mods icon indicating copy to clipboard operation
mods copied to clipboard

Document how to use with Azure OpenAI

Open st0le opened this issue 1 year ago • 3 comments

I'm not able to get it to work. Keep hitting 404.

  azure:
    # Set to 'azure-ad' to use Active Directory
    # Azure OpenAI setup: https://learn.microsoft.com/en-us/azure/cognitive-services/openai/how-to/create-resource
    base-url: https://<resourcename>.openai.azure.com
    api-key-env: AZURE_OPENAI_KEY
    models:
      gpt-4:
        aliases: ["az4"]
        max-input-chars: 24500
        fallback:
      gpt-35-turbo-16k:
        aliases: ["az35t"]
        max-input-chars: 12250
        fallback:
      gpt-35:
        aliases: ["az35"]
        max-input-chars: 12250
        fallback:

image

Despite this I see a 404

echo say hi | mods -a azure -m gpt-35-turbo-16k

   ERROR  Missing model 'gpt-35-turbo-16k' for API 'azure'.

  error, status code: 404, message: The API deployment for this resource does not exist. If you created the deployment within the last 5 minutes, please wait a moment and try again.

Am I missing something?

st0le avatar Sep 26 '23 16:09 st0le

Hello, It seems you've encountered this issue because your deployment named "gpt-35-turbo-16k" hasn't been deployed with this name on Azure OpenAI. After providing the base URL, you will need to also specify your particular deployment name within the "models" section of your configuration.

For instance, if you've deployed a GPT-4 model named "MyGPT4" and a GPT-3.5 model named "MyGPT35" on Azure OpenAI, your configuration should resemble the following:

models:
  MyGPT4:
    aliases: ["az4"]
    max-input-chars: 24500
    fallback: MyGPT35
  MyGPT35:
    aliases: ["az35"]
    max-input-chars: 12250
    fallback:

ctrl-vfr avatar Oct 01 '23 16:10 ctrl-vfr

sorry to hijack this issue ? Is there a way of using a copilot api key with mod could not find anything ont he docs

lrfurtado avatar Oct 10 '23 16:10 lrfurtado

I think this issue also raises a good point on improving UX here. We should see about listing off what models are available for that API, if the one specified doesn't exist. Although, it is usually a good idea to specify fallback models in case one fails, again something we could highlight better in the docs. Thanks for letting us know :)

bashbunni avatar Mar 07 '24 00:03 bashbunni