feat: Support for v1/completions endpoint in OpenAI-compatible server

[sestinj]

Problem Continue.dev users are increasingly turning to Jan as their goto method of running a language model. Unfortunately, they can't use it for tab-autocomplete because this requires raw, non-chat-formatted requests. This is only possible through the v1/completions endpoint, which Jan doesn't seem to support.

Success Criteria Jan's OpenAI server adheres to the OpenAI spec for v1/completions. Then users can rely on Jan for local tab-autocomplete : )

Additional context

[clement-igonet]

I fully agree.

Also, /v1/completions is officialy declared as deprecated by the openai API doc, but still supported, and used by many clients of the openai API.

It would be nice to support it, for the adoption of jan (I'm not able to plug to continue dev autocomplete feature, but hope to succeed one day, and I'm fond of jan open/free license adoption).

[clement-igonet]

Any news to support /v1/completions ?

[controldev]

Agreed, many third party applications rely on the completions endpoint. This is a huge blocker.

[freelerobot]

Deepest apologies, we will be forward compatible only.