How to Use a Self-Hosted Model with Custom Headers

[mio4kon]

Feature request

Our company provides an API similar to OpenAI's ({host}/v1/chat/completions)。
but requires custom headers in the API request.

Does pr-agent support this?

Motivation

Use a Self-Hosted Model

[mrT23]

We support all the different models and APIs via litellm.

Are you able to utilize these custom headers via litellm ? if so, share a link, and we can check if it can be supported

[mio4kon]

@mrT23
Perhaps my explanation wasn't very clear.

Our AI provider is internal within the company, and they provide us with an API. This API is quite different from the standard OpenAI API—it requires extra headers and the return format is also different.

I noticed that the project uses litellm at its core to call large models.
I would like to know if litellm supports this custom API calling method, or do we need to deploy our own proxy service to handle protocol translation?

[mrT23]

review this
#1564