multi stream websocket connection for ElevenLabs TTS

[simllll]

Description

Implements multi stream websocket connection for ElevenLabs TTS

Based on the work of #828
fix #824

Changes Made

• implement connection handler
• implement alignment
• implement cpu friendly audio and transcript processing

Pre-Review Checklist

• Build passes: All builds (lint, typecheck, tests) pass locally
• AI-generated code reviewed: Removed unnecessary comments and ensured code quality
• Changes explained: All changes are properly documented and justified above
• Scope appropriate: All changes relate to the PR title, or explanations provided for why they're included
• Video demo: A small video demo showing changes works as expected and did not break any existing functionality using Agent Playground (if applicable)

Testing

• Automated tests added/updated (if applicable)
• All tests pass
• Make sure both restaurant_agent.ts and realtime_agent.ts work properly (for major changes)

Additional Notes

not sure how to test the alignment part, but I followed the python implementation. We use the audio part of this implementation already in production.

---

Note to reviewers: Please ensure the pre-review checklist is completed before starting your review.

[changeset-bot]

🦋 Changeset detected

Latest commit: a4c1eff

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@livekit/agents-plugin-elevenlabs	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[Devesh36]

LG!! please review it sir @toubatbrian :)

[Devesh36]

LG !!

[simllll]

There are still some issues with ending the websocket manager. Right now I'm not even sure if we can actually solve this with the current architecture, but let's see what else I can find out

I'm experiencing similar issues with the current deepgram implementation (on stt side).

[simllll]

Okay, I've currently not so much time to dig into it or fix it, but here are my current findings:

• in general, there is no real cleanup happening in typescript (Maybe this is by desing, due to the task fork approach?)
• tts is by design by "request" (one text per request - short living), stt is by design by session (one audio stream - long running).
• cleanup in deepgram's stt implementaiton is also broken, the wsMonitor promise is never resolved. (quite easy fixable). which means even though the process only needs to get cleaned up at the end, it's not really happening and the sub process gets either killed or is running endless (or at least way longer than needed)
• main issue for the TTS multi stream websocket case: As we have a "per request" design, I can actually establish the websocket at the beginning, but I have no idea when I should close it again.. there is no "close()" hook like in python the "aclose()" https://github.com/livekit/agents/blob/a9e43fec7fc2a752658cf80506d901d0af622e38/livekit-agents/livekit/agents/tts/tts.py#L140
• in python there is a close hook also for stt, just saying..not so much of relevance, because we have the AbortHandler at least in the STT implementation that fullfills a simliar goal.

so what we need is either:

1. AbortController in the Base TTS that "gets cancelled" at the end of the session.
2. close() hook in the Base TTS that gets called when the session ends.
   to get this thing "clean".

[simllll]

this one would be ready @toubatbrian
Also added #861 changes

[toubatbrian]

Thanks @simllll, will take a look!