docs/how_to/tool_runtime/

[giscus[bot]]

docs/how_to/tool_runtime/

<Prerequisites titlesAndLinks={[

https://python.langchain.com/docs/how_to/tool_runtime/

[iriacardiel]

Hi,

I am having trouble including arguments in tool functions that I want the LLM to ignore. According to this post, the LLM only receives information about the tool_call_schema and not the get_input_schema.
So any variable annotated with InjectedToolArg should not be in the LLM's context, right? That is what I understand from the documentation, but it's not what occurs in my tests. My model hallucinates with the Injected Tool Argument, and it's not supposed to know about its existence.

Thank you!

[balaji830]

I was able to successfully inject a runtime parameter. I am also trying to get structured output from LLM. I am using the option 2 approach from this documentation https://langchain-ai.github.io/langgraph/how-tos/react-agent-structured-output/#define-graph_1.
Any help on how to achieve this?