Different AI models seem to have different response types!

[jelveh]

The AI driver seems to return different responses for different models such claude's, gpt, ... This breaks the consistency expected from the driver.

Claude response:

{
    "success": true,
    "service": {
        "name": "claude"
    },
    "result": {
        "message": {
            "id": "msg_016XXX1pauQjLiu",
            "type": "message",
            "role": "assistant",
            "model": "claude-3-5-sonnet-20240620",
            "content": [
                {
                    "type": "text",
                    "text": "Rick and Morty is an adult animated science fiction ..."
                }
            ],
            "stop_reason": "end_turn",
            "stop_sequence": null,
            "usage": {
                "input_tokens": 84,
                "output_tokens": 767
            }
        },
        "finish_reason": "stop"
    }
}

Same function call but with gpt4o mini:

{
    "success": true,
    "service": {
        "name": "openai-completion"
    },
    "result": {
        "index": 0,
        "message": {
            "role": "assistant",
            "content": "An animated television series created by Justin Roiland and Dan Harmon ...",
            "refusal": null
        },
        "logprobs": null,
        "finish_reason": "stop"
    }
}

Either AI driver or Puter.js should normalize the response. Prefereably the AI driver.