Fixing dockey/doc_id mismatch when no metadata is found (#1288)

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: jamesbraza

src/paperqa/clients/__init__.py

@@ -251,5 +251,13 @@ async def upgrade_doc_to_doc_details(self, doc: Doc, **kwargs) -> DocDetails:
             return provided_doc_details + doc_details
 
         # if we can't get metadata, just return the doc, but don't overwrite any fields
-        orig_fields = doc.model_dump() | {"fields_to_overwrite_from_metadata": set()}
+        overwrite_fields: set[str] = set()
+        if doc.dockey == doc.content_hash:
+            # This allows DocDetails validator on fields_to_overwrite_from_metadata
+            # to sync dockey with doc_id. Otherwise, dockey remains the raw
+            # content_hash and won't match the computed doc_id.
+            overwrite_fields.add("doc_id")
+        orig_fields = doc.model_dump() | {
+            "fields_to_overwrite_from_metadata": overwrite_fields
+        }
         return DocDetails(**(orig_fields | provided_fields))

tests/test_agents.py

@@ -103,7 +103,9 @@ async def test_get_directory_index(
             results = await index.query(query="who is Frederick Bates?", min_score=5)
             assert results
             target_doc_path = (paper_dir / "bates.txt").absolute()
-            assert results[0].docs.keys() == {md5sum(target_doc_path)}, (
+            assert results[0].docs.keys() == {
+                compute_unique_doc_id(None, md5sum(target_doc_path))
+            }, (
                 f"Expected to find {target_doc_path.name!r}, got citations"
                 f" {[d.formatted_citation for d in results[0].docs.values()]}."
             )