Support `ChatCompletion` Endpoint in OpenAI demo server

[zhuohan123]

@infwinston Feel free to use FastChat's completion template to implement a chat completion endpoint in our demo server. You can use the completion API as a reference:

vllm/vllm/entrypoints/openai/api_server.py

Lines 88 to 101 in 9d27b09

	@app.post("/v1/completions")
	async def create_completion(raw_request: Request):
	"""Completion API similar to OpenAI's API.
	See https://platform.openai.com/docs/api-reference/completions/create
	for the API specification. This API mimics the OpenAI Completion API.
	NOTE: Currently we do not support the following features:
	- echo (since the vLLM engine does not currently support
	getting the logprobs of prompt tokens)
	- suffix (the language models we currently support do not support
	suffix)
	- logit_bias (to be supported by vLLM engine)
	"""

[gesanqiu]

I noticed that FastChat already have a vllm_worker, so I think we could just use FastChat as frontend which provided OpenAI-Compatible APIs and vLLM as backend.

[zhuohan123]

I noticed that FastChat already have a vllm_worker, so I think we could just use FastChat as frontend which provided OpenAI-Compatible APIs and vLLM as backend.

Yes! FastChat is a great frontend to enjoy the benefit of vLLM. However, there are some more advanced options (e.g., beam search) that requires we implement a separate end point to support.

[gesanqiu]

I didn't run vllm_worker succeed yesterday and I found it use a older version of vllm, so I implement a ChatCompletion API at (#330 )

[zhuohan123]

Fixed by #330