Azure OpenAI support

[hooman-bayer]

hi again and thanks again for chainlit! 🥳

Beside OpenAI, a lot of people like me, use Azure OpenAI API (Please see langchain docs). But currently it cant be easily used in chainlit.

[hooman-bayer]

Nevermind

According to Langchain simply set the following env variables for embedding models:

os.environ["OPENAI_API_TYPE"] = "azure"
os.environ["OPENAI_API_BASE"] = "https://<your-endpoint.openai.azure.com/"
os.environ["OPENAI_API_KEY"] = "your AzureOpenAI key"
os.environ["OPENAI_API_VERSION"] = "2023-03-15-preview"

and similarly for the chat_model.

[hooman-bayer]

Reopening this because prompt playground is a very useful feature which does not work with Azure OpenAI yet. Any hints how to solve it?

Please see below:

openai.error.InvalidRequestError: Must provide an 'engine' or 'deployment_id' parameter to create a <class 'openai.api_resources.chat_completion.ChatCompletion'>

Simply you need to pass that environmental variables to openai.ChatCompletion.create in case of Azure OpenAI API:

# Other env variables
openai.api_type = "azure"
openai.api_key ="<KEY>" 
openai.api_base = "<URL of Endpoint>"
openai.api_version = "2023-03-15-preview"

openai.ChatCompletion.create(
        deployment_id=settings.azure_deployment_id,
        model=model,
    )

[willydouhard]

For a quick local hack you can hardcode your deployment_id here https://github.com/Chainlit/chainlit/blob/main/src/chainlit/server.py#L91 in your local installation.

For the real fix, we need to enhance the prompt playground to make it more flexible and compatible with more LLMs.