[Refactor] [11/N] to simplify the mcp architecture (#32396)

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

tests/entrypoints/openai/responses/test_mcp_tools.py

@@ -7,7 +7,7 @@ import pytest_asyncio
 from openai import OpenAI
 from openai_harmony import ToolDescription, ToolNamespaceConfig
 
-from vllm.entrypoints.tool_server import MCPToolServer
+from vllm.entrypoints.mcp.tool_server import MCPToolServer
 
 from ....utils import RemoteOpenAIServer
 

tests/entrypoints/openai/test_gptoss_structural_tags_integration.py

@@ -8,7 +8,7 @@ from unittest.mock import Mock
 
 import pytest
 
-from vllm.entrypoints.tool_server import ToolServer
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.reasoning.gptoss_reasoning_parser import (
     GptOssReasoningParser,
 )

tests/entrypoints/openai/test_serving_responses.py

@@ -13,15 +13,15 @@ from openai.types.responses.tool import (
     Tool,
 )
 
-from vllm.entrypoints.context import ConversationContext
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.entrypoints.openai.engine.protocol import ErrorResponse
+from vllm.entrypoints.openai.responses.context import ConversationContext
 from vllm.entrypoints.openai.responses.protocol import ResponsesRequest
 from vllm.entrypoints.openai.responses.serving import (
     OpenAIServingResponses,
     _extract_allowed_tools_from_mcp_requests,
     extract_tool_types,
 )
-from vllm.entrypoints.tool_server import ToolServer
 from vllm.inputs.data import TokensPrompt
 
 

tests/entrypoints/test_context.py

@@ -6,7 +6,7 @@ from unittest.mock import MagicMock, patch
 import pytest
 from openai_harmony import Author, Message, Role, StreamState, TextContent
 
-from vllm.entrypoints.context import (
+from vllm.entrypoints.openai.responses.context import (
     HarmonyContext,
     StreamingHarmonyContext,
     TurnMetrics,
@@ -71,7 +71,7 @@ async def generate_mock_outputs(
 def mock_parser():
     """Set up a mock parser for tests."""
     with patch(
-        "vllm.entrypoints.context.get_streamable_parser_for_assistant"
+        "vllm.entrypoints.openai.responses.context.get_streamable_parser_for_assistant"
     ) as mock_parser_factory:
         # Create a mock parser object
         parser = MagicMock()
@@ -284,7 +284,7 @@ async def test_negative_tool_tokens_edge_case():
     """Test edge case where calculation could result in negative tool
     tokens. We should log an error and clamp the value to 0."""
     # Use patch to check if logger.error was called
-    with patch("vllm.entrypoints.context.logger.error") as mock_log:
+    with patch("vllm.entrypoints.openai.responses.context.logger.error") as mock_log:
         context = HarmonyContext(messages=[], available_tools=["browser"])
 
         # First turn

tests/entrypoints/test_responses_utils.py

@@ -16,7 +16,7 @@ from openai.types.responses.response_reasoning_item import (
 )
 
 from vllm.entrypoints.constants import MCP_PREFIX
-from vllm.entrypoints.responses_utils import (
+from vllm.entrypoints.openai.responses.utils import (
     _construct_single_message_from_response_item,
     _maybe_combine_reasoning_and_tool_call,
     construct_chat_messages_with_tool_call,

tests/v1/structured_output/test_gptoss_structural_tags.py

@@ -8,7 +8,7 @@ from unittest.mock import Mock
 
 import pytest
 
-from vllm.entrypoints.tool_server import ToolServer
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.reasoning.gptoss_reasoning_parser import (
     GptOssReasoningParser,
     from_builtin_tool_to_tag,

vllm/entrypoints/mcp/__init__.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project

vllm/entrypoints/tool.py → vllm/entrypoints/mcp/tool.py

@@ -15,7 +15,7 @@ from vllm.utils import random_uuid
 
 if TYPE_CHECKING:
     # Avoid circular import.
-    from vllm.entrypoints.context import ConversationContext
+    from vllm.entrypoints.openai.responses.context import ConversationContext
 
 logger = init_logger(__name__)
 
@@ -81,7 +81,7 @@ class HarmonyBrowserTool(Tool):
         logger.info_once("Browser tool initialized")
 
     async def get_result(self, context: "ConversationContext") -> Any:
-        from vllm.entrypoints.context import HarmonyContext
+        from vllm.entrypoints.openai.responses.context import HarmonyContext
 
         assert isinstance(context, HarmonyContext)
         last_msg = context.messages[-1]
@@ -141,7 +141,7 @@ class HarmonyPythonTool(Tool):
         logger.info_once("Code interpreter tool initialized")
 
     async def get_result(self, context: "ConversationContext") -> Any:
-        from vllm.entrypoints.context import HarmonyContext
+        from vllm.entrypoints.openai.responses.context import HarmonyContext
 
         assert isinstance(context, HarmonyContext)
         last_msg = context.messages[-1]
@@ -155,7 +155,7 @@ class HarmonyPythonTool(Tool):
         This function converts parsable context types to harmony and
         back so we can use GPTOSS demo python tool
         """
-        from vllm.entrypoints.context import ParsableContext
+        from vllm.entrypoints.openai.responses.context import ParsableContext
 
         assert isinstance(context, ParsableContext)
 

vllm/entrypoints/tool_server.py → vllm/entrypoints/mcp/tool_server.py

@@ -6,7 +6,7 @@ from typing import TYPE_CHECKING, Any
 
 from openai_harmony import ToolDescription, ToolNamespaceConfig
 
-from vllm.entrypoints.tool import HarmonyBrowserTool, HarmonyPythonTool, Tool
+from vllm.entrypoints.mcp.tool import HarmonyBrowserTool, HarmonyPythonTool, Tool
 from vllm.logger import init_logger
 
 logger = init_logger(__name__)

vllm/entrypoints/openai/api_server.py

@@ -36,6 +36,7 @@ from vllm.engine.protocol import EngineClient
 from vllm.entrypoints.anthropic.serving import AnthropicServingMessages
 from vllm.entrypoints.launcher import serve_http
 from vllm.entrypoints.logger import RequestLogger
+from vllm.entrypoints.mcp.tool_server import DemoToolServer, MCPToolServer, ToolServer
 from vllm.entrypoints.openai.chat_completion.serving import OpenAIServingChat
 from vllm.entrypoints.openai.cli_args import make_arg_parser, validate_parsed_serve_args
 from vllm.entrypoints.openai.completion.serving import OpenAIServingCompletion
@@ -62,7 +63,6 @@ from vllm.entrypoints.serve.elastic_ep.middleware import (
     ScalingMiddleware,
 )
 from vllm.entrypoints.serve.tokenize.serving import OpenAIServingTokenization
-from vllm.entrypoints.tool_server import DemoToolServer, MCPToolServer, ToolServer
 from vllm.entrypoints.utils import (
     cli_env_setup,
     log_non_default_args,

vllm/entrypoints/openai/engine/serving.py

@@ -31,12 +31,6 @@ from vllm.entrypoints.chat_utils import (
     parse_chat_messages_futures,
     resolve_chat_template_content_format,
 )
-from vllm.entrypoints.context import (
-    ConversationContext,
-    HarmonyContext,
-    ParsableContext,
-    StreamingHarmonyContext,
-)
 from vllm.entrypoints.logger import RequestLogger
 from vllm.entrypoints.openai.chat_completion.protocol import (
     ChatCompletionNamedToolChoiceParam,
@@ -54,10 +48,19 @@ from vllm.entrypoints.openai.engine.protocol import (
     FunctionDefinition,
 )
 from vllm.entrypoints.openai.models.serving import OpenAIServingModels
+from vllm.entrypoints.openai.responses.context import (
+    ConversationContext,
+    HarmonyContext,
+    ParsableContext,
+    StreamingHarmonyContext,
+)
 from vllm.entrypoints.openai.responses.protocol import (
     ResponseInputOutputItem,
     ResponsesRequest,
 )
+from vllm.entrypoints.openai.responses.utils import (
+    construct_input_messages,
+)
 from vllm.entrypoints.openai.translations.protocol import (
     TranscriptionRequest,
     TranscriptionResponse,
@@ -85,9 +88,6 @@ from vllm.entrypoints.pooling.score.protocol import (
     ScoreResponse,
 )
 from vllm.entrypoints.renderer import BaseRenderer, CompletionRenderer, RenderConfig
-from vllm.entrypoints.responses_utils import (
-    construct_input_messages,
-)
 from vllm.entrypoints.serve.disagg.protocol import GenerateRequest, GenerateResponse
 from vllm.entrypoints.serve.tokenize.protocol import (
     DetokenizeRequest,

vllm/entrypoints/context.py → vllm/entrypoints/openai/responses/context.py

@@ -22,6 +22,8 @@ from vllm.entrypoints.chat_utils import (
     ChatTemplateContentFormatOption,
 )
 from vllm.entrypoints.constants import MCP_PREFIX
+from vllm.entrypoints.mcp.tool import Tool
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.entrypoints.openai.engine.protocol import (
     FunctionCall,
 )
@@ -38,9 +40,7 @@ from vllm.entrypoints.openai.responses.protocol import (
     ResponseRawMessageAndToken,
     ResponsesRequest,
 )
-from vllm.entrypoints.responses_utils import construct_tool_dicts
-from vllm.entrypoints.tool import Tool
-from vllm.entrypoints.tool_server import ToolServer
+from vllm.entrypoints.openai.responses.utils import construct_tool_dicts
 from vllm.outputs import RequestOutput
 from vllm.reasoning.abs_reasoning_parsers import ReasoningParser
 from vllm.tokenizers import TokenizerLike

vllm/entrypoints/openai/responses/serving.py

@@ -64,14 +64,8 @@ from vllm.entrypoints.chat_utils import (
     ChatCompletionMessageParam,
     ChatTemplateContentFormatOption,
 )
-from vllm.entrypoints.context import (
-    ConversationContext,
-    HarmonyContext,
-    ParsableContext,
-    SimpleContext,
-    StreamingHarmonyContext,
-)
 from vllm.entrypoints.logger import RequestLogger
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.entrypoints.openai.engine.protocol import (
     DeltaMessage,
     ErrorResponse,
@@ -94,6 +88,13 @@ from vllm.entrypoints.openai.parser.harmony_utils import (
     parse_response_input,
     render_for_completion,
 )
+from vllm.entrypoints.openai.responses.context import (
+    ConversationContext,
+    HarmonyContext,
+    ParsableContext,
+    SimpleContext,
+    StreamingHarmonyContext,
+)
 from vllm.entrypoints.openai.responses.protocol import (
     InputTokensDetails,
     OutputTokensDetails,
@@ -108,13 +109,12 @@ from vllm.entrypoints.openai.responses.protocol import (
     ResponseUsage,
     StreamingResponsesResponse,
 )
-from vllm.entrypoints.responses_utils import (
+from vllm.entrypoints.openai.responses.utils import (
     construct_input_messages,
     construct_tool_dicts,
     extract_tool_types,
     should_continue_final_message,
 )
-from vllm.entrypoints.tool_server import ToolServer
 from vllm.exceptions import VLLMValidationError
 from vllm.inputs.data import TokensPrompt
 from vllm.logger import init_logger

vllm/reasoning/abs_reasoning_parsers.py

@@ -8,7 +8,7 @@ from collections.abc import Callable, Sequence
 from functools import cached_property
 from typing import TYPE_CHECKING, Any
 
-from vllm.entrypoints.tool_server import ToolServer
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.logger import init_logger
 from vllm.utils.collection_utils import is_list_of
 from vllm.utils.import_utils import import_from_path

vllm/reasoning/gptoss_reasoning_parser.py

@@ -5,12 +5,12 @@ from collections.abc import Sequence
 
 from transformers import PreTrainedTokenizerBase
 
+from vllm.entrypoints.mcp.tool_server import ToolServer
 from vllm.entrypoints.openai.chat_completion.protocol import (
     ChatCompletionRequest,
 )
 from vllm.entrypoints.openai.engine.protocol import DeltaMessage
 from vllm.entrypoints.openai.parser.harmony_utils import parse_chat_output
-from vllm.entrypoints.tool_server import ToolServer
 from vllm.logger import init_logger
 from vllm.reasoning import ReasoningParser