fix(ci): restore green backend gate after engine pin

Ruff: declare adapter re-exports via __all__ in docling_extractor
(F401) and sort the engine import block in the adapter test (I001).
Pytest: engine-dependent tests now importorskip the pinned engine —
CI installs light deps only and the engine repo is private, so they
skip visibly there and run in Docker/dev envs with the engine.

Verified: ruff clean; pytest 74 passed + 2 skipped without engine,
80 passed with the pinned engine v0.1.0 installed.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.github/workflows/ci.yml

@@ -40,10 +40,13 @@ jobs:
         run: python -m compileall -q app scripts tests
 
       - name: Pytest (unit only — heavy deps excluded)
+        # test_chunker.py skips here: it needs the pinned document recognition
+        # engine (private TeamHUB_Engines repo, no PAT secret configured).
+        # It runs in the Docker image and in dev envs with the engine installed.
         run: |
           pytest tests/test_hashing.py tests/test_quality.py \
                  tests/test_chunker.py tests/test_duplicates.py \
-                 tests/test_hybrid_search.py -q
+                 tests/test_hybrid_search.py -q -rs
 
       - name: docker compose config
         run: docker compose config --quiet

app/ingestion/docling_extractor.py

@@ -15,6 +15,17 @@ from teamhub_document_recognition_engine import extract as _engine_extract
 
 from app.config import settings
 
+# Re-exported engine models: chunker, table/figure processors and tests import
+# them from this adapter instead of depending on the engine package directly.
+__all__ = [
+    "ExtractedBlock",
+    "ExtractedFigure",
+    "ExtractedPage",
+    "ExtractedTable",
+    "ExtractionResult",
+    "extract",
+]
+
 
 def extract(pdf_path: Path) -> ExtractionResult:
     return _engine_extract(pdf_path, docling_ocr_enabled=settings.docling_ocr_enabled)

tests/test_chunker.py

@@ -1,7 +1,14 @@
 from __future__ import annotations
 
-from app.ingestion.chunker import chunk_extraction
-from app.ingestion.docling_extractor import (
+import pytest
+
+# CI installs light deps only; the pinned engine repo is private (no PAT), so
+# these tests run wherever the engine is installed (Docker image, dev env) and
+# skip visibly elsewhere. See app/ingestion/ENGINE_SOURCE.md.
+pytest.importorskip("teamhub_document_recognition_engine")
+
+from app.ingestion.chunker import chunk_extraction  # noqa: E402
+from app.ingestion.docling_extractor import (  # noqa: E402
     ExtractedBlock,
     ExtractedFigure,
     ExtractedPage,

tests/test_document_recognition_engine_adapter.py

@@ -1,5 +1,11 @@
 from pathlib import Path
 
+import pytest
+
+# Skip cleanly where the pinned engine is not installed (e.g. light-deps CI);
+# see app/ingestion/ENGINE_SOURCE.md for the focused verification command.
+pytest.importorskip("teamhub_document_recognition_engine")
+
 
 def test_legacy_ocr_adapter_reexports_engine_model_and_settings(monkeypatch, tmp_path: Path) -> None:
     from teamhub_document_recognition_engine import OcrResult as EngineOcrResult
@@ -34,6 +40,8 @@ def test_legacy_ocr_adapter_reexports_engine_model_and_settings(monkeypatch, tmp
 def test_legacy_docling_adapter_reexports_engine_models_and_settings(monkeypatch, tmp_path: Path) -> None:
     from teamhub_document_recognition_engine import (
         ExtractedBlock as EngineExtractedBlock,
+    )
+    from teamhub_document_recognition_engine import (
         ExtractionResult as EngineExtractionResult,
     )