Description

In the agent, I set function_calling_llm intending to pass tool information through function calls rather than injecting it into prompts. However, through event listener monitoring, I found that this didn't take effect—it always used the prompt method.

After reviewing the source code, I discovered that the issue arises because litellm.utils.supports_function_calling(self.model, custom_llm_provider=provider) is used to determine whether the model supports function calls. Since this model is newly developed by our company and not included in the list, how should I address this problem?

Steps to Reproduce

null

Expected behavior

null

Screenshots/Code snippets

null

Operating System

Ubuntu 20.04

Python Version

3.10

crewAI Version

0.193.2

crewAI Tools Version

0.73.1

Virtual Environment

Venv

Evidence

null

Possible Solution

null

Additional context

null