Fix pdf_infer_table_strategy being silently changed in API (#411)

Changes the logic which disabled pdf_infer_table_strategy for auto
strategy.

Author: hubert-rutkowski85

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 0.0.67-dev1
+
+* Change pdf_infer_table_structure parameter from being disabled in auto strategy.
+
 ## 0.0.66
 
 * Add support for `unique_element_ids` parameter.

prepline_general/api/app.py

@@ -12,7 +12,7 @@
 app = FastAPI(
     title="Unstructured Pipeline API",
     summary="Partition documents with the Unstructured library",
-    version="0.0.66",
+    version="0.0.67",
     docs_url="/general/docs",
     openapi_url="/general/openapi.json",
     servers=[

prepline_general/api/general.py

@@ -596,7 +596,8 @@ def _validate_chunking_strategy(chunking_strategy: Optional[str]) -> Optional[st
 
 
 def _set_pdf_infer_table_structure(pdf_infer_table_structure: bool, strategy: str) -> bool:
-    return strategy == "hi_res" and pdf_infer_table_structure
+    """Avoids table inference in "fast" and "ocr_only" runs."""
+    return strategy in ("hi_res", "auto") and pdf_infer_table_structure
 
 
 def get_validated_mimetype(file: UploadFile) -> Optional[str]:
@@ -703,7 +704,7 @@ def return_content_type(filename: str):
 
 
 @router.get("/general/v0/general", include_in_schema=False)
-@router.get("/general/v0.0.66/general", include_in_schema=False)
+@router.get("/general/v0.0.67/general", include_in_schema=False)
 async def handle_invalid_get_request():
     raise HTTPException(
         status_code=status.HTTP_405_METHOD_NOT_ALLOWED, detail="Only POST requests are supported."
@@ -718,7 +719,7 @@ async def handle_invalid_get_request():
     description="Description",
     operation_id="partition_parameters",
 )
-@router.post("/general/v0.0.66/general", include_in_schema=False)
+@router.post("/general/v0.0.67/general", include_in_schema=False)
 def general_partition(
     request: Request,
     # cannot use annotated type here because of a bug described here:

preprocessing-pipeline-family.yaml

@@ -1,2 +1,2 @@
 name: general
-version: 0.0.66
+version: 0.0.67

scripts/smoketest.py

@@ -207,18 +207,10 @@ def test_strategy_performance():
     hi_res_time = time.monotonic() - start_time
     assert response.status_code == 200
 
-    start_time = time.monotonic()
-    response = send_document(filenames=[test_file], content_type="application/pdf", strategy="auto")
-    auto_time = time.monotonic() - start_time
-    assert response.status_code == 200
-
-    assert hi_res_time > performance_ratio * auto_time
-
     start_time = time.monotonic()
     response = send_document(filenames=[test_file], content_type="application/pdf", strategy="fast")
     fast_time = time.monotonic() - start_time
     assert response.status_code == 200
-
     assert hi_res_time > performance_ratio * fast_time
 
 

test_general/api/test_app.py

@@ -860,14 +860,17 @@ def test_chunking_strategy_param():
 @pytest.mark.parametrize(
     ("multipage_sections", "combine_under_n_chars", "new_after_n_chars", "max_characters"),
     [
-        (False, None, None, 500),  # test multipage_sections
+        (False, None, None, 600),  # test multipage_sections
         (True, 1000, None, 5000),  # test combine_under_n_chars
         (True, None, 10, 500),  # test new_after_n_chars
         (True, None, None, 100),  # test max__characters
     ],
 )
 def test_chunking_strategy_additional_params(
-    multipage_sections, combine_under_n_chars, new_after_n_chars, max_characters
+    multipage_sections: bool,
+    combine_under_n_chars: int,
+    new_after_n_chars: int,
+    max_characters: int,
 ):
     client = TestClient(app)
     test_file = Path("sample-docs") / "layout-parser-paper-fast.pdf"