Is there any way to configure extra headers to be passed to any model?

[stronk7]

I've seen @ #2110 (and some issues/PRs), that some of the providers (a few) support passing custom extra_headers. But many others don't.

It may be super-useful to be able to extend that feature to other providers. Here there is my current exact use case (just in case that there is some way to achive it that I've not been able to find):

• Ollama installed in remote server, running behind reverse proxy (right now restricting by ip/domain).
• We want to change that restriction to become auth based (say Basic Auth, or Bearer, or xxxx).
• For any of those cases, we need to be able to add some new headers (or, at very least, pass the LiteLLM Auth Bearer token). That way the reverse proxy may decide, log, ...
• I've been looking to the ollama/ollama_chat providers and haven't found anything obvious, neither to create new headers neither to passthrough the LiteLLM bearer token/key.

And that's my current use case and its enormous utility (IMO) for a number of other use cases.

Ciao :-)

[stronk7]

Related: I saw that ollama chat models were accepting things like headers: {"hey": "cross fingers"} within litellm_params and I tried... but those headers don't become part of the request to the provider, so far.