add param to set query timeout when do vector/full-text search

api/.env.example

@@ -297,6 +297,7 @@ LINDORM_URL=http://ld-*******************-proxy-search-pub.lindorm.aliyuncs.com:
 LINDORM_USERNAME=admin
 LINDORM_PASSWORD=admin
 USING_UGC_INDEX=False
+QUERY_TIMEOUT=1
 
 # OceanBase Vector configuration
 OCEANBASE_VECTOR_HOST=127.0.0.1

api/configs/middleware/vdb/lindorm_config.py

@@ -32,3 +32,8 @@ class LindormConfig(BaseSettings):
         description="Using UGC index will store the same type of Index in a single index but can retrieve separately.",
         default=False,
     )
+    QUERY_TIMEOUT: Optional[float] = Field(
+        description="The lindorm search request timeout (s)",
+        default=2.0
+    )
+

api/core/rag/datasource/vdb/lindorm/lindorm_vector.py

@@ -32,6 +32,7 @@ class LindormVectorStoreConfig(BaseModel):
     username: Optional[str] = None
     password: Optional[str] = None
     using_ugc: Optional[bool] = False
+    request_timeout: Optional[float] = 1.0      # timeout units: s
 
     @model_validator(mode="before")
     @classmethod
@@ -81,12 +82,12 @@ class LindormVectorStore(BaseVector):
         self._client.indices.refresh(index=self._collection_name)
 
     def add_texts(
-        self,
-        documents: list[Document],
-        embeddings: list[list[float]],
-        batch_size: int = 64,
-        timeout: int = 60,
-        **kwargs,
+            self,
+            documents: list[Document],
+            embeddings: list[list[float]],
+            batch_size: int = 64,
+            timeout: int = 60,
+            **kwargs,
     ):
         logger.info(f"Total documents to add: {len(documents)}")
         uuids = self._get_uuids(documents)
@@ -251,7 +252,7 @@ class LindormVectorStore(BaseVector):
         query = default_vector_search_query(query_vector=query_vector, k=top_k, filters=filters, **kwargs)
 
         try:
-            params = {}
+            params = {"timeout": self._client_config.request_timeout}
             if self._using_ugc:
                 params["routing"] = self._routing
             response = self._client.search(index=self._collection_name, body=query, params=params)
@@ -304,8 +305,8 @@ class LindormVectorStore(BaseVector):
             routing=routing,
             routing_field=self._routing_field,
         )
-
-        response = self._client.search(index=self._collection_name, body=full_text_query)
+        params = {"timeout": self._client_config.request_timeout}
+        response = self._client.search(index=self._collection_name, body=full_text_query, params=params)
         docs = []
         for hit in response["hits"]["hits"]:
             docs.append(
@@ -444,17 +445,17 @@ def default_text_mapping(dimension: int, method_name: str, **kwargs: Any) -> dic
 
 
 def default_text_search_query(
-    query_text: str,
-    k: int = 4,
-    text_field: str = Field.CONTENT_KEY.value,
-    must: Optional[list[dict]] = None,
-    must_not: Optional[list[dict]] = None,
-    should: Optional[list[dict]] = None,
-    minimum_should_match: int = 0,
-    filters: Optional[list[dict]] = None,
-    routing: Optional[str] = None,
-    routing_field: Optional[str] = None,
-    **kwargs,
+        query_text: str,
+        k: int = 4,
+        text_field: str = Field.CONTENT_KEY.value,
+        must: Optional[list[dict]] = None,
+        must_not: Optional[list[dict]] = None,
+        should: Optional[list[dict]] = None,
+        minimum_should_match: int = 0,
+        filters: Optional[list[dict]] = None,
+        routing: Optional[str] = None,
+        routing_field: Optional[str] = None,
+        **kwargs,
 ) -> dict:
     query_clause: dict[str, Any] = {}
     if routing is not None:
@@ -501,17 +502,17 @@ def default_text_search_query(
 
 
 def default_vector_search_query(
-    query_vector: list[float],
-    k: int = 4,
-    min_score: str = "0.0",
-    ef_search: Optional[str] = None,  # only for hnsw
-    nprobe: Optional[str] = None,  # "2000"
-    reorder_factor: Optional[str] = None,  # "20"
-    client_refactor: Optional[str] = None,  # "true"
-    vector_field: str = Field.VECTOR.value,
-    filters: Optional[list[dict]] = None,
-    filter_type: Optional[str] = None,
-    **kwargs,
+        query_vector: list[float],
+        k: int = 4,
+        min_score: str = "0.0",
+        ef_search: Optional[str] = None,  # only for hnsw
+        nprobe: Optional[str] = None,  # "2000"
+        reorder_factor: Optional[str] = None,  # "20"
+        client_refactor: Optional[str] = None,  # "true"
+        vector_field: str = Field.VECTOR.value,
+        filters: Optional[list[dict]] = None,
+        filter_type: Optional[str] = None,
+        **kwargs,
 ) -> dict:
     if filters is not None:
         filter_type = "pre_filter" if filter_type is None else filter_type
@@ -554,6 +555,7 @@ class LindormVectorStoreFactory(AbstractVectorFactory):
             username=dify_config.LINDORM_USERNAME,
             password=dify_config.LINDORM_PASSWORD,
             using_ugc=dify_config.USING_UGC_INDEX,
+            request_timeout=dify_config.QUERY_TIMEOUT
         )
         using_ugc = dify_config.USING_UGC_INDEX
         if using_ugc is None:

docker/.env.example

@@ -553,6 +553,7 @@ VIKINGDB_SOCKET_TIMEOUT=30
 LINDORM_URL=http://lindorm:30070
 LINDORM_USERNAME=lindorm
 LINDORM_PASSWORD=lindorm
+QUERY_TIMEOUT=1
 
 # OceanBase Vector configuration, only available when VECTOR_STORE is `oceanbase`
 OCEANBASE_VECTOR_HOST=oceanbase