fix lindorm vdb bugs

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

@@ -49,10 +49,10 @@ def to_opensearch_params(self) -> dict[str, Any]:
 
 
 class LindormVectorStore(BaseVector):
-    def __init__(self, collection_name: str, config: LindormVectorStoreConfig, **kwargs):
+    def __init__(self, collection_name: str, config: LindormVectorStoreConfig, using_ugc: bool, **kwargs):
         self._routing = None
         self._routing_field = None
-        if config.using_ugc:
+        if using_ugc:
             routing_value: str = kwargs.get("routing_value")
             if routing_value is None:
                 raise ValueError("UGC index should init vector with valid 'routing_value' parameter value")
@@ -64,7 +64,7 @@ def __init__(self, collection_name: str, config: LindormVectorStoreConfig, **kwa
             super().__init__(collection_name.lower())
         self._client_config = config
         self._client = OpenSearch(**config.to_opensearch_params())
-        self._using_ugc = config.using_ugc
+        self._using_ugc = using_ugc
         self.kwargs = kwargs
 
     def get_type(self) -> str:
@@ -354,17 +354,17 @@ 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:
     if routing is not None:
         query_clause = {
@@ -410,17 +410,17 @@ 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 = "post_filter" if filter_type is None else filter_type
@@ -467,12 +467,16 @@ def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings
         using_ugc = dify_config.USING_UGC_INDEX
         routing_value = None
         if dataset.index_struct:
-            if using_ugc:
+            # if an existed record's index_struct_dict doesn't contain using_ugc field,
+            # it actually stores in the normal index format
+            stored_in_ugc = dataset.index_struct_dict.get("using_ugc", False)
+            using_ugc = stored_in_ugc
+            if stored_in_ugc:
                 dimension = dataset.index_struct_dict["dimension"]
                 index_type = dataset.index_struct_dict["index_type"]
                 distance_type = dataset.index_struct_dict["distance_type"]
-                index_name = f"{UGC_INDEX_PREFIX}_{dimension}_{index_type}_{distance_type}"
                 routing_value = dataset.index_struct_dict["vector_store"]["class_prefix"]
+                index_name = f"{UGC_INDEX_PREFIX}_{dimension}_{index_type}_{distance_type}"
             else:
                 index_name = dataset.index_struct_dict["vector_store"]["class_prefix"]
         else:
@@ -487,11 +491,12 @@ def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings
                 "index_type": index_type,
                 "dimension": dimension,
                 "distance_type": distance_type,
+                "using_ugc": using_ugc
             }
             dataset.index_struct = json.dumps(index_struct_dict)
             if using_ugc:
                 index_name = f"{UGC_INDEX_PREFIX}_{dimension}_{index_type}_{distance_type}"
                 routing_value = class_prefix
             else:
                 index_name = class_prefix
-        return LindormVectorStore(index_name, lindorm_config, routing_value=routing_value)
+        return LindormVectorStore(index_name, lindorm_config, routing_value=routing_value, using_ugc=using_ugc)