Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(cli): warn when init --resume reuses pages without populating decisions (#670) (#2482)

Rishabh Mishra5e40f9921d95
diff --git a/packages/core/src/repowise/core/persistence/_interfaces/job_store.py b/packages/core/src/repowise/core/persistence/_interfaces/job_store.py
index 82ebeefb1..20b499caa 100644
--- a/packages/core/src/repowise/core/persistence/_interfaces/job_store.py
+++ b/packages/core/src/repowise/core/persistence/_interfaces/job_store.py
@@ -78,6 +78,7 @@ class JobStore(ABC):
*,
cursor: str | None = None,
error: str | None = None,
+ metadata: dict | None = None,
) -> JobRecord:
"""Update lifecycle state. ``cursor`` advances on each checkpoint;
``error`` is populated on failure transitions."""
diff --git a/packages/core/src/repowise/core/persistence/stores/sql_job_store.py b/packages/core/src/repowise/core/persistence/stores/sql_job_store.py
index de5021274..2751d55d0 100644
--- a/packages/core/src/repowise/core/persistence/stores/sql_job_store.py
+++ b/packages/core/src/repowise/core/persistence/stores/sql_job_store.py
@@ -73,6 +73,7 @@ class SqlJobStore(JobStore):
*,
cursor: str | None = None,
error: str | None = None,
+ metadata: dict | None = None,
) -> JobRecord:
row = await self._session.get(PipelineJob, job_id)
if row is None:
@@ -82,6 +83,13 @@ class SqlJobStore(JobStore):
row.cursor = cursor
if error is not None:
row.error = error
+ if metadata is not None:
+ try:
+ current_meta = json.loads(row.metadata_json) if row.metadata_json else {}
+ except json.JSONDecodeError:
+ current_meta = {}
+ current_meta.update(metadata)
+ row.metadata_json = json.dumps(current_meta)
row.updated_at = _now_utc()
await self._session.flush()
return _to_record(row)
diff --git a/packages/core/src/repowise/core/pipeline/orchestrator.py b/packages/core/src/repowise/core/pipeline/orchestrator.py
index 3233223e2..08d39cd0b 100644
--- a/packages/core/src/repowise/core/pipeline/orchestrator.py
+++ b/packages/core/src/repowise/core/pipeline/orchestrator.py
@@ -505,6 +505,34 @@ async def run_pipeline(
logger.warning("resume_rehydrate_analysis_failed_recomputing", error=str(exc))
skip_analysis = False
+ # If analysis was skipped, but decision extraction never completed (e.g. earlier
+ # run was interrupted during decision extraction or returned None), re-run decision
+ # extraction specifically so that decisions are backfilled without re-running dead code / health.
+ if (
+ skip_analysis
+ and resume_controller
+ and not await resume_controller.has_completed_decision_extraction()
+ ):
+ if progress:
+ progress.on_message(
+ "info",
+ " ↳ Resumed analysis missing decision extraction — running extraction",
+ )
+ decision_report = await _run_decision_extraction(
+ repo_path,
+ llm_client=llm_client,
+ graph_builder=graph_builder,
+ git_meta_map=git_meta_map,
+ parsed_files=parsed_files,
+ source_map=source_map,
+ progress=progress,
+ )
+ if decision_report is not None:
+ gen_decision_report = decision_report
+ await resume_controller.checkpoint_decision_backfill(
+ decision_report, progress=progress
+ )
+
if not skip_analysis:
# The four analyses share read-only inputs (graph, git_meta_map,
# parsed_files; the lazy metric caches were warmed during ingestion)
diff --git a/packages/core/src/repowise/core/pipeline/resume/controller.py b/packages/core/src/repowise/core/pipeline/resume/controller.py
index f86b2f12f..8749b39fe 100644
--- a/packages/core/src/repowise/core/pipeline/resume/controller.py
+++ b/packages/core/src/repowise/core/pipeline/resume/controller.py
@@ -230,10 +230,77 @@ class ResumeController:
"a resumed run will have to recompute dead code, health and decisions.",
)
return
- await self._ledger.mark_completed(ResumePhase.ANALYSIS)
+ extraction_completed = decision_report is not None
+ decisions_count = (
+ len(getattr(decision_report, "decisions", []) or []) if extraction_completed else 0
+ )
+ await self._ledger.mark_completed(
+ ResumePhase.ANALYSIS,
+ metadata={
+ "decision_extraction_completed": extraction_completed,
+ "decisions_count": decisions_count,
+ },
+ )
logger.info("resume_checkpoint_analysis", repo_id=self._repo_id)
+ async def has_completed_decision_extraction(self) -> bool:
+ """True iff the completed ANALYSIS phase recorded that decision extraction ran."""
+ job = await self._ledger.get_completed_job(ResumePhase.ANALYSIS)
+ if job is None:
+ return False
+ if job.metadata.get("decision_extraction_completed") is True:
+ return True
+ from repowise.core.persistence import get_session
+ from repowise.core.persistence.crud import list_decisions
+
+ try:
+ async with get_session(self._sf) as session:
+ rows = await list_decisions(session, self._repo_id, limit=1)
+ if rows:
+ return True
+ except Exception as exc:
+ logger.debug("check_decision_records_failed", error=str(exc))
+ return False
+
+ async def checkpoint_decision_backfill(
+ self,
+ decision_report: Any,
+ progress: Any | None = None,
+ ) -> None:
+ """Persist backfilled decisions for a resumed run and record completion."""
+ from repowise.core.persistence import get_session
+
+ view = SimpleNamespace(
+ dead_code_report=None,
+ health_report=None,
+ decision_report=decision_report,
+ doc_drift_report=None,
+ git_metadata_list=[],
+ generated_pages=None,
+ vector_store=None,
+ )
+ try:
+ async with get_session(self._sf) as session:
+ await persist_analysis(view, session, self._repo_id)
+ decisions_count = len(getattr(decision_report, "decisions", []) or [])
+ await self.mark_decision_extraction_completed(count=decisions_count)
+ except Exception as exc:
+ logger.warning("resume_backfill_decision_persist_failed", error=str(exc))
+ emit_warning(
+ progress,
+ f"Backfilled decisions not saved ({exc}); "
+ "a subsequent resumed run will re-run decision extraction.",
+ )
+
+ async def mark_decision_extraction_completed(self, count: int = 0) -> None:
+ """Update the completed ANALYSIS job record to reflect that decision extraction finished."""
+ await self._ledger.mark_completed(
+ ResumePhase.ANALYSIS,
+ metadata={"decision_extraction_completed": True, "decisions_count": count},
+ )
+
async def mark_phase_complete(self, phase: ResumePhase) -> None:
"""Record *phase* as completed (analysis / generation persisted by the
normal end-of-run persist; this just stamps the ledger)."""
await self._ledger.mark_completed(phase)
+
diff --git a/packages/core/src/repowise/core/pipeline/resume/ledger.py b/packages/core/src/repowise/core/pipeline/resume/ledger.py
index 51af83695..473121b66 100644
--- a/packages/core/src/repowise/core/pipeline/resume/ledger.py
+++ b/packages/core/src/repowise/core/pipeline/resume/ledger.py
@@ -72,7 +72,22 @@ class ResumeLedger:
except Exception as exc:
logger.debug("resume_ledger_start_failed", phase=str(phase), error=str(exc))
- async def mark_completed(self, phase: ResumePhase) -> None:
+ async def get_completed_job(self, phase: ResumePhase) -> Any | None:
+ """Return the latest COMPLETED job record for *phase*, if any."""
+ from repowise.core.persistence import get_session
+ from repowise.core.persistence._interfaces.job_store import JobState
+
+ try:
+ async with get_session(self._sf) as session:
+ jobs = await _store(session).list_jobs(
+ repository_id=self._repo_id, phase=str(phase), state=JobState.COMPLETED
+ )
+ return jobs[0] if jobs else None
+ except Exception as exc:
+ logger.debug("resume_ledger_get_job_failed", phase=str(phase), error=str(exc))
+ return None
+
+ async def mark_completed(self, phase: ResumePhase, metadata: dict | None = None) -> None:
from repowise.core.persistence import get_session
from repowise.core.persistence._interfaces.job_store import JobState
@@ -83,8 +98,12 @@ class ResumeLedger:
if job_id is None:
# No open job (e.g. mark_started failed) — open one so the
# COMPLETED state still lands.
- job = await store.create_job(repository_id=self._repo_id, phase=str(phase))
+ job = await store.create_job(
+ repository_id=self._repo_id,
+ phase=str(phase),
+ metadata=metadata,
+ )
job_id = job.id
- await store.update_state(job_id, JobState.COMPLETED)
+ await store.update_state(job_id, JobState.COMPLETED, metadata=metadata)
except Exception as exc:
logger.debug("resume_ledger_complete_failed", phase=str(phase), error=str(exc))
diff --git a/tests/unit/pipeline/test_resume.py b/tests/unit/pipeline/test_resume.py
index 2a408a4da..a9afd7b16 100644
--- a/tests/unit/pipeline/test_resume.py
+++ b/tests/unit/pipeline/test_resume.py
@@ -331,3 +331,50 @@ async def test_failed_symbol_analysis_reconciliation_keeps_analysis_retryable(sf
"a resumed run will have to recompute dead code, health and decisions.",
)
]
+
+
+async def test_decision_extraction_completion_tracking(sf, monkeypatch):
+ """Test that ResumeController correctly records and checks decision extraction status."""
+ from types import SimpleNamespace
+
+ async def _ok(*_a: object, **_k: object) -> None:
+ return None
+
+ monkeypatch.setattr("repowise.core.pipeline.resume.controller.persist_symbol_analysis", _ok)
+ monkeypatch.setattr("repowise.core.pipeline.resume.controller.persist_analysis", _ok)
+
+ repo_id = await _make_repo(sf)
+ progress = _RecordingProgress()
+
+ # Case 1: checkpoint_analysis without decision_report (e.g. extraction not completed)
+ ctrl1 = ResumeController(sf, repo_id, resume=False)
+ await ctrl1.checkpoint_analysis(
+ parsed_files=[],
+ dead_code_report=None,
+ health_report=None,
+ decision_report=None,
+ git_metadata_list=[],
+ progress=progress,
+ )
+
+ assert await ctrl1.has_completed_decision_extraction() is False
+
+ # Mark decision extraction completed
+ await ctrl1.mark_decision_extraction_completed(count=5)
+ assert await ctrl1.has_completed_decision_extraction() is True
+
+ # Case 2: fresh controller with genuine empty decisions
+ repo_id_2 = await _make_repo(sf)
+ ctrl2 = ResumeController(sf, repo_id_2, resume=False)
+ empty_decision_report = SimpleNamespace(decisions=[])
+ await ctrl2.checkpoint_analysis(
+ parsed_files=[],
+ dead_code_report=None,
+ health_report=None,
+ decision_report=empty_decision_report,
+ git_metadata_list=[],
+ progress=progress,
+ )
+
+ assert await ctrl2.has_completed_decision_extraction() is True
+