What is the recommended way of dealing with multiple functionCall requests?

Gemini API
,
1

When working against either the Flash or Pro endpoints, both of them will invoke tools we have configured correctly.

My question is: What is the recommended approach to take if a model sends a response that consists of, say, 1 text part and 3 functionCall parts? As I see it, we’ve two choices:

  1. We run all 3 function calls and then send back a functionResponse with 3 parts in the array.
  2. We send a single functionResponse back 3 times, one for each request.

Note that we’re using model.generateContent in the TypeScript API, not the ‘chat’ API as we need to manage the context history ourselves.

Cheers,

Rich

3

@Richard_Davey,

depending on your application, if you want to call all the 3 function calls simultaniously, you can do so in a loop and add the responses to a list/array and send back to the model

following is an pseudo example for this:

for fc in tool_call.function_calls:
                    function_response = types.FunctionResponse(
                        id=fc.id,
                        name=fc.name,
                        response={ "result": "ok" } # simple, hard-coded function response
                    )
                    function_responses.append(function_response)

but if you want the output of 1st function call to use as input for the next you can use compositional function calling(Function Calling with the Gemini API  |  Google AI for Developers) .