@cloudflare/voice@0.2.0

cloudflare/agents@cloudflare/voice@0.2.0May 13, 2026by github-actions[bot]

AI Summary

This release adds an `enabled` option to `useVoiceAgent` for React apps to delay VoiceClient creation, along with bug fixes for Workers AI STT session edge cases in Flux and Nova 3 models, and a fix for `withVoice` text streaming with AI SDK's `textStream` responses.

Key Highlights

Added `enabled` option to `useVoiceAgent` for delaying VoiceClient creation until async prerequisites like capability tokens are ready
Fixed Flux STT to preserve latest non-empty transcript so empty `EndOfTurn` events still emit completed utterances
Fixed Flux `StartOfTurn` to drive server-side barge-in for prompt user speech abort during LLM/TTS playback
Fixed Nova 3 STT to defensively normalize finalized segment state to avoid stale teardown errors during abnormal close
Fixed `withVoice` text streaming for AI SDK `textStream` responses to properly produce TTS audio

New Features

New `enabled` option in `useVoiceAgent` hook to conditionally enable/disable voice agent initialization based on async prerequisites

Full Release Notes

### Minor Changes

-   [#1478](https://github.com/cloudflare/agents/pull/1478) [`2c7d91b`](https://github.com/cloudflare/agents/commit/2c7d91b7dd2aed73b1871f72b79e6e59b89e2ce8) Thanks [@whoiskatrin](https://github.com/whoiskatrin)! - Add an `enabled` option to `useVoiceAgent` so React apps can delay creating and connecting a `VoiceClient` until async prerequisites such as capability tokens are ready.

### Patch Changes

-   [#1458](https://github.com/cloudflare/agents/pull/1458) [`84cb429`](https://github.com/cloudflare/agents/commit/84cb429f7f41becc5e6ff0592f0308c52a5134f1) Thanks [@whoiskatrin](https://github.com/whoiskatrin)! - Fix Workers AI STT session edge cases for Flux and Nova 3.

    Flux now preserves the latest non-empty turn transcript from turn lifecycle events so an `EndOfTurn` event with an empty `transcript` can still emit the completed utterance. Flux `StartOfTurn` also drives server-side barge-in so model-detected user speech aborts active LLM/TTS playback promptly. Nova 3 now defensively normalizes finalized segment state before reading it to avoid stale teardown messages throwing during abnormal close paths.

-   [#1462](https://github.com/cloudflare/agents/pull/1462) [`5f6214d`](https://github.com/cloudflare/agents/commit/5f6214dccfe3ba9bf243ec15d291b37b9659a54c) Thanks [@whoiskatrin](https://github.com/whoiskatrin)! - Fix `withVoice` text streaming for AI SDK `textStream` responses so TTS audio is produced when `onTurn()` returns `streamText(...).textStream` directly.