remove chunk_rerank_top_k

README-zh.md

@@ -295,15 +295,10 @@ class QueryParam:
     """Number of top items to retrieve. Represents entities in 'local' mode and relationships in 'global' mode."""
 
     chunk_top_k: int = int(os.getenv("CHUNK_TOP_K", "5"))
-    """Number of text chunks to retrieve initially from vector search.
+    """Number of text chunks to retrieve initially from vector search and keep after reranking.
     If None, defaults to top_k value.
     """
 
-    chunk_rerank_top_k: int = int(os.getenv("CHUNK_RERANK_TOP_K", "5"))
-    """Number of text chunks to keep after reranking.
-    If None, keeps all chunks returned from initial retrieval.
-    """
-
     max_entity_tokens: int = int(os.getenv("MAX_ENTITY_TOKENS", "10000"))
     """Maximum number of tokens allocated for entity context in unified token control system."""
 
@@ -340,6 +335,11 @@ class QueryParam:
     """User-provided prompt for the query.
     If proivded, this will be use instead of the default vaulue from prompt template.
     """
+
+    enable_rerank: bool = True
+    """Enable reranking for retrieved text chunks. If True but no rerank model is configured, a warning will be issued.
+    Default is True to enable reranking when rerank model is available.
+    """
 ```
 
 > top_k的默认值可以通过环境变量TOP_K更改。

README.md

@@ -302,15 +302,10 @@ class QueryParam:
     """Number of top items to retrieve. Represents entities in 'local' mode and relationships in 'global' mode."""
 
     chunk_top_k: int = int(os.getenv("CHUNK_TOP_K", "5"))
-    """Number of text chunks to retrieve initially from vector search.
+    """Number of text chunks to retrieve initially from vector search and keep after reranking.
     If None, defaults to top_k value.
     """
 
-    chunk_rerank_top_k: int = int(os.getenv("CHUNK_RERANK_TOP_K", "5"))
-    """Number of text chunks to keep after reranking.
-    If None, keeps all chunks returned from initial retrieval.
-    """
-
     max_entity_tokens: int = int(os.getenv("MAX_ENTITY_TOKENS", "10000"))
     """Maximum number of tokens allocated for entity context in unified token control system."""
 
@@ -341,6 +336,11 @@ class QueryParam:
     """User-provided prompt for the query.
     If proivded, this will be use instead of the default vaulue from prompt template.
     """
+
+    enable_rerank: bool = True
+    """Enable reranking for retrieved text chunks. If True but no rerank model is configured, a warning will be issued.
+    Default is True to enable reranking when rerank model is available.
+    """
 ```
 
 > default value of Top_k can be change by environment  variables  TOP_K.

env.example

@@ -58,14 +58,13 @@ OLLAMA_EMULATING_MODEL_TAG=latest
 # COSINE_THRESHOLD=0.2
 ### Number of entities or relations to retrieve from KG
 # TOP_K=60
-### Number of text chunks to retrieve initially from vector search
+### Number of text chunks to retrieve initially from vector search and keep after reranking
 # CHUNK_TOP_K=5
 
 ### Rerank Configuration
-# ENABLE_RERANK=False
-### Number of text chunks to keep after reranking (should be <= CHUNK_TOP_K)
-# CHUNK_RERANK_TOP_K=5
-### Rerank model configuration (required when ENABLE_RERANK=True)
+### Note: Reranking is now controlled per query via the 'enable_rerank' parameter (default: true)
+### The following configuration is only needed when you want to use reranking
+### Rerank model configuration (required when enable_rerank=true in query parameters)
 # RERANK_MODEL=BAAI/bge-reranker-v2-m3
 # RERANK_BINDING_HOST=https://api.your-rerank-provider.com/v1/rerank
 # RERANK_BINDING_API_KEY=your_rerank_api_key_here

lightrag/api/config.py

@@ -168,20 +168,8 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--chunk-top-k",
         type=int,
-        default=get_env_value("CHUNK_TOP_K", 15, int),
-        help="Number of text chunks to retrieve initially from vector search (default: from env or 15)",
-    )
-    parser.add_argument(
-        "--chunk-rerank-top-k",
-        type=int,
-        default=get_env_value("CHUNK_RERANK_TOP_K", 5, int),
-        help="Number of text chunks to keep after reranking (default: from env or 5)",
-    )
-    parser.add_argument(
-        "--enable-rerank",
-        action="store_true",
-        default=get_env_value("ENABLE_RERANK", False, bool),
-        help="Enable rerank functionality (default: from env or False)",
+        default=get_env_value("CHUNK_TOP_K", 5, int),
+        help="Number of text chunks to retrieve initially from vector search and keep after reranking (default: from env or 5)",
     )
     parser.add_argument(
         "--cosine-threshold",

lightrag/api/lightrag_server.py

@@ -292,9 +292,9 @@ def create_app(args):
         ),
     )
 
-    # Configure rerank function if enabled
+    # Configure rerank function if model and API are configured
     rerank_model_func = None
-    if args.enable_rerank and args.rerank_binding_api_key and args.rerank_binding_host:
+    if args.rerank_binding_api_key and args.rerank_binding_host:
         from lightrag.rerank import custom_rerank
 
         async def server_rerank_func(
@@ -312,10 +312,12 @@ def create_app(args):
             )
 
         rerank_model_func = server_rerank_func
-        logger.info(f"Rerank enabled with model: {args.rerank_model}")
-    elif args.enable_rerank:
-        logger.warning(
-            "Rerank enabled but RERANK_BINDING_API_KEY or RERANK_BINDING_HOST not configured. Rerank will be disabled."
+        logger.info(
+            f"Rerank model configured: {args.rerank_model} (can be enabled per query)"
+        )
+    else:
+        logger.info(
+            "Rerank model not configured. Set RERANK_BINDING_API_KEY and RERANK_BINDING_HOST to enable reranking."
         )
 
     # Initialize RAG
@@ -351,7 +353,6 @@ def create_app(args):
             },
             enable_llm_cache_for_entity_extract=args.enable_llm_cache_for_extract,
             enable_llm_cache=args.enable_llm_cache,
-            enable_rerank=args.enable_rerank,
             rerank_model_func=rerank_model_func,
             auto_manage_storages_states=False,
             max_parallel_insert=args.max_parallel_insert,
@@ -381,7 +382,6 @@ def create_app(args):
             },
             enable_llm_cache_for_entity_extract=args.enable_llm_cache_for_extract,
             enable_llm_cache=args.enable_llm_cache,
-            enable_rerank=args.enable_rerank,
             rerank_model_func=rerank_model_func,
             auto_manage_storages_states=False,
             max_parallel_insert=args.max_parallel_insert,
@@ -512,11 +512,13 @@ def create_app(args):
                     "enable_llm_cache": args.enable_llm_cache,
                     "workspace": args.workspace,
                     "max_graph_nodes": args.max_graph_nodes,
-                    # Rerank configuration
-                    "enable_rerank": args.enable_rerank,
-                    "rerank_model": args.rerank_model if args.enable_rerank else None,
+                    # Rerank configuration (based on whether rerank model is configured)
+                    "enable_rerank": rerank_model_func is not None,
+                    "rerank_model": args.rerank_model
+                    if rerank_model_func is not None
+                    else None,
                     "rerank_binding_host": args.rerank_binding_host
-                    if args.enable_rerank
+                    if rerank_model_func is not None
                     else None,
                 },
                 "auth_mode": auth_mode,

lightrag/api/routers/query_routes.py

@@ -52,13 +52,7 @@ class QueryRequest(BaseModel):
     chunk_top_k: Optional[int] = Field(
         ge=1,
         default=None,
-        description="Number of text chunks to retrieve initially from vector search.",
-    )
-
-    chunk_rerank_top_k: Optional[int] = Field(
-        ge=1,
-        default=None,
-        description="Number of text chunks to keep after reranking.",
+        description="Number of text chunks to retrieve initially from vector search and keep after reranking.",
     )
 
     max_entity_tokens: Optional[int] = Field(
@@ -99,6 +93,11 @@ class QueryRequest(BaseModel):
         description="User-provided prompt for the query. If provided, this will be used instead of the default value from prompt template.",
     )
 
+    enable_rerank: Optional[bool] = Field(
+        default=None,
+        description="Enable reranking for retrieved text chunks. If True but no rerank model is configured, a warning will be issued. Default is True.",
+    )
+
     @field_validator("query", mode="after")
     @classmethod
     def query_strip_after(cls, query: str) -> str: