livekit-agents@1.8.8

livekit/agentslivekit-agents@1.8.8Oct 15, 2026by github-actions[bot]

Upgrading?

AI Summary

This release enables running GPT-Live and Google Gemini Live directly via LiveKit Inference, eliminating the need for separate API keys. It introduces new decision-model plugins like TypeSafe Jev and updates model lists for xAI and Google.

Key Highlights

  • GPT-Live integration via `inference.DuplexModel`
  • Support for Google Gemini Live (3.x models) and TypeSafe Jev decision model
  • xAI Grok Voice 3.0 replaces Think Fast 2.0 in Inference lists

Breaking Changes

  • `inference.RealtimeModel` no longer subclasses the plugin's `openai.realtime.RealtimeModel`; update `isinstance` checks
  • `cedar` voice replaces the unsupported `aster` voice in GPT-Live

New Features

  • TypeSafe Jev decision model plugin for background probability evaluations
  • Inference model-string support for Google Gemini Live models

Full Release Notes

## GPT-Live through LiveKit Inference

You can now run GPT-Live with your LiveKit credentials instead of a separate OpenAI key. `inference.DuplexModel` handles its protocol, and you can adopt it directly with a model string:

```python
session = AgentSession(llm="openai/gpt-live-1")
```

## New providers and models

- **OpenAI GPT-Live 1**: now available through LiveKit Inference.
- **Google Gemini Live**: adds Inference model-string support for `google/gemini-3.8-live`, `google/gemini-3.8-live-extended-thinking`, and `google/gemini-live`.
- **xAI Grok Voice 3.0**: replaces Think Fast 2.0 in the Inference and xAI plugin model lists.
- **TypeSafe Jev**: a new decision-model plugin supports `jev-1.13.0` and OpenRouter’s Decisions API for background probability, choice, and score evaluations without delaying replies.

## Breaking changes

- **OpenAI realtime**: `inference.RealtimeModel` no longer subclasses the plugin’s `openai.realtime.RealtimeModel`; update `isinstance` checks. Incidental wildcard exports such as `realtime_model.logger` are no longer available.
- **GPT-Live voices**: `cedar` replaces the unsupported `aster` voice; update voice selections.

## Behavior changes

- **LiveKit Inference**: retired GPT Realtime 1.5/2 and Grok Voice Think Fast 2.0 IDs remain accepted with deprecation warnings. Realtime `inference_class` is deprecated and no longer sent.
- **Google Gemini**: 3.x reply instructions now use wrapped user turns.
- **xAI**: hosted sessions now emit one final user transcript per turn.

XML-aware TTS chunking, realtime startup retries, turn-audio flushing, and interruption handling are more reliable.
## What's Changed
* fix(tokenize): ignore stray closing tags in the XML depth check by @ANDROIDHASSAN in https://github.com/livekit/agents/pull/7742
* feat(inference, google): run GPT-Live as inference.DuplexModel, wrap Gemini 3.x reply instructions, update the realtime model list by @russellmartin-livekit in https://github.com/livekit/agents/pull/7673
* feat(agents): add DecisionModel for background decisions by @tinalenguyen in https://github.com/livekit/agents/pull/7714
* feat(typesafe): add TypeSafe Jev decision model plugin by @tinalenguyen in https://github.com/livekit/agents/pull/7756
* livekit-agents@1.8.8 by @github-actions[bot] in https://github.com/livekit/agents/pull/7844


**Full Changelog**: https://github.com/livekit/agents/compare/livekit-agents@1.8.7...livekit-agents@1.8.8

Related Repositories