Providing a ToolNode tool access to state

[austinmw]

Hi, I am currently using ToolNode like this:

tools = [generate_workout]
tool_node = ToolNode(tools, handle_tool_errors=True)

How can I make it so that generate_workout also has access to state?

---

For more context, currently my generate_workout function only takes as input the following:

class WorkoutInputs(BaseModel):
    """Any messages to send the workout planner"""
    messages: list[str] = Field(
        description=("List of goals and preferences specified from the user."),
    )

def generate_workout(request: WorkoutInputs):
   ...

However I'd also like to provide it access with state['current_workout_program'] (a previous tool output).

[isahers1]

You can check out this how to for info on how to pass graph state to a tool, lmk if you have any further questions.

[austinmw]

Thanks!

[austinmw]

Hey @isahers1 , actually I have one more question. Is it possible for a tool to save something to the state as well?

For example, if the tool returns a dict, but I'd only like to pass a subset of the dict back to the chat LLM, and save the rest in state.