document : try to avoid possible race conditions with minio and db on document creation

backend/app/api/routes/collections.py

@@ -188,8 +188,17 @@ def handle_pdf_upload(session: Any, collection_id: uuid.UUID, file: UploadFile)
             n_pages=len(doc.pages),
             title=doc_title,
         )
-        created_doc = cast(Document, crud.create_generic_item(session=session, model=Document, item=new_doc))
+        # Important : don't use create_generic_item because we want to commit
+        # data only when file has been written to s3 to avoid potential race condition
+        created_doc = Document.model_validate(new_doc)
+        session.add(created_doc)
+        # Use flush to insert the document in db without closing the transaction
+        session.flush()
+        session.refresh(created_doc)
         created_doc.path.write_bytes(file.file.read())
+        assert created_doc.path.exists()
+        session.commit()
+        session.refresh(created_doc)
         return created_doc
     except Exception as exc:
         session.rollback()

backend/app/models.py

@@ -265,6 +265,7 @@ class Document(DocumentCreate, table=True):
     def qdrant_collection_name(self) -> str:
         return settings.QDRANT.collection_names[self.collection.type]
 
+    @computed_field  # type: ignore[prop-decorator]
     @property
     def path(self) -> Any:
         return client.CloudPath(f"s3://{COLLECTIONS_BUCKET_NAME}/{self.collection.title}/{self.id}/raw_document.pdf")