feat: url add knowledge

backend/api/quivr_api/modules/knowledge/service/knowledge_service.py

@@ -231,8 +231,15 @@ async def create_knowledge(
                         "knowledge_id": knowledge_db.id,
                     },
                 )
+                return knowledge_db
+            else:
+                knowledge_db = await self.repository.update_knowledge(
+                    knowledge_db,
+                    KnowledgeUpdate(status=KnowledgeStatus.UPLOADED),
+                    autocommit=autocommit,
+                )
+                return knowledge_db
 
-            return knowledge_db
         except Exception as e:
             logger.exception(
                 f"Error uploading knowledge {knowledgedb.id} to storage : {e}"

backend/api/quivr_api/modules/knowledge/tests/test_knowledge_service.py

@@ -517,6 +517,30 @@ async def test_create_knowledge_file(session: AsyncSession, user: User):
     storage.knowledge_exists(km)
 
 
+@pytest.mark.asyncio(loop_scope="session")
+async def test_create_knowledge_web(session: AsyncSession, user: User):
+    assert user.id
+    storage = FakeStorage()
+    repository = KnowledgeRepository(session)
+    service = KnowledgeService(repository, storage)
+
+    km_to_add = AddKnowledge(
+        url="http://quivr.app",
+        source=KnowledgeSource.WEB,
+        is_folder=False,
+        parent_id=None,
+    )
+
+    km = await service.create_knowledge(
+        user_id=user.id, knowledge_to_add=km_to_add, upload_file=None
+    )
+
+    assert km.id
+    assert km.url == km_to_add.url
+    assert km.status == KnowledgeStatus.UPLOADED
+    assert not km.is_folder
+
+
 @pytest.mark.asyncio(loop_scope="session")
 async def test_create_knowledge_folder(session: AsyncSession, user: User):
     assert user.id
@@ -555,7 +579,7 @@ async def test_create_knowledge_folder(session: AsyncSession, user: User):
 
 
 @pytest.mark.asyncio(loop_scope="session")
-async def test_create_knowledge_file_in_folder(
+async def test_create_knowledge_file_in_folder_in_brain(
     monkeypatch, session: AsyncSession, user: User, folder_km_brain: KnowledgeDB
 ):
     tasks = {}
@@ -948,10 +972,6 @@ async def test_list_knowledge_root(session: AsyncSession, user: User):
     assert len(root_kms) == 2
     assert {k.id for k in root_kms} == {root_folder.id, root_file.id}
 
-    # check order
-    assert root_kms[0].file_name == "folder"
-    assert root_kms[1].file_name == "file_1"
-
 
 @pytest.mark.asyncio(loop_scope="session")
 async def test_list_knowledge(session: AsyncSession, user: User):

backend/worker/quivr_worker/process/README.md

@@ -49,8 +49,17 @@ If an exception occurs during the parsing loop, the following steps are taken:
      - This operation should be rolled back if an error occurs afterward. Otherwise, the knowledge could remain in `Processing` or `ERROR` status with associated vectors.
      - Reprocessing the knowledge would result in reinserting the vectors into the database, leading to duplicate vectors for the same knowledge.
 
-2. Set the knowledge status to `ERROR`.
-3. Continue processing.
+**Transaction Safety for Each Operation:**
+
+- **Creating knowledge and linking to brains**: These operations can be retried safely. Knowledge is only recreated if it does not already exist in the database, allowing for safe retry.
+- **Downloading sync files**: This operation is idempotent but is safe to retry. If a change has occured, we would download the last version of the file.
+- **Linking knowledge to brains**: Only links the brain if it is not already associated with the knowledge. Safe for retry.
+- **Creating vectors**:
+  - This operation should be rolled back if an error occurs afterward. Otherwise, the knowledge could remain in `Processing` or `ERROR` status with associated vectors.
+  - Reprocessing the knowledge would result in reinserting the vectors into the database, leading to duplicate vectors for the same knowledge.
+
+1. Set the knowledge status to `ERROR`.
+2. Continue processing.
 
 | Note: This means that some knowledges will remain in an errored state. Currently, they are not automatically rescheduled for processing.