[Bug]: Mistral function calling is not working with LM Studio

[rjambrecic]

Describe the bug

If we register a function to agent with Mistral llm config, the conversation will raise BadRequestError

Steps to reproduce

from autogen import ConversableAgent, register_function

llm_config = {
    "config_list": [
        {
            "model": "lmstudio-community/mistral-7b-instruct-v0.3",
            "base_url": "http://localhost:1234/v1",
            "api_key": "lm-studio",
            "price":[0.0,0.0],
        },
    ],
    "cache_seed": None,  # Disable caching.
}

agent = ConversableAgent(
    name="Agent",
    llm_config=llm_config,
)

executor = ConversableAgent(
    name="user_proxy",
    human_input_mode="NEVER",
)

def get_airplane_ticket_price():
    return "Ticket is 100€"

register_function(
    get_airplane_ticket_price,
    caller=agent,
    executor=executor,
    description="Get the price of an airplane ticket",
)


executor.initiate_chat(
    recipient=agent,
    message="I need to buy a plane ticket for Paris",
)

user_proxy (to Agent):

I need to buy a plane ticket for Paris

--------------------------------------------------------------------------------

>>>>>>>> USING AUTO REPLY...
---------------------------------------------------------------------------
BadRequestError                           Traceback (most recent call last)
Cell In[1], [line 36](vscode-notebook-cell:?execution_count=1&line=36)
     [26](vscode-notebook-cell:?execution_count=1&line=26)     return "Ticket is 100€"
     [28](vscode-notebook-cell:?execution_count=1&line=28) register_function(
     [29](vscode-notebook-cell:?execution_count=1&line=29)     get_airplane_ticket_price,
     [30](vscode-notebook-cell:?execution_count=1&line=30)     caller=agent,
     [31](vscode-notebook-cell:?execution_count=1&line=31)     executor=executor,
     [32](vscode-notebook-cell:?execution_count=1&line=32)     description="Get the price of an airplane ticket",
     [33](vscode-notebook-cell:?execution_count=1&line=33) )
---> [36](vscode-notebook-cell:?execution_count=1&line=36) executor.initiate_chat(
     [37](vscode-notebook-cell:?execution_count=1&line=37)     recipient=agent,
     [38](vscode-notebook-cell:?execution_count=1&line=38)     message="I need to buy a plane ticket for Paris",
     [39](vscode-notebook-cell:?execution_count=1&line=39) )

File ~/projects/ag2/autogen/agentchat/conversable_agent.py:1499, in ConversableAgent.initiate_chat(self, recipient, clear_history, silent, cache, max_turns, summary_method, summary_args, message, **kwargs)
   [1497](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1497)     else:
   [1498](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1498)         msg2send = self.generate_init_message(message, **kwargs)
-> [1499](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1499)     self.send(msg2send, recipient, silent=silent)
   [1500](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1500) summary = self._summarize_chat(
   [1501](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1501)     summary_method,
   [1502](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1502)     summary_args,
   [1503](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1503)     recipient,
   [1504](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1504)     cache=cache,
   [1505](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/autogen/agentchat/conversable_agent.py:1505) )
...
   (...)
   [1079](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/.venv/lib/python3.11/site-packages/openai/_base_client.py:1079)     retries_taken=retries_taken,
   [1080](https://file+.vscode-resource.vscode-cdn.net/Users/robert/projects/ag2/~/projects/ag2/.venv/lib/python3.11/site-packages/openai/_base_client.py:1080) )

BadRequestError: Error code: 400 - {'error': 'Error rendering prompt with jinja template: Error: Only user and assistant roles are supported!'}

Model Used

No response

Expected Behavior

No response

Screenshots and logs

No response

Additional Information

No response