LoginSign Up
← Back to all sparks
L

LiveKit Agents

AI-ASSISTANTS
Velocity5.0

Framework for building real-time voice and multimodal AI agents over WebRTC.

docs.livekit.io/agents ↗

LiveKit Agents ships weekly, and the work is nearly all provider breadth and failure paths.

voice-agentsprovider-pluginsturn-detectionweekly-releasescli-migration
◆Current state
The two newest entries are bare release tags — 1.6.10 and a 1.6.11 release candidate — with no notes. The releases with content follow one pattern: add speech and avatar vendors (Deepgram Flux, Bland TTS, Spatius avatars, Phonic and Smallest STT options), then fix the ways conversations go wrong. Endpointing and turn detection recur constantly — max_delay held constant under dynamic endpointing, primary speaker preserved through unattributed segments, an Ink-2 speech-onset delay that pushed back STT interruptions. 1.6.8 deprecates the console and dev modes in favour of the lk agent CLI.
◆Where it's heading
This is a maintenance train rather than a roadmap: breadth across providers plus relentless correctness work on turn-taking, which is where voice agents actually fail. The one demand on users in the window is the CLI migration. Cadence is roughly weekly and the contributor list is largely external — vendors landing their own plugins.
◆Prediction
More provider plugins and more endpointing fixes; the deprecated console and dev modes are the next thing likely to be removed outright.

◆Recent moves

  1. 5d ago

    livekit-agents@1.6.11.rc1 (#6847)

    A 1.6.11 release candidate published with no notes beyond the tag. Nothing here indicates what it contains.

    View source ↗
  2. 6d ago

    livekit-agents@1.6.10

    A 1.6.10 release whose body is just the version and PR number. Contentless in the feed, consistent with this train tagging faster than it documents.

    View source ↗
  3. 11d ago

    Agents 1.6.9 adds Deepgram Flux and Bland TTS plugins

    Two more speech vendors join the roster with Deepgram Flux and a Bland TTS plugin, continuing the breadth strategy. The supporting fixes are characteristically about failure paths, including giving each TTS fallback recovery its own task slot so a degraded provider does not stall the others.

    View source ↗
  4. 15d ago

    Agents 1.6.8 deprecates console/dev modes for the lk agent CLI

    This is the release that asks users to do work: console and dev modes are deprecated in favor of the lk agent CLI, so anything shelling out to python agent.py needs migrating. It also fixes an Ink-2 speech onset delay that pushed back STT-based interruption for setups running without a local VAD, which is the turn-detection theme again.

    View source ↗
  5. 25d ago

    Agents 1.6.7 adds Spatius avatar plugin, fixes endpointing delays

    A Spatius avatar plugin extends the roster beyond speech into rendered presence. The remaining changes cluster on conversation-state correctness, keeping max_delay constant under dynamic endpointing and holding the primary speaker through unattributed segments.

    View source ↗
  6. 1mo ago

    Agents 1.6.6 pools FishAudio sockets to fix cold-start crackle

    FishAudio gains pooled websocket reuse and prebuffering to remove cold-start crackle, a concrete audio-quality gain rather than a new capability. ElevenLabs begins deprecating its STT model_id in favor of model, one of several small provider-API alignments in this release.

    View source ↗