feat: add support for OpenAI completion API

[Syst3m1cAn0maly]

Add support for OpenAI completion API in order for tabby to be plugged to other OpenAI compatible backends (for example vLLM).

Tested with the vLLM backend, autocompletion is a lot faster and can support more concurrent requests than the llama.cpp backend.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 60 lines in your changes missing coverage. Please review.

Project coverage is 58.24%. Comparing base (ef8542c) to head (17d7a04).
Report is 3 commits behind head on main.

Files	Patch %	Lines
crates/http-api-bindings/src/completion/openai.rs	0.00%	53 Missing ⚠️
crates/http-api-bindings/src/completion/mod.rs	0.00%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2604      +/-   ##
==========================================
- Coverage   58.45%   58.24%   -0.22%     
==========================================
  Files         167      168       +1     
  Lines       15154    15214      +60     
==========================================
+ Hits         8858     8861       +3     
- Misses       6296     6353      +57     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[wsxiaoys]

Thank you for the pull request—I'm happy to merge it. The only concern is that since OpenAI has deprecated its completion API, referring to it as OpenAICompletion is a bit tricky because its definition is no longer present in any standard OpenAI client library. If there's no better alternative, perhaps consider renaming it to something like vllm/completion or VllmCompletion?

[wsxiaoys]

Besides - could you also add an example for newly added completion config in https://github.com/TabbyML/tabby/edit/main/website/docs/administration/model.md? thank you!

[Syst3m1cAn0maly]

Thank you for the pull request—I'm happy to merge it. The only concern is that since OpenAI has deprecated its completion API, referring to it as OpenAICompletion is a bit tricky because its definition is no longer present in any standard OpenAI client library. If there's no better alternative, perhaps consider renaming it to something like vllm/completion or VllmCompletion?

I have mixed feelings about the OpenAI completion API as well as it is deprecated in their documentation...
However, it has become and still is a standard for many backends and is specifically mentioned as OpenAI completion API or OpenAI compatible API in their documentation and it is not limited to vLLM.
Some examples :

• in vLLM docs : https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html#api-reference
• in Nvidia NIM docs : https://docs.nvidia.com/nim/large-language-models/latest/getting-started.html#openai-completion-request
• in LocalAI docs : https://localai.io/features/text-generation/#completions