LLMToolSelectorMiddleware Class — langchain Architecture

Architecture documentation for the LLMToolSelectorMiddleware class in tool_selection.py from the langchain codebase.

Class python

Entity Profile

Dependency Diagram

graph TD
  888c953d_8f6f_ac7f_4274_ca04589e186a["LLMToolSelectorMiddleware"]
  2769f2d0_e2a8_5ab6_dc35_d5d0d6406ac4["HumanMessage"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|extends| 2769f2d0_e2a8_5ab6_dc35_d5d0d6406ac4
  740a7da1_ca4f_8c85_71e0_4f92696f1237["tool_selection.py"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|defined in| 740a7da1_ca4f_8c85_71e0_4f92696f1237
  74e81f5f_ce46_4e5d_763f_ed260d0dbec3["__init__()"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|method| 74e81f5f_ce46_4e5d_763f_ed260d0dbec3
  bdf758af_80be_0125_ef7b_855624a8a30a["_prepare_selection_request()"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|method| bdf758af_80be_0125_ef7b_855624a8a30a
  963a6e36_74b3_e68c_c642_f81d47a7666b["_process_selection_response()"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|method| 963a6e36_74b3_e68c_c642_f81d47a7666b
  79ce050f_9134_368b_2f3c_59dd0f33524c["wrap_model_call()"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|method| 79ce050f_9134_368b_2f3c_59dd0f33524c
  914b6a96_704c_ef23_6236_b6d24e64245f["awrap_model_call()"]
  888c953d_8f6f_ac7f_4274_ca04589e186a -->|method| 914b6a96_704c_ef23_6236_b6d24e64245f

Relationship Graph

Source Code

libs/langchain_v1/langchain/agents/middleware/tool_selection.py lines 93–358

class LLMToolSelectorMiddleware(AgentMiddleware[AgentState[ResponseT], ContextT, ResponseT]):
    """Uses an LLM to select relevant tools before calling the main model.

    When an agent has many tools available, this middleware filters them down
    to only the most relevant ones for the user's query. This reduces token usage
    and helps the main model focus on the right tools.

    Examples:
        !!! example "Limit to 3 tools"

            ```python
            from langchain.agents.middleware import LLMToolSelectorMiddleware

            middleware = LLMToolSelectorMiddleware(max_tools=3)

            agent = create_agent(
                model="openai:gpt-4o",
                tools=[tool1, tool2, tool3, tool4, tool5],
                middleware=[middleware],
            )
            ```

        !!! example "Use a smaller model for selection"

            ```python
            middleware = LLMToolSelectorMiddleware(model="openai:gpt-4o-mini", max_tools=2)
            ```
    """

    def __init__(
        self,
        *,
        model: str | BaseChatModel | None = None,
        system_prompt: str = DEFAULT_SYSTEM_PROMPT,
        max_tools: int | None = None,
        always_include: list[str] | None = None,
    ) -> None:
        """Initialize the tool selector.

        Args:
            model: Model to use for selection.

                If not provided, uses the agent's main model.

                Can be a model identifier string or `BaseChatModel` instance.
            system_prompt: Instructions for the selection model.
            max_tools: Maximum number of tools to select.

                If the model selects more, only the first `max_tools` will be used.

                If not specified, there is no limit.
            always_include: Tool names to always include regardless of selection.

                These do not count against the `max_tools` limit.
        """
        super().__init__()
        self.system_prompt = system_prompt
        self.max_tools = max_tools
        self.always_include = always_include or []

        if isinstance(model, (BaseChatModel, type(None))):
            self.model: BaseChatModel | None = model
        else:
            self.model = init_chat_model(model)

    def _prepare_selection_request(
        self, request: ModelRequest[ContextT]
    ) -> _SelectionRequest | None:
        """Prepare inputs for tool selection.

        Args:
            request: the model request.

        Returns:
            `SelectionRequest` with prepared inputs, or `None` if no selection is
            needed.

        Raises:
            ValueError: If tools in `always_include` are not found in the request.
            AssertionError: If no user message is found in the request messages.
        """