protocol: show "Generating" status while it is really generating

[sykp241095]

If it is generating answer, use formal font-style, and prompting: "Generating answer..."

[634750802]

This depends on a reproducible stream api.

As discussed before, we could use redis's XADD and XREAD to dispatch generating stream to multiple clients.

• client 1 call chat api to generate chat
  • Server create a stream response to client 1
  • Server create a redis stream and duplicate each chunks from original stream
  • Optional Server store the stream to database or file storage for debug
• client N call the new read stream api to read chunks from redis stream