Responses max_tool_calls

[iamemilio]

🚀 Describe the new functionality needed

Responses API can accept a max_tool_calls parameter that limits the number of tool calls allowed to be executed for a given response. There are a few things to consider when implementing this:

Requirements

• When handling inference calls that get converted into chat completions:
  • If a list of tool call requests is returned from the model, truncate that list down to max_tool_calls, then execute the calls for each tool.
  • If max_tool_calls < 0, return Bad Request

💡 Why is this needed? What if we don't build it?

This is a key functionality of responses and it allows the tool calls made to be filtered down so the model doesn't get overwhelmed by context. Also this is a feature gap.

Other thoughts

No response