fix: upgrade pydantic to 2.10 to fix NodeWithScore serialization error (#587)

Also:
- fix typo for `enable_kg_enhance_query_refine` parameter

Author: Mini256

backend/app/api/admin_routes/models.py

@@ -44,4 +44,4 @@ class ChatEngineBasedRetrieveRequest(BaseModel):
     top_k: Optional[int] = 5
     similarity_top_k: Optional[int] = None
     oversampling_factor: Optional[int] = 5
-    enable_kg_enchance_query_refine: Optional[bool] = True
+    enable_kg_enhance_query_refine: Optional[bool] = False

backend/app/api/admin_routes/retrieve.py

@@ -15,15 +15,15 @@
 
 
 @router.get("/admin/retrieve/documents")
-async def retrieve_documents(
+def retrieve_documents(
     session: SessionDep,
     user: CurrentSuperuserDep,
     question: str,
     chat_engine: str = "default",
     top_k: Optional[int] = 5,
     similarity_top_k: Optional[int] = None,
     oversampling_factor: Optional[int] = 5,
-    enable_kg_enchance_query_refine: Optional[bool] = True,
+    enable_kg_enhance_query_refine: Optional[bool] = True,
 ) -> List[Document]:
     try:
         return retrieve_service.chat_engine_retrieve_documents(
@@ -33,7 +33,7 @@ async def retrieve_documents(
             chat_engine_name=chat_engine,
             similarity_top_k=similarity_top_k,
             oversampling_factor=oversampling_factor,
-            enable_kg_enchance_query_refine=enable_kg_enchance_query_refine,
+            enable_kg_enhance_query_refine=enable_kg_enhance_query_refine,
         )
     except KBNotFound as e:
         raise e
@@ -43,26 +43,27 @@ async def retrieve_documents(
 
 
 @router.get("/admin/embedding_retrieve")
-async def embedding_retrieve(
+def embedding_retrieve(
     session: SessionDep,
     user: CurrentSuperuserDep,
     question: str,
     chat_engine: str = "default",
     top_k: Optional[int] = 5,
     similarity_top_k: Optional[int] = None,
     oversampling_factor: Optional[int] = 5,
-    enable_kg_enchance_query_refine: Optional[bool] = True,
+    enable_kg_enhance_query_refine=False,
 ) -> List[NodeWithScore]:
     try:
-        return retrieve_service.chat_engine_retrieve_chunks(
+        nodes = retrieve_service.chat_engine_retrieve_chunks(
             session,
             question=question,
             top_k=top_k,
             chat_engine_name=chat_engine,
             similarity_top_k=similarity_top_k,
             oversampling_factor=oversampling_factor,
-            enable_kg_enchance_query_refine=enable_kg_enchance_query_refine,
+            enable_kg_enhance_query_refine=enable_kg_enhance_query_refine,
         )
+        return nodes
     except KBNotFound as e:
         raise e
     except Exception as e:
@@ -71,7 +72,7 @@ async def embedding_retrieve(
 
 
 @router.post("/admin/embedding_retrieve")
-async def embedding_search(
+def embedding_search(
     session: SessionDep,
     user: CurrentSuperuserDep,
     request: ChatEngineBasedRetrieveRequest,
@@ -83,7 +84,7 @@ async def embedding_search(
             top_k=request.top_k,
             similarity_top_k=request.similarity_top_k,
             oversampling_factor=request.oversampling_factor,
-            enable_kg_enchance_query_refine=request.enable_kg_enchance_query_refine,
+            enable_kg_enhance_query_refine=request.enable_kg_enhance_query_refine,
         )
     except KBNotFound as e:
         raise e

backend/app/rag/retrieve.py

@@ -27,7 +27,7 @@ def chat_engine_retrieve_documents(
         chat_engine_name: str = "default",
         similarity_top_k: Optional[int] = None,
         oversampling_factor: Optional[int] = None,
-        enable_kg_enchance_query_refine: bool = False,
+        enable_kg_enhance_query_refine: bool = False,
     ) -> List[DBDocument]:
         chat_engine_config = ChatEngineConfig.load_from_db(db_session, chat_engine_name)
         if not chat_engine_config.knowledge_base:
@@ -40,7 +40,7 @@ def chat_engine_retrieve_documents(
             chat_engine_name=chat_engine_name,
             similarity_top_k=similarity_top_k,
             oversampling_factor=oversampling_factor,
-            enable_kg_enchance_query_refine=enable_kg_enchance_query_refine,
+            enable_kg_enhance_query_refine=enable_kg_enhance_query_refine,
         )
 
         linked_knowledge_base = chat_engine_config.knowledge_base.linked_knowledge_base
@@ -59,15 +59,15 @@ def chat_engine_retrieve_chunks(
         chat_engine_name: str = "default",
         similarity_top_k: Optional[int] = None,
         oversampling_factor: Optional[int] = None,
-        enable_kg_enchance_query_refine: bool = False,
+        enable_kg_enhance_query_refine: bool = False,
     ) -> List[NodeWithScore]:
         retriever = ChatEngineBasedRetriever(
             db_session=db_session,
             engine_name=chat_engine_name,
             top_k=top_k,
             similarity_top_k=similarity_top_k,
             oversampling_factor=oversampling_factor,
-            enable_kg_enchance_query_refine=enable_kg_enchance_query_refine,
+            enable_kg_enhance_query_refine=enable_kg_enhance_query_refine,
         )
         return retriever.retrieve(question)
 

backend/app/rag/retrievers/LegacyChatEngineRetriever.py

@@ -33,14 +33,14 @@ def __init__(
         top_k: int = 10,
         similarity_top_k: int = None,
         oversampling_factor: int = 5,
-        enable_kg_enchance_query_refine: bool = False,
+        enable_kg_enhance_query_refine: bool = False,
     ):
         self.db_session = db_session
         self.engine_name = engine_name
         self.top_k = top_k
         self.similarity_top_k = similarity_top_k
         self.oversampling_factor = oversampling_factor
-        self.enable_kg_enchance_query_refine = enable_kg_enchance_query_refine
+        self.enable_kg_enhance_query_refine = enable_kg_enhance_query_refine
 
         self.chat_engine_config = chat_engine_config or ChatEngineConfig.load_from_db(
             db_session, engine_name
@@ -128,7 +128,7 @@ def _refine_query(self, query: str) -> str:
                 )
                 graph_knowledges_context = graph_knowledges.template
             else:
-                entities, relations = graph_index.retrieve_with_weight(
+                entities, relations, _ = graph_index.retrieve_with_weight(
                     query,
                     [],
                     depth=kg_config.depth,

backend/app/repositories/chunk.py

@@ -23,7 +23,7 @@ def document_exists_chunks(self, session: Session, document_id: int) -> bool:
         )
 
     def get_documents_by_chunk_ids(
-        self, session: Session, chunk_ids: list[int]
+        self, session: Session, chunk_ids: list[str]
     ) -> list[DBDocument]:
         stmt = select(DBDocument).where(
             DBDocument.id.in_(

backend/pyproject.toml

@@ -12,9 +12,9 @@ dependencies = [
     "celery>=5.4.0",
     "dspy-ai>=2.4.9",
     "langfuse>=2.48.0",
-    "llama-index>=0.11.10",
+    "llama-index==0.12.10",
     "alembic>=1.13.1",
-    "pydantic>=2.8.2",
+    "pydantic==2.10.5",
     # Update Check: https://github.com/pydantic/pydantic/issues/8061
     "pydantic-settings>=2.3.3",
     "sentry-sdk>=2.5.1",
@@ -23,7 +23,8 @@ dependencies = [
     "tenacity~=8.4.0",
     "redis>=5.0.5",
     "flower>=2.0.1",
-    "llama-index-llms-gemini>=0.1.11",
+    # Notice: Ensure bug issue https://github.com/run-llama/llama_index/issues/17475 be fixed before upgrade to 0.4.3
+    "llama-index-llms-gemini==0.4.2",
     "tidb-vector>=0.0.14",
     "deepdiff>=7.0.1",
     "python-dotenv>=1.0.1",

backend/requirements-dev.lock

@@ -361,7 +361,7 @@ llama-index-llms-anthropic==0.6.3
 llama-index-llms-bedrock==0.3.3
 llama-index-llms-gemini==0.4.2
 llama-index-llms-ollama==0.4.2
-llama-index-llms-openai==0.3.12
+llama-index-llms-openai==0.3.13
     # via llama-index
     # via llama-index-agent-openai
     # via llama-index-cli
@@ -532,7 +532,7 @@ pyasn1-modules==0.4.0
     # via google-auth
 pycparser==2.22
     # via cffi
-pydantic==2.8.2
+pydantic==2.10.5
     # via anthropic
     # via cohere
     # via deepeval
@@ -551,7 +551,7 @@ pydantic==2.8.2
     # via pydantic-settings
     # via ragas
     # via sqlmodel
-pydantic-core==2.20.1
+pydantic-core==2.27.2
     # via pydantic
 pydantic-settings==2.6.1
     # via langchain-community

backend/requirements.lock

@@ -354,7 +354,7 @@ llama-index-llms-anthropic==0.6.3
 llama-index-llms-bedrock==0.3.3
 llama-index-llms-gemini==0.4.2
 llama-index-llms-ollama==0.4.2
-llama-index-llms-openai==0.3.12
+llama-index-llms-openai==0.3.13
     # via llama-index
     # via llama-index-agent-openai
     # via llama-index-cli
@@ -520,7 +520,7 @@ pyasn1-modules==0.4.0
     # via google-auth
 pycparser==2.22
     # via cffi
-pydantic==2.8.2
+pydantic==2.10.5
     # via anthropic
     # via cohere
     # via deepeval
@@ -539,7 +539,7 @@ pydantic==2.8.2
     # via pydantic-settings
     # via ragas
     # via sqlmodel
-pydantic-core==2.20.1
+pydantic-core==2.27.2
     # via pydantic
 pydantic-settings==2.6.1
     # via langchain-community