Hi, is it currently possible for agent to join the room with multiple participants and use stt to get transcription of the conversation? Or do i need an agent for every participant? Or should such case be handled in a different way?

We do need more agents-js examples. Which Python example(s) were you looking to replicate in agents-js? We have the https://github.com/livekit-examples/python-agents-examples/blob/main/complex-agents/note-taking-assistant/agent.py but that doesn't do anything special in the agent since by default it will listen to all room participants, the logic in that demo is more to send the transcript over RPC to multiple participants. We also have https://github.com/livekit-examples/python-agents-examples/blob/main/pipeline-stt/diarization.py to show diarization, but that doesn't have any logic to send the transcription to the front end.

I needed the functionality of https://github.com/livekit/agents/blob/main/examples/other/transcription/multi-user-transcriber.py. In the end, I rewrote our agent in python based on this the only thing we are now missing is access to previous transcripts (ie. if someone joins later). We don't want to store the transcript somewhere so will probably try to just save it to some variable in agent and then send the full history to new participants in the next.js app (who would have to dedupe it)

That sounds like a good approach