pipecat-ai
llm.handleFunctionCall only receives functionName, no arguments
pipecat version
latest main branch
Python version
3.12
Operating System
macos 15
Issue description
On the latest main branch, the llm.handleFunctionCall(async (props) => { ... }) function is invoked only with the functionName, but none of the expected arguments (arguments, args, etc.) are included in props. However, the onLLMFunctionCall handler correctly receives the arguments object (typed incorrectl though )with the expected data.
Reproduction steps
Clone the repo and check out the latest main branch.
Set up an LLM instance with both llm.handleFunctionCall and onLLMFunctionCall defined.
Trigger a tool/function call via the LLM.
Observe the props received by llm.handleFunctionCall.
Expected behavior
llm.handleFunctionCall should receive a props object that includes the arguments or args, similar to what is passed to onLLMFunctionCall.
Actual behavior
llm.handleFunctionCall only receives the functionName, and the arguments are missing entirely from the props object.
Logs
Just to clarify, when you are referring to llm.handleFunctionCall, are you referring to the client-js and react API?
Just to clarify, when you are referring to
llm.handleFunctionCall, are you referring to the client-js and react API?
Yes