search_mlx_models

Instructions

Implementation Reference

• src/mcp_mlx_launcher/server.py:131-190 (handler)

  Main handler for the search_mlx_models tool. Extracts arguments (search_query, limit), validates them, then uses HfApi.list_models with tag 'mlx' and sort by downloads to fetch results. Runs the blocking HF API call via asyncio.to_thread. Returns a JSON list of model IDs, downloads, and likes.

  @server.call_tool()
  async def handle_call_tool(
      name: str, arguments: dict[str, Any] | None
  ) -> list[types.TextContent | types.ImageContent | types.EmbeddedResource]:
      """AIエージェントから呼び出されたツールを実際に実行します"""
      if arguments is None and name not in ("list_running_servers", "search_mlx_models", "check_system_environment"):
          raise ValueError("Arguments are required")
  
      if name == "check_system_environment":
          info = process_manager.get_system_info()
          return [types.TextContent(type="text", text=json.dumps(info, indent=2))]
  
      elif name == "check_llm_status":
          port = arguments.get("port")
          if not isinstance(port, int):
              raise ValueError("Port must be an integer")
          
          is_running = process_manager.is_port_in_use(port)
          return [types.TextContent(type="text", text=str(is_running).lower())]
  
      elif name == "list_running_servers":
          servers = process_manager.get_running_servers()
          if not servers:
              return [types.TextContent(type="text", text="No running servers found.")]
          return [types.TextContent(type="text", text=json.dumps(servers, indent=2))]
  
      elif name == "search_mlx_models":
          query = arguments.get("search_query") if arguments else None
          limit = arguments.get("limit", 10) if arguments else 10
  
          if limit is not None and not isinstance(limit, int):
              raise ValueError("limit must be an integer")
          if query is not None and not isinstance(query, str):
              raise ValueError("search_query must be a string")
  
          def _search():
              api = HfApi()
              models = api.list_models(
                  search=query if query else None,
                  tags="mlx",
                  sort="downloads",
                  direction=-1,
                  limit=limit
              )
              results = []
              for m in models:
                  results.append({
                      "modelId": m.id,
                      "downloads": getattr(m, 'downloads', 0),
                      "likes": getattr(m, 'likes', 0)
                  })
              return results
  
          try:
              results = await asyncio.to_thread(_search)
              if not results:
                  return [types.TextContent(type="text", text="No MLX models found.")]
              return [types.TextContent(type="text", text=json.dumps(results, indent=2))]
          except Exception as e:
              return [types.TextContent(type="text", text=f"Error searching models: {str(e)}")]

• src/mcp_mlx_launcher/server.py:48-63 (schema)

  Registration and input schema for search_mlx_models. Defines the tool name, description, and inputSchema with optional 'search_query' (string) and 'limit' (integer, default 10) properties.

  types.Tool(
      name="search_mlx_models",
      description="Hugging Faceからダウンロード可能なMLXフォーマットのLLMモデルを検索・リストアップします。",
      inputSchema={
          "type": "object",
          "properties": {
              "search_query": {
                  "type": "string",
                  "description": "検索キーワード（例: 'llama', 'qwen'）。未指定の場合は人気のMLXモデルを返します。"
              },
              "limit": {
                  "type": "integer",
                  "description": "取得する最大件数。デフォルトは10。"
              }
          },
      },

• src/mcp_mlx_launcher/server.py:48-64 (registration)

  Tool registration inside handle_list_tools(). The search_mlx_models tool is listed alongside 7 other tools in the MCP tool registry.

  types.Tool(
      name="search_mlx_models",
      description="Hugging Faceからダウンロード可能なMLXフォーマットのLLMモデルを検索・リストアップします。",
      inputSchema={
          "type": "object",
          "properties": {
              "search_query": {
                  "type": "string",
                  "description": "検索キーワード（例: 'llama', 'qwen'）。未指定の場合は人気のMLXモデルを返します。"
              },
              "limit": {
                  "type": "integer",
                  "description": "取得する最大件数。デフォルトは10。"
              }
          },
      },
  ),

• src/mcp_mlx_launcher/server.py:136-137 (registration)

  Special registration: search_mlx_models is exempted from the 'arguments required' check, since both search_query and limit are optional.

  if arguments is None and name not in ("list_running_servers", "search_mlx_models", "check_system_environment"):
      raise ValueError("Arguments are required")