support Azure OpenAI api

[wxxb789]

will this package support Azure OpenAI api?

[karthink]

Not explicitly, no. But I had a look at the API:

curl https://YOUR_RESOURCE_NAME.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT_NAME/chat/completions?api-version=2023-03-15-preview \
  -H "Content-Type: application/json" \
  -H "api-key: YOUR_API_KEY" \
  -d '{"messages":[{"role": "system", "content": "You are a helpful assistant."},{"role": "user", "content": "Does Azure OpenAI support customer managed keys?"},{"role": "assistant", "content": "Yes, customer managed keys are supported by Azure OpenAI."},{"role": "user", "content": "Do other Azure Cognitive Services support this too?"}]}'

It looks very similar to the OpenAI API, so you might be able to use gptel with minimal tweaking: Replace the request URL and one of the headers (Authorization: Bearer sk-... → api-key: sk-...). If that works I could add a PR.

[jcs090218]

Maybe worth mentioning, openai.el now supports this. 😄 See https://github.com/emacs-openai/openai/issues/14.

[karthink]

Maybe worth mentioning, openai.el now supports this. smile See emacs-openai/openai#14.

Thanks for the heads-up. If all that's needed is making the base url configurable, this should be simple to add.

[TeMPOraL]

It's not that simple - see https://github.com/emacs-openai/openai/issues/18. Basically, Azure OpenAI API is slightly different:

POST https://YOUR_RESOURCE_NAME.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT_NAME/completions?api-version=2022-12-01

In particular, you:

• Don't pass model parameter in the payload, but istead it's the YOUR_DEPLOYMENT_NAME;
• You don't have v1, but a api-version URL parameter;
• You don't pass the API key as Authorization: Bearer $key, but as api-key: $key header.

I would dearly love to see this implemented too.

[karthink]

Support for Azure has been implemented in v0.5.0, along with support for local models using GPT4All and Ollama. Please refer to the README for instructions.