[feature request] add prompt styles support

[jiacheo]

There are many different styles of prompts for different LLMs, such like openai/llama2 style (especially support SYSTEM role prompt), pure text style, ziya, etc.
From the api_server.py 's parameter, we can see it just support the pure text style.
Maybe it will be a good feature to support these styles.

[shihaobai]

Thank you very much for your suggestion. We will support this feature very soon.

[XHPlus]

Also welcome a contribution from the community, thanks!

[Valdanitooooo]

Is it possible to integrate in fastchat like vllm?
https://github.com/lm-sys/FastChat/blob/main/docs/vllm_integration.md

[WANDY666]

Hi, we have now added the /v1/chat/completions api, the usage is the same as that of openai.
We use fastchat to target different model types for different styles of prompt stitching.
Maybe you can give it a try.