Merge pull request #2218 from danielaskdd/issue-2215

Refact: Improve query result with semantic null returns

lightrag/lightrag.py

@@ -21,6 +21,7 @@ from typing import (
     List,
     Dict,
 )
+from lightrag.prompt import PROMPTS
 from lightrag.constants import (
     DEFAULT_MAX_GLEANING,
     DEFAULT_FORCE_LLM_SUMMARY_ON_MERGE,
@@ -2287,20 +2288,35 @@ class LightRAG:
         else:
             raise ValueError(f"Unknown mode {data_param.mode}")
 
-        # Extract raw_data from QueryResult
-        final_data = query_result.raw_data if query_result else {}
-
-        # Log final result counts - adapt to new data format from convert_to_user_format
-        if final_data and "data" in final_data:
-            data_section = final_data["data"]
-            entities_count = len(data_section.get("entities", []))
-            relationships_count = len(data_section.get("relationships", []))
-            chunks_count = len(data_section.get("chunks", []))
-            logger.debug(
-                f"[aquery_data] Final result: {entities_count} entities, {relationships_count} relationships, {chunks_count} chunks"
-            )
+        if query_result is None:
+            no_result_message = "Query returned no results"
+            if data_param.mode == "naive":
+                no_result_message = "No relevant document chunks found."
+            final_data: dict[str, Any] = {
+                "status": "failure",
+                "message": no_result_message,
+                "data": {},
+                "metadata": {
+                    "failure_reason": "no_results",
+                    "mode": data_param.mode,
+                },
+            }
+            logger.info("[aquery_data] Query returned no results.")
         else:
-            logger.warning("[aquery_data] No data section found in query result")
+            # Extract raw_data from QueryResult
+            final_data = query_result.raw_data or {}
+
+            # Log final result counts - adapt to new data format from convert_to_user_format
+            if final_data and "data" in final_data:
+                data_section = final_data["data"]
+                entities_count = len(data_section.get("entities", []))
+                relationships_count = len(data_section.get("relationships", []))
+                chunks_count = len(data_section.get("chunks", []))
+                logger.debug(
+                    f"[aquery_data] Final result: {entities_count} entities, {relationships_count} relationships, {chunks_count} chunks"
+                )
+            else:
+                logger.warning("[aquery_data] No data section found in query result")
 
         await self._query_done()
         return final_data
@@ -2403,16 +2419,19 @@ class LightRAG:
                     "status": "failure",
                     "message": "Query returned no results",
                     "data": {},
-                    "metadata": {},
+                    "metadata": {
+                        "failure_reason": "no_results",
+                        "mode": param.mode,
+                    },
                     "llm_response": {
-                        "content": None,
+                        "content": PROMPTS["fail_response"],
                         "response_iterator": None,
                         "is_streaming": False,
                     },
                 }
 
             # Extract structured data from query result
-            raw_data = query_result.raw_data if query_result else {}
+            raw_data = query_result.raw_data or {}
             raw_data["llm_response"] = {
                 "content": query_result.content
                 if not query_result.is_streaming

lightrag/operate.py

@@ -2247,7 +2247,7 @@ async def kg_query(
     hashing_kv: BaseKVStorage | None = None,
     system_prompt: str | None = None,
     chunks_vdb: BaseVectorStorage = None,
-) -> QueryResult:
+) -> QueryResult | None:
     """
     Execute knowledge graph query and return unified QueryResult object.
 
@@ -2264,7 +2264,7 @@ async def kg_query(
         chunks_vdb: Document chunks vector database
 
     Returns:
-        QueryResult: Unified query result object containing:
+        QueryResult | None: Unified query result object containing:
             - content: Non-streaming response text content
             - response_iterator: Streaming response iterator
             - raw_data: Complete structured data (including references and metadata)
@@ -2275,6 +2275,8 @@ async def kg_query(
         - only_need_prompt=True: content contains complete prompt
         - stream=True: response_iterator contains streaming response, raw_data contains complete data
         - default: content contains LLM response text, raw_data contains complete data
+
+        Returns None when no relevant context could be constructed for the query.
     """
     if not query:
         return QueryResult(content=PROMPTS["fail_response"])
@@ -2322,7 +2324,8 @@ async def kg_query(
     )
 
     if context_result is None:
-        return QueryResult(content=PROMPTS["fail_response"])
+        logger.info("[kg_query] No query context could be built; returning no-result.")
+        return None
 
     # Return different content based on query parameters
     if query_param.only_need_context and not query_param.only_need_prompt:
@@ -3985,7 +3988,7 @@ async def naive_query(
     global_config: dict[str, str],
     hashing_kv: BaseKVStorage | None = None,
     system_prompt: str | None = None,
-) -> QueryResult:
+) -> QueryResult | None:
     """
     Execute naive query and return unified QueryResult object.
 
@@ -3998,11 +4001,13 @@ async def naive_query(
         system_prompt: System prompt
 
     Returns:
-        QueryResult: Unified query result object containing:
+        QueryResult | None: Unified query result object containing:
             - content: Non-streaming response text content
             - response_iterator: Streaming response iterator
             - raw_data: Complete structured data (including references and metadata)
             - is_streaming: Whether this is a streaming result
+
+        Returns None when no relevant chunks are retrieved.
     """
 
     if not query:
@@ -4023,16 +4028,10 @@ async def naive_query(
     chunks = await _get_vector_context(query, chunks_vdb, query_param, None)
 
     if chunks is None or len(chunks) == 0:
-        # Build empty raw data structure for naive mode
-        empty_raw_data = convert_to_user_format(
-            [],  # naive mode has no entities
-            [],  # naive mode has no relationships
-            [],  # no chunks
-            [],  # no references
-            "naive",
+        logger.info(
+            "[naive_query] No relevant document chunks found; returning no-result."
         )
-        empty_raw_data["message"] = "No relevant document chunks found."
-        return QueryResult(content=PROMPTS["fail_response"], raw_data=empty_raw_data)
+        return None
 
     # Calculate dynamic token limit for chunks
     max_total_tokens = getattr(