[Frontend] OpenAI Responses API supports Tool/Function calling

Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>

Author: chaunceyjiang

vllm/entrypoints/openai/api_server.py

@@ -1631,6 +1631,8 @@ 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,
+        exclude_tools_when_tool_choice_none=args.
+        exclude_tools_when_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/serving_engine.py

@@ -68,7 +68,6 @@
     MultiModalDataDict)
 from vllm.outputs import PoolingRequestOutput, RequestOutput
 from vllm.pooling_params import PoolingParams
-from vllm.prompt_adapter.request import PromptAdapterRequest
 from vllm.reasoning import ReasoningParser, ReasoningParserManager
 from vllm.sampling_params import BeamSearchParams, SamplingParams
 from vllm.sequence import Logprob, PromptLogprobs

vllm/entrypoints/openai/serving_responses.py

@@ -60,6 +60,7 @@ def __init__(
         return_tokens_as_token_ids: bool = False,
         reasoning_parser: str = "",
         enable_auto_tools: bool = False,
+        exclude_tools_when_tool_choice_none: bool = False,
         tool_parser: Optional[str] = None,
         enable_prompt_tokens_details: bool = False,
         enable_force_include_usage: bool = False,
@@ -81,6 +82,9 @@ def __init__(
 
         self.tool_parser = self._get_tool_parser(
             tool_parser_name=tool_parser, enable_auto_tools=enable_auto_tools)
+        self.exclude_tools_when_tool_choice_none = (
+            exclude_tools_when_tool_choice_none
+        )
         self.enable_prompt_tokens_details = enable_prompt_tokens_details
         self.enable_force_include_usage = enable_force_include_usage
         self.default_sampling_params = (
@@ -138,7 +142,9 @@ async def create_responses(
             lora_request = self._maybe_get_adapters(request)
             model_name = self._get_model_name(request.model, lora_request)
             tokenizer = await self.engine_client.get_tokenizer(lora_request)
-            if request.tools is None:
+            if (request.tools is None or
+                (request.tool_choice == "none"
+                        and self.exclude_tools_when_tool_choice_none)):
                 tool_dicts = None
             else:
                 tool_dicts = [tool.model_dump() for tool in request.tools]

vllm/entrypoints/openai/tool_parsers/glm4_moe_tool_parser.py

@@ -6,12 +6,15 @@
 from typing import Union
 
 import regex as re
-
+# yapf conflicts with isort for this block
+# yapf: disable
 from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
                                               DeltaFunctionCall, DeltaMessage,
                                               DeltaToolCall,
                                               ExtractedToolCallInformation,
-                                              FunctionCall, ToolCall)
+                                              FunctionCall, ResponsesRequest,
+                                              ToolCall)
+# yapf: enable
 from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
     ToolParser, ToolParserManager)
 from vllm.logger import init_logger
@@ -90,7 +93,7 @@ def _parse_arguments(self, args_text: str) -> str:
     def extract_tool_calls(
         self,
         model_output: str,
-        request: ChatCompletionRequest,
+        request: Union[ChatCompletionRequest, ResponsesRequest],
     ) -> ExtractedToolCallInformation:
 
         # sanity check; avoid unnecessary processing

vllm/entrypoints/openai/tool_parsers/hunyuan_a13b_tool_parser.py

@@ -7,12 +7,15 @@
 from typing import Any, Optional, Union
 
 import regex as re
-
+# yapf conflicts with isort for this block
+# yapf: disable
 from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
                                               DeltaFunctionCall, DeltaMessage,
                                               DeltaToolCall,
                                               ExtractedToolCallInformation,
-                                              FunctionCall, ToolCall)
+                                              FunctionCall, ResponsesRequest,
+                                              ToolCall)
+# yapf: enable
 from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
     ToolParser, ToolParserManager)
 from vllm.entrypoints.openai.tool_parsers.utils import consume_space
@@ -86,8 +89,10 @@ def preprocess_model_output(
         return model_output, None
 
     def extract_tool_calls(
-            self, model_output: str,
-            request: ChatCompletionRequest) -> ExtractedToolCallInformation:
+        self,
+        model_output: str,
+        request: Union[ChatCompletionRequest, ResponsesRequest],
+    ) -> ExtractedToolCallInformation:
         """
         Extract tool calls from a complete model output.
         """

vllm/entrypoints/openai/tool_parsers/qwen3coder_tool_parser.py

@@ -7,13 +7,16 @@
 from typing import Any, Optional, Union
 
 import regex as re
-
+# yapf conflicts with isort for this block
+# yapf: disable
 from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
                                               ChatCompletionToolsParam,
                                               DeltaFunctionCall, DeltaMessage,
                                               DeltaToolCall,
                                               ExtractedToolCallInformation,
-                                              FunctionCall, ToolCall)
+                                              FunctionCall, ResponsesRequest,
+                                              ToolCall)
+# yapf: enable
 from vllm.entrypoints.openai.tool_parsers.abstract_tool_parser import (
     ToolParser, ToolParserManager)
 from vllm.logger import init_logger
@@ -269,7 +272,7 @@ def _get_function_calls(self, model_output: str) -> list[str]:
     def extract_tool_calls(
         self,
         model_output: str,
-        request: ChatCompletionRequest,
+        request: Union[ChatCompletionRequest, ResponsesRequest],
     ) -> ExtractedToolCallInformation:
         # Quick check to avoid unnecessary processing
         if self.tool_call_prefix not in model_output: