in API, is there any way to list the available/applicable models specific to a capability (chat, embeddings etc)?

[joshuamosesb]

from openai import OpenAI
client = OpenAI()

client.models.list()

this snippet lists all available models for the client instance.
But, is there any way to list available/valid models at respective capability level like client.chat.models.list() or client.embeddings.models.list() ?
this will help in passing a suitable model name while calling the respective create method (e.g., client.chat.completion.create)

if we pass model="text-embedding-ada-002" not suitable for chat.completion.create

response = client.chat.completions.create(
  model="text-embedding-ada-002",
  messages=[
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Introduce yourself."}
  ]
)

we get an error:

NotFoundError: Error code: 404 - {'error': {'message': 'This is not a chat model and thus not supported in the v1/chat/completions endpoint. Did you mean to use v1/completions?', 'type': 'invalid_request_error', 'param': 'model', 'code': None}}

[judahpaul16]

Could maybe do something like this to fallback to another endpoint if the first try fails (ignore the display var):

async def query_openai(text, display, retries=3):
    # Load settings from settings.json
    settings = load_settings()
    max_tokens = settings.get("max_tokens")
    temperature = settings.get("temperature")

    for i in range(retries):
        try:
            response = openai.ChatCompletion.create(
                model=settings.get("model"),
                messages=[
                    {"role": "system", "content": f"You are a helpful assistant. {settings.get('custom_instructions')}"},
                    {"role": "user", "content": f"Human: {text}\nAI:"}
                ],
                max_tokens=max_tokens,
                temperature=temperature
            )
            response_content = response['choices'][0]['message']['content'].strip()
            if response_content:  # Check if the response is not empty
                message = response_content
                return message
            else:
                logger.warning(f"Retry {i+1}: Received empty response from OpenAI.")
        except Exception as e:
            if 'Did you mean to use v1/completions?' in str(e):
                # Re-query using v1/completions
                prompt = f"You are a helpful assistant. {settings.get('custom_instructions')}\nHuman: {text}"
                response = openai.Completion.create(
                    model=settings.get("model"),
                    prompt=prompt,
                    max_tokens=max_tokens,
                    temperature=temperature
                )
                response_content = response['choices'][0]['text'].strip()
                if response_content:
                    message = response_content
                    return message
            else:
                logger.error(f"Error on try {i+1}: {e}")
            if i == retries - 1:  # If this was the last retry
                error_message = f"Something went wrong after {retries} retries: {e}"
                await handle_error(error_message, None, display)
        await asyncio.sleep(0.5)  # Wait before retrying

This function could be modified to filter whatever models you want to use:

async def available_models():
    try:
        # Get available models from OpenAI
        openai.api_key = os.getenv("OPENAI_API_KEY")
        model_list = openai.Model.list()
        
        # Filter to only keep supported models.
        supported_models = [model['id'] for model in model_list['data'] if "gpt" in model['id'].lower()]

        return JSONResponse(content={"models": supported_models})
    except Exception as e:
        return HTTPException(status_code=500, detail=f"An error occurred: {str(e)}")

Check out my project GPT Home for more context: https://github.com/judahpaul16/gpt-home. First function is from src/actions.py second function is from src/backend.py.