Repository navigation
3.0.0: Upgrade to React Native 0.88 and Yarn 4 with TurboModule migration - #24
Merged
Merged
Conversation
BREAKING CHANGE: ESM-only build output; requires React >= 19 and React Native >= 0.79.
BREAKING CHANGE: requires the New Architecture and React Native >= 0.80. Native device events are replaced by Deepgram.onAudioPCM, onAudioLevel, onAudioRouteChange and onAudioInterruption. The iOS MicPermission module is replaced by Deepgram.requestRecordPermission().
…mplate Use the monorepo Metro/Babel config, autolink the library through example/react-native.config.js and load the config plugin by relative path (drops the postinstall symlink). Migrate to the new expo-file-system API and react-native-safe-area-context, add Android/iOS example builds to CI, and make Deepgram.stopRecording() resolve null consistently.
Extract ManagedSocket (auth, reconnect backoff, stale-event guard), createKeepAlive, useSessionStats, PCM helpers, listen/speak URL builders and agent-settings helpers into src/internal, and rebuild the speech-to-text, text-to-speech and Voice Agent hooks on top of them. Adds characterization tests for all three hooks plus unit tests for the internals, and fixes the races and bugs listed in the changelog.
- configure() sends mid-stream Configure (keyterms/numerals on Nova-3, thresholds/keyterms/language_hints/numerals on Flux) - forceEndTurn() for Flux - onTurnInfo and onMessage callbacks; typed TurnInfo/Configure options - Flux URLs forward numerals and redact; language_hint for Nova-3 multi - nova-3-pharma model, aggressive_numbers redaction BREAKING CHANGE: the deprecated diarize option is removed, use diarizeModel.
- flux-* voices stream from /v2/speak and use the v2 REST endpoint - interrupt() and configureStream() for Flux; expressivity and stream tag - onSpeechStarted, onSpeechMetadata, onSpeechInterrupted, onStreamMessage - Flux sendText keeps whitespace; clearStream interrupts on Flux - Aura-2 de/fr/nl/it/ja voices and the Flux English catalog in model types BREAKING CHANGE: the deprecated top-level TTS options, http.format, the DeepgramTextToSpeechEncoding alias and the Text stream message are removed.
- forceEndTurn(), updateThinkLive(), dynamic and interrupt injections - onLatencyReport, onFunctionCallCancelled, onFrustrationDetected, onThinkLiveUpdated and onCustomFromThinkProvider callbacks - typed listen v1/v2 and speak (flux/aura) providers, think_live, defer_until_eot, reasoning_mode and agent configuration ids - createAgentSettings builds Flux-correct providers and takes speakModel BREAKING CHANGE: agent.language is removed from the settings types and createAgentSettings puts the language on the listen provider instead.
- iOS: track the mic capture rate separately from the playback rate so PCM
events and WAV headers stay at 16 kHz after startPlayer
- iOS: keep the session while a streaming player is prepared, restart a
stopped engine on feedAudio, resolve one-shots after playback
- iOS: run spec methods and session notifications on one serial queue
- iOS: fix simulator Voice Agent capture ("Invalid input format"), no
RedBox for rejected promises or session fallbacks
- synthesize() sets the player rate, respects autoPlayAudio, skips
non-PCM playback; Voice Agent caps the declared mic input rate at 16 kHz
- Android: non-sticky keep-alive service, startPlayer never throws
- keep test helpers and __tests__ declarations out of the published types - README: Migrating from 2.x section, telemetry (trackStats/getStats/stats) and onAudioConfig docs, React Native 0.80 iOS/Xcode minimums, Flux TTS - release-it tags match the existing bare-version tags and the podspec - release workflow writes the npm token to .npmrc
- speech-to-text: Nova-3 / Nova-3 multilingual / Flux picker, Flux turn indicator and End turn, live keyterm configure() with server status - text-to-speech: Flux interrupt, mid-stream speed, expressivity, turn events; 24/48 kHz rates and 24 kHz defaults - voice agent: Flux listen models, Flux/Aura voice picker with live switch, latency report, end-my-turn and agent message injection - missing API key banner, Android-safe monospace font
- Bottom tab bar (Agent/Listen/Speak/Intel/Manage) with safe-area insets and a compact header (eyebrow, screen title, one-line description). - New SegmentedControl for audio routes, STT engines and caption modes. - Decorative capability chips replaced by a mono "pipeline" pill showing the active models. - Compact heroes; Intel/Manage heroes dropped in favour of card headers with status badges, removing the duplicated screen titles. - Even two-column button grids (flexButton) instead of ragged rows.
Ported from a downstream app's patch-package fix. Other libraries commonly declare FOREGROUND_SERVICE, so with backgroundAudio: false on Android 10-13 the permission check passed and startForegroundService() ran. startForeground() then threw because the service declares no foregroundServiceType, and the unfulfilled startForegroundService() crashed the app. The keep-alive service now also requires the merged manifest to declare the mediaPlayback type, and only adds the microphone type when declared.
The iOS Simulator has no VPIO and the Android emulator exposes no AcousticEchoCanceler, so the Voice Agent hears itself through the host speakers and barges in on its own audio. Show a banner on the Agent screen when running on a simulator/emulator (expo-device) suggesting headphones or a physical device.
itsRares
force-pushed
the
feature/3.0.0
branch
from
November 5, 2026 16:25
8bcc286 to
451cb1b
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Upgrade the toolchain to Yarn 4, React Native 0.88, and TypeScript 6. Migrate the native module to a TurboModule architecture, introducing typed event emitters and new APIs. Update the example app to showcase the new features and ensure compatibility with the latest Expo SDK. Address various bugs and improve performance across platforms. Include breaking changes and a migration guide for users.