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".
| params.set("mapping_url", result.mapping_url || ""); | ||
| const derived = result.derived_artifact_ids || {}; | ||
| if (derived.element_mapping) params.set("mapping_artifact_id", derived.element_mapping); | ||
| const target = `http://127.0.0.1:8004?${params.toString()}`; |
There was a problem hiding this comment.
The new handoff action builds http://127.0.0.1:8004?..., but the coordinator only serves its console on /ui and /dev-console (checked bim-review-coordinator/src/app.ts), and its dev-console code does not read these query params when creating a session. In the worker UI success path, clicking “開啟 USDC Review” therefore either lands on the coordinator 404/root or drops the converted usdc_url/mapping_url, so the promised handoff does not open the just-converted model. Point this to a real route that imports these params or post an artifact_bindings payload.
Useful? React with 👍 / 👎.
| started = perf_counter() | ||
| phase_timings = _pending_phase_timings() | ||
| store = WorkerStore(settings, converter=converter) | ||
| source_path, resolved_item = resolve_dev_ifc_source(settings.dev_storage_root, source_id) |
There was a problem hiding this comment.
In non-timeout runs, resolving the selected source now happens before _run_single_fixture enters its per-fixture try block. If an IFC is modified or removed between list_dev_ifc_sources() and processing, resolve_dev_ifc_source() raises ValueError("Unknown or stale dev IFC source.") and aborts the whole batch instead of recording that fixture as failed and continuing as the previous implementation did. Move this resolution into the existing try so transient storage races produce a normal failed fixture record.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: dec971861a
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".
| result = completed.get("result") or {} | ||
| quality = result.get("quality_metrics") or {} | ||
| phase_timings = _merge_phase_timings(phase_timings, quality.get("phase_timings")) |
There was a problem hiding this comment.
When a converter fails without hitting the timeout path, complete_conversion_job() returns a failed result with no quality_metrics, so this merge never picks up the .phase.json progress that the converter wrote. In that scenario the batch record only shows source_read/artifact_intake/conversion_total, losing the ifc_open or source_entity_enumeration phase that actually failed; this contradicts the new phase-timing diagnostics contract and makes non-timeout canonical blockers opaque. Read the phase progress file on failed results as well, not only in _timeout_record.
Useful? React with 👍 / 👎.
| for source_item in selected_sources: | ||
| started = perf_counter() | ||
| record: dict[str, Any] = { | ||
| "filename": source_item["filename"], | ||
| "relative_path": source_item["relative_path"], | ||
| "size_bytes": source_item["size_bytes"], | ||
| } | ||
| try: | ||
| source_path, resolved_item = resolve_dev_ifc_source(settings.dev_storage_root, source_item["source_id"]) | ||
| source_artifact = store.create_source_artifact( | ||
| ArtifactIntakeRequest( | ||
| tenant_id="tenant_batch_verification", | ||
| project_id="project_batch_verification", | ||
| model_version_id="version_batch_verification", | ||
| source_system="dev_storage", | ||
| uploaded_by="batch_verification", | ||
| filename=resolved_item["filename"], | ||
| source_format="ifc", | ||
| content_base64=base64.b64encode(source_path.read_bytes()).decode("ascii"), | ||
| ) | ||
| ) | ||
| job = store.create_conversion_job( | ||
| source_artifact["source_artifact_id"], | ||
| {"target_format": "usdc", "generate_mapping": True}, | ||
| if timeout_seconds is not None and timeout_seconds > 0: | ||
| record = _run_single_fixture_with_timeout( | ||
| settings, | ||
| source_item["source_id"], | ||
| source_item, | ||
| converter=converter, | ||
| timeout_seconds=timeout_seconds, | ||
| ) |
There was a problem hiding this comment.
With the CLI default --limit unset, selected_sources contains the full fixture set, and after one fixture times out this loop continues launching the remaining conversions. That bypasses the new current spec in openspec/specs/worker-artifact-pipeline/spec.md:261, which says the helper should try the full 13-file batch only after the canonical --limit 1 run passes or records a deterministic blocker; in the current 600s timeout case this can turn the known first-fixture blocker into many hours of unnecessary conversions and artifacts.
Useful? React with 👍 / 👎.
Code review 結論未發現需要阻擋合併的風險。 已檢查：
注意：完整 _worker API tests 在目前全域 FastAPI/Starlette 相容性下無法完成 collection，已用 focused tests 覆蓋本次主要 runtime 修改面。GitHub 不允許同帳號 approve 自己的 PR，因此此留言作為本次 code review 記錄。 |
變更摘要
本 PR 將
worker-canonical-storage-batch-baseline依使用者指示先歸檔，並建立新的 OpenSpec changeoptimize-worker-source-entity-enumeration，專門處理 canonical 89MB IFC fixture 在source_entity_enumeration階段 timeout 的 blocker。修改原因
worker-canonical-storage-batch-baseline已補入 batch timeout/status semantics、phase timing、worker UI handoff 與相關 evidence requirement，但 canonical--limit 1 --timeout-seconds 600仍卡在第一個 89MB fixture 的source_entity_enumeration，full 13-file batch 與 visual preview 仍未通過。為避免已歸檔規格被誤認為 runtime passed，本 PR 將 blocked 狀態寫入 roadmap，並把下一步收斂成更小的 enumeration optimization change。主要變更
openspec/changes/worker-canonical-storage-batch-baseline/移至openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/。openspec/specs/worker-artifact-pipeline/spec.md，納入 canonical storage batch status、timeout diagnostics、single-fixture gate、phase timing 與 baseline lock 條件。openspec/specs/runtime-verification-evidence/spec.md，明確區分 conversion success、visual preview blocker、canonical batch passed / blocked / timed_out / partial 狀態。openspec/specs/worker-demo-upload-convert-ui/spec.md，補入 worker UI lineage / quality view 與 review viewer handoff 邊界。openspec/changes/optimize-worker-source-entity-enumeration/，包含 proposal、design、delta specs 與 26 項 tasks。docs/plans/AI-BIM-governance-saas-roadmap-2026-05.md與同名 HTML 檢視版，標示 canonical batch archive 仍 runtime blocked，下一個 active change 改為 enumeration optimization。驗證方式
openspec validate optimize-worker-source-entity-enumeration --strict：通過。openspec validate --all --strict：12 passed / 0 failed。git diff --cached --check：通過，無 whitespace error。gitnexus analyze --embeddings --skills --skip-agents-md：完成，目前 worktree index 顯示 up-to-date。gitnexus status：目前 worktree index up-to-date。gitnexus detect-changes --scope staged：未完成；GitNexus resolver 仍未在 detect-changes repo 選單中刷新出本次c161worktree，即使gitnexus list與gitnexus status已可看到/確認該 index。尚未執行 canonical runtime conversion、browser visual preview、full 13-file batch 或 Kit/GPU/WebRTC 驗證。
風險與影響
worker-canonical-storage-batch-baseline已歸檔，但 runtime evidence 仍 blocked atsource_entity_enumeration；不得將此 PR 解讀為 canonical batch passed。_workersource entity enumeration，後續若更動 converter symbol，仍需依 AGENTS 執行 GitNexus impact analysis。.md。回滾方式
若需撤回，revert 本 PR 即可還原：
openspec/specs/的 canonical batch requirement 更新。openspec/changes/archive/2026-05-12-worker-canonical-storage-batch-baseline/的歸檔移動。openspec/changes/optimize-worker-source-entity-enumeration/新 change。後續建議
optimize-worker-source-entity-enumeration實作前，先針對_worker/app/converters.py相關 symbols 做 GitNexus impact analysis。_source_entities(model)/ all-entity enumeration 的耗時與 last-known operation。minimum_coverage_locked=true。Summary by CodeRabbit
New Features
Bug Fixes / Improvements
CLI
Tests
Documentation