langgraph_agent_toolkit.core.tool_selection module

Configure optional LangChain tool selection without shared request state.

class langgraph_agent_toolkit.core.tool_selection.ToolSelectionMiddleware(config, *, selector_model=None, protected_tools=())

Bases: AgentMiddleware

Adapt native selectors to the tools in each model request.

Each call creates a native middleware instance. Required Deep Agents tools stay visible when they exist in that request. Concurrent calls do not change one another’s tool lists or selection configuration.

Parameters:

• config (ToolSelectionConfig)

• selector_model (BaseChatModel | None)

• protected_tools (Sequence[str])

transformers = (<class 'langchain.agents.middleware.internal_call_transformer.InternalCallTransformer'>,)

Stream transformer factories registered by the middleware.

Each entry is a scope-aware factory invoked as factory(scope) so every invocation receives a fresh instance. Factories are merged with the transformers argument of [create_agent][langchain.agents.create_agent] at graph compile time, after the ToolCallTransformer and before any user-supplied entries.

__init__(config, *, selector_model=None, protected_tools=())

Parameters:

• config (ToolSelectionConfig)

• selector_model (BaseChatModel | None)

• protected_tools (Sequence[str])

Return type:

None

wrap_model_call(request, handler)

Select tools for one synchronous model request.

Parameters:

• request (ModelRequest)

• handler (Callable[[ModelRequest], ModelResponse])

Return type:

ModelResponse

async awrap_model_call(request, handler)

Select tools without blocking the event loop.

Parameters:

• request (ModelRequest)

• handler (Callable[[ModelRequest], Awaitable[ModelResponse]])

Return type:

ModelResponse

langgraph_agent_toolkit.core.tool_selection.create_tool_selection_middleware(config=None, *, selector_model=None, protected_tools=())

Resolve settings and optional managed model configuration at graph setup.

Parameters:

• config (ToolSelectionConfig | None)

• selector_model (BaseChatModel | None)

• protected_tools (Sequence[str])

Return type:

list[AgentMiddleware]