chore: fix CI — ruff formatting and minio service startup

.github/workflows/eval.yml

@@ -26,13 +26,6 @@ jobs:
       qdrant:
         image: qdrant/qdrant:v1.12.4
         ports: ["6333:6333"]
-      minio:
-        image: minio/minio:RELEASE.2024-10-13T13-34-11Z
-        env:
-          MINIO_ROOT_USER: kbminio
-          MINIO_ROOT_PASSWORD: kbminio-secret
-        ports: ["9000:9000"]
-        options: --health-cmd "curl -fsS http://localhost:9000/minio/health/live" --health-interval 3s --health-retries 20
 
     steps:
       - uses: actions/checkout@v4
@@ -41,6 +34,21 @@ jobs:
           version: "0.11.17"
       - run: uv python install 3.11
       - run: uv sync --all-extras
+      - name: Start MinIO
+        # minio/minio image requires "server /data" as the CMD argument; GitHub Actions service
+        # containers have no way to pass CMD args, so we start minio as a background container step.
+        run: |
+          docker run -d --name minio \
+            -p 9000:9000 \
+            -e MINIO_ROOT_USER=kbminio \
+            -e MINIO_ROOT_PASSWORD=kbminio-secret \
+            minio/minio:RELEASE.2024-10-13T13-34-11Z server /data
+          for i in $(seq 1 20); do
+            if curl -fsS http://localhost:9000/minio/health/live; then
+              echo "MinIO is ready"; break
+            fi
+            sleep 3
+          done
       - name: Lint + unit tests
         run: |
           uv run ruff check src tests

src/kb/api/routes/infer.py

@@ -139,9 +139,7 @@ async def infer_from_files(
             )
             remaining = max(sample_count - len(samples), 0)
             if remaining:
-                samples.extend(
-                    _sample_texts_from_elements(elements, sample_count=remaining)
-                )
+                samples.extend(_sample_texts_from_elements(elements, sample_count=remaining))
         except Exception as e:
             errors.append({"filename": filename, "error": str(e)})
 

src/kb/api/routes/ingest_data.py

@@ -223,7 +223,9 @@ async def ingest_record(body: RecordIn) -> RecordOut:
                     "is_parent": False,
                     "source": "record",
                 },
-                content_hash=chunk_content_hash(f"record:{body.project}:{body.kind}:{body.type}:{excerpt}"),
+                content_hash=chunk_content_hash(
+                    f"record:{body.project}:{body.kind}:{body.type}:{excerpt}"
+                ),
             )
         )
         upserted += 1
@@ -233,7 +235,9 @@ async def ingest_record(body: RecordIn) -> RecordOut:
         await store.delete_by_file(body.kind, file_id)
         await store.upsert(body.kind, chunks)
     except Exception as e:
-        await repo.set_file_status(file_id, "failed", error=f"record vector indexing failed: {e}"[:500])
+        await repo.set_file_status(
+            file_id, "failed", error=f"record vector indexing failed: {e}"[:500]
+        )
         raise HTTPException(503, f"record stored but vector indexing failed: {e}") from e
 
     return RecordOut(

src/kb/api/routes/schemas.py

@@ -118,9 +118,7 @@ async def apply_draft(draft_id: str, body: ApplyDraftIn | None = None) -> dict:
 @router.post("/drafts/{draft_id}/discard")
 async def discard_draft(draft_id: str, body: DraftProjectIn | None = None) -> dict:
     body = body or DraftProjectIn()
-    row = await repo.update_schema_draft_status(
-        draft_id, project=body.project, status="discarded"
-    )
+    row = await repo.update_schema_draft_status(draft_id, project=body.project, status="discarded")
     if not row:
         raise HTTPException(404, "schema draft not found")
     return row
@@ -147,7 +145,9 @@ async def reprocess_for_active_schema(domain: str, body: ReprocessIn | None = No
     body = body or ReprocessIn()
     sch = await get_active_schema(domain, project=body.project)
     if not sch:
-        raise HTTPException(404, f"No active schema for domain '{domain}' in project '{body.project}'")
+        raise HTTPException(
+            404, f"No active schema for domain '{domain}' in project '{body.project}'"
+        )
     enqueued = await reindex_domain_with_schema(
         project=body.project,
         domain=domain,

src/kb/query/engine.py

@@ -679,8 +679,7 @@ def _section_score(h: Any) -> float:
         and (
             hard_citation_gate
             or not (
-                crag_score >= verify_skip_threshold
-                and confidence_value >= verify_skip_threshold
+                crag_score >= verify_skip_threshold and confidence_value >= verify_skip_threshold
             )
         )
     )

src/kb/query/search.py

@@ -24,7 +24,9 @@
 _TOKEN_RE = re.compile(r"[A-Za-z0-9][A-Za-z0-9_-]{2,}")
 
 
-def _explicit_filters(scope: dict[str, Any] | None, filters: dict[str, Any] | None) -> dict[str, Any]:
+def _explicit_filters(
+    scope: dict[str, Any] | None, filters: dict[str, Any] | None
+) -> dict[str, Any]:
     out: dict[str, Any] = {}
     if scope:
         for key in ("entity_id", "file_id", "parent_id"):

src/kb/vector/qdrant_store.py

@@ -176,12 +176,10 @@ async def upsert(self, domain: str, chunks: list[Chunk]) -> None:
                 collection_name=_collection(domain),
                 scroll_filter=qm.Filter(
                     must=[
-                        qm.FieldCondition(
-                            key="project", match=qm.MatchValue(value=project)
-                        ),
+                        qm.FieldCondition(key="project", match=qm.MatchValue(value=project)),
                         qm.FieldCondition(
                             key="content_hash", match=qm.MatchAny(any=list(set(hashes)))
-                        )
+                        ),
                     ]
                 ),
                 limit=len(hashes) * 2,