Supporting Azure OpenAI

[monuminu]

Supporting Azure OpenAI

[ralzinov]

LiteLLM potentially might help https://docs.litellm.ai/docs/

[nikhil-pandey]

We should be able to bake in Azure OpenAI support directly from the openai package — likely just a few lines of changes.

import { AzureOpenAI, OpenAI } from "openai";

I might take a stab at it later today, or wire it up through GitHub agent mode if I get the chance.

[rgregg]

I asked codex this question on it's own repo - and it says you should be able to set two ENV vars to redirect it:

 • Set OPENAI_BASE_URL to your Azure OpenAI endpoint (e.g. https://<your‑resource>.openai.azure.com)
 • Set OPENAI_API_KEY (or codex config set apiKey …) to your Azure API key

I haven't given it a try yet though.

[guygregory]

I asked codex this question on it's own repo - and it says you should be able to set two ENV vars to redirect it:

 • Set OPENAI_BASE_URL to your Azure OpenAI endpoint (e.g. https://<your‑resource>.openai.azure.com)
 • Set OPENAI_API_KEY (or codex config set apiKey …) to your Azure API key

I haven't given it a try yet though.

Deployment name for the model will also be required for AOAI

[SeaDude]

Please also consider implementing azure_ad_token (or the like) so that enterprises don’t have to communicate and manage AOAI API keys.

Reference

[thegovind]

Working on it - #92

[surim0n]

Following, if someone successfully gets it running please update this thread. thank you x 1000000.

[cavanaug]

It would be a godsend if you supported the GH copilot api similar to how vscode utilizes models. GH copilot also has a generous free tier and is well utilized in Enterprises.

The GH OpenAi api is a slight variant on the generic OpenAI api. More information and how litellm supported it available at BerriAI/litellm#6564

[jslitzkerttcu]

PR is being worked on in #769