[Frontend] Expand tools even if tool_choice="none"

docs/features/tool_calling.md

@@ -97,6 +97,14 @@ vLLM supports the `tool_choice='required'` option in the chat completion API. Si
 
 When tool_choice='required' is set, the model is guaranteed to generate one or more tool calls based on the specified tool list in the `tools` parameter. The number of tool calls depends on the user's query. The output format strictly follows the schema defined in the `tools` parameter.
 
+## None Function Calling
+
+vLLM supports the `tool_choice='none'` option in the chat completion API. When this option is set, the model will not generate any tool calls and will respond with regular text content only, even if tools are defined in the request.
+
+By default, when `tool_choice='none'` is specified, vLLM excludes tool definitions from the prompt to optimize context usage. To include tool definitions even with `tool_choice='none'`, use the `--expand-tools-even-if-tool-choice-none` option.
+
+Note: This behavior will change in v0.10.0, where tool definitions will be included by default even with `tool_choice='none'`.
+
 ## Automatic Function Calling
 
 To enable this feature, you should set the following flags:

vllm/entrypoints/openai/api_server.py

@@ -1187,6 +1187,8 @@ async def init_app_state(
         chat_template_content_format=args.chat_template_content_format,
         return_tokens_as_token_ids=args.return_tokens_as_token_ids,
         enable_auto_tools=args.enable_auto_tool_choice,
+        expand_tools_even_if_tool_choice_none=args.
+        expand_tools_even_if_tool_choice_none,
         tool_parser=args.tool_call_parser,
         reasoning_parser=args.reasoning_parser,
         enable_prompt_tokens_details=args.enable_prompt_tokens_details,

vllm/entrypoints/openai/cli_args.py

@@ -223,6 +223,17 @@ def make_arg_parser(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
         default=False,
         help="Enable auto tool choice for supported models. Use "
         "``--tool-call-parser`` to specify which parser to use.")
+    parser.add_argument(
+        "--expand-tools-even-if-tool-choice-none",
+        action="store_true",
+        default=False,
+        deprecated=True,
+        help="Include tool definitions in prompts "
+        "even when tool_choice='none'. "
+        "This is a transitional option that will be removed in v0.10.0. "
+        "In v0.10.0, tool definitions will always be included regardless of "
+        "tool_choice setting. Use this flag now to test the new behavior "
+        "before the breaking change.")
 
     valid_tool_parsers = ToolParserManager.tool_parsers.keys()
     parser.add_argument(

vllm/entrypoints/openai/protocol.py

@@ -686,10 +686,8 @@ def check_tool_usage(cls, data):
         if "tool_choice" not in data and data.get("tools"):
             data["tool_choice"] = "auto"
 
-        # if "tool_choice" is "none" -- ignore tools if present
+        # if "tool_choice" is "none" -- no validation is needed for tools
         if "tool_choice" in data and data["tool_choice"] == "none":
-            # ensure that no tools are present
-            data.pop("tools", None)
             return data
 
         # if "tool_choice" is specified -- validation

vllm/entrypoints/openai/serving_chat.py

@@ -62,6 +62,7 @@ def __init__(
         return_tokens_as_token_ids: bool = False,
         reasoning_parser: str = "",
         enable_auto_tools: bool = False,
+        expand_tools_even_if_tool_choice_none: bool = False,
         tool_parser: Optional[str] = None,
         enable_prompt_tokens_details: bool = False,
     ) -> None:
@@ -108,6 +109,8 @@ def __init__(
                 raise TypeError("Error: --enable-auto-tool-choice requires "
                                 f"tool_parser:'{tool_parser}' which has not "
                                 "been registered") from e
+        self.expand_tools_even_if_tool_choice_none = (
+            expand_tools_even_if_tool_choice_none)
 
         self.enable_prompt_tokens_details = enable_prompt_tokens_details
         self.default_sampling_params = (
@@ -172,9 +175,24 @@ async def create_chat_completion(
                     "--enable-auto-tool-choice and --tool-call-parser to be set"
                 )
 
-            tool_dicts = None if request.tools is None else [
-                tool.model_dump() for tool in request.tools
-            ]
+            if request.tools is None:
+                tool_dicts = None
+            elif (request.tool_choice == "none"
+                  and not self.expand_tools_even_if_tool_choice_none):
+                if len(request.tools) > 0:
+                    logger.warning_once(
+                        "Tools are specified but tool_choice is set to 'none' "
+                        "and --expand-tools-even-if-tool-choice-none is not "
+                        "enabled. Tool definitions will be excluded from the "
+                        "prompt. This behavior will change in vLLM v0.10 where "
+                        "tool definitions will be included by default even "
+                        "with tool_choice='none'. To adopt the new behavior "
+                        "now, use --expand-tools-even-if-tool-choice-none. "
+                        "To suppress this warning, either remove tools from "
+                        "the request or set tool_choice to a different value.")
+                tool_dicts = None
+            else:
+                tool_dicts = [tool.model_dump() for tool in request.tools]
 
             (
                 conversation,