# Changelog

All notable changes to `vue-text-to-speech` are documented here.

The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

---

## [2.0.6] — 2026-08-13

### Documentation
- **README — repositioned** — tagline updated to *"The unified speech layer for Vue 3 — TTS, STT and AI streaming voice through one composable API"* across both root and package READMEs
- **README — provider architecture diagram** — new `## Provider Architecture` section with ASCII diagram showing the abstraction layer over all 4 providers
- **README — capability matrix** — new `## Capabilities` table listing 12 verified capabilities (all ✅ including SSR safe)
- **README — LLM streaming** — enhanced `## LLM Streaming` section with token → sentence detection → queue → audio flow diagram; noted compatibility with any `AsyncIterable<string>` source
- **README — "Why vue-text-to-speech?"** — new section with Without vs With code comparison illustrating the abstraction value
- **README — "Built for"** — new use-cases table covering AI assistants, accessibility, education, enterprise, and rapid prototyping
- **README — security section** — expanded from an inline warning into a full `## Security` section with the proxy pattern diagram and `baseURL` example
- **README — star CTA** — added subtle contributor star request before the License section
- **LICENSE** — added root-level `LICENSE` file (MIT, Copyright © 2024–2026 kunchamvasu); fixes the license badge link that previously 404'd

### Playground
- **Dynamic version chip** — `AppHeader.vue` now reads the version from `packages/vue-text-to-speech/package.json` at build time via Vite `define` (`__VTS_VERSION__`); the header chip always reflects the current release without manual edits
- **Dynamic test count** — `OverviewTab.vue` test count chip now reads the live test count from the test suite at build time via Vite `define` (`__VTS_TEST_COUNT__`); corrected stale count (was `192`, now accurate `207`)

### Docs
- **Dynamic test count** — VitePress `config.ts` now counts tests at build time and patches the `index.md` hero feature card via `transformPageData`, keeping the displayed count accurate without manual edits (was `189`)

## [2.0.5] — 2026-08-07

### Documentation
- **README — playground video** — added a clickable YouTube thumbnail to both the root and package READMEs linking to the full playground walkthrough ([▶ youtu.be/lcBE4AGz6EI](https://youtu.be/lcBE4AGz6EI))
- **Docs site aligned with codebase** — all API pages updated to match the current implementation: corrected `VoiceInfo` type (`id`, `label` replacing `voiceURI`/`localService`), `SpeechError.code` enum, `useSpeechSynthesis` options parameter, `speak()` overrides signature, `loadVoices()` async return, `useVoiceQueue` `skip()`/`dequeue()` methods, `VueSpeechVoiceSelect` required `voices` prop, `VueSpeechPlayer` visibility props, ElevenLabs provider defaults
- **New docs page** — `api/utilities` documents the public `extractCompleteSentences()` and `splitSentences()` exports
- **README updates** — corrected test count to 207, fixed ElevenLabs default `voiceId`, added `VueSpeechVoiceSelect` and `useSpeechRecognition` usage examples, documented `components: true` global registration option

## [2.0.4] — 2026-08-04

### Added
- **`VueSpeechPlayer` visibility props** — `showVoiceSelect`, `showRate`, `showPitch`, `showVolume` (all `boolean`, default `true`) let callers hide individual UI sections without replacing slots
- **`VueSpeechVoiceSelect` new props** — `voices: readonly VoiceInfo[]` (now required, passed from `useSpeechSynthesis`), `disabled?: boolean`, `loading?: boolean` (shows "Loading voices…" placeholder while voices are being fetched)
- **`useVoiceQueue` — `skip()` and `dequeue()`** — `skip()` stops the current utterance and triggers auto-advance to the next item; `dequeue()` manually pops the next item without speaking it
- **`useSpeechSynthesis` options parameter** — accepts `UseSpeechSynthesisOptions { provider?, rate?, pitch?, volume? }` so a component can override the injected provider or set initial playback values
- **`useSpeechSynthesis` — `speak()` overrides** — second argument `overrides?: Partial<SpeakOptions>` allows per-call voice/rate/pitch/volume without mutating the reactive refs
- **`useStreamingTTS` options parameter** — accepts `UseStreamingTTSOptions { provider? }` to override the injected provider per composable instance
- **Plugin `components` option** — `VueSpeechOptions.components?: boolean`; when `true`, registers all three UI components globally (opt-in, default `false`)
- **`createWebSpeechProvider()`** — synchronous factory exported from the public API; returns a `WebSpeechTTSProvider` without the async `createVueSpeech()` path
- **Utility exports** — `extractCompleteSentences(buffer)` and `splitSentences(text)` are now public API for consumers building custom streaming pipelines
- **`SPEECH_PROVIDER_KEY`** — injection key exported for advanced usage (`provide`/`inject` without the full plugin)
- **`SpeechErrorCode`** exported as a standalone type alongside `SpeechError`
- **`WebSpeechSTTOptions`** exported as a public type
- 12 new tests (207 total, up from 195)

### Changed
- **`VoiceInfo` shape** — `voiceURI: string` and `localService: boolean` replaced by `id: string` (unique identifier) and `label: string` (locale-aware display name via `Intl.DisplayNames`). **Breaking** for code that reads `voice.voiceURI` or `voice.localService`
- **`SpeechError.code`** — enum is now `'NOT_SUPPORTED' | 'PERMISSION_DENIED' | 'NETWORK' | 'API_ERROR' | 'RATE_LIMIT' | 'AUDIO_PLAYBACK' | 'CANCELLED' | 'UNKNOWN'`; removed `'SYNTHESIS_FAILED'`, `'NETWORK_ERROR'`, `'ABORTED'`
- **`useSpeechSynthesis` — `loadVoices()`** — return type changed from `void` to `Promise<void>`
- **`useSpeechSynthesis` — `selectedVoice`** — type changed from `Ref<VoiceInfo | null>` to `Ref<VoiceInfo | undefined>`
- **`ElevenLabsProvider` defaults** — default `voiceId` changed to `'EXAVITQu4vr4xnSDxMaL'`; default `modelId` changed from `'eleven_monolingual_v1'` to `'eleven_multilingual_v2'`
- **`VueSpeechVoiceSelect` — `voices` prop is now required** — pass `useSpeechSynthesis().voices` directly; the component no longer fetches voices internally

### Fixed
- `VueSpeechVoiceSelect` now identifies the selected voice by `voice.id` instead of the deprecated `voiceURI`, fixing selection state after a voice list reload


## [2.0.3] — 2026-07-27

### Fixed
- **`useSpeechRecognition` stops on short pauses** — `continuous` now defaults to `true` in `WebSpeechSTTProvider` (was `false`), preventing the browser from terminating the session after every sentence-length silence
- **`isListening` desync after natural end** — `WebSpeechSTTProvider` now wires `rec.onend` and exposes an `onEnd` callback on the `STTProvider` interface; `useSpeechRecognition` uses this to keep `isListening` accurate when the browser ends recognition on its own
- **Auto-restart on unexpected browser stop** — `useSpeechRecognition` now transparently restarts the provider when the session ends unexpectedly (network hiccup, tab backgrounding, etc.) while `isListening` is still `true`; calling `stop()` still terminates cleanly with no restart
- **`stop()` / `onUnmounted` ordering** — `isListening` is now set to `false` before `provider.stop()` is called, ensuring the auto-restart guard never triggers during intentional teardown

### Changed
- `WebSpeechSTTOptions.continuous` default changed from `false` → `true`
- `STTProvider` interface gains `onEnd(cb: () => void): void`

### Added
- 6 new tests covering `onEnd` callback, natural-end detection, auto-restart behaviour, and the manual-stop guard (195 total, up from 189)



## [2.0.2] — 2026-07-25

### Fixed
- Bug Fixes


## [2.0.1] — 2026-07-25

### Fixed
- Provider test mocks updated to avoid jsdom `Blob.stream()` incompatibility with Node.js 22
- Added missing `README.md` to published package
- Fixed GitHub Actions pnpm version conflict and Node.js version requirement


## [2.0.0] — 2026-07-25

### Breaking Changes

- **Named export** — `VueSpeech` is now a named export, not the default export
- **`ProviderConfig` required** — `app.use(VueSpeech, { provider: 'web' })` config is now required
- **`useTTS()` removed** — replaced by `useSpeechSynthesis()`
- **`useSTT()` removed** — replaced by `useSpeechRecognition()`
- **Component renames** — `<SpeechPlayer>` → `<VueSpeechPlayer>`, `<SpeechRecorder>` → `<VueSpeechRecorder>`, `<VoiceSelect>` → `<VueSpeechVoiceSelect>`
- **Event renames** — `@playing` → `@start`, `@stopped` → `@end`, `@failed` → `@error` (with structured `SpeechError` payload), `@result` → `@transcript`, `@done` → `@final-transcript`
- **Prop renames** — `content` → `text`, `auto` → `autoSpeak`, `language` → `lang`, `loop` → `continuous`
- **Error payload** — errors are now `SpeechError` objects (`{ code, message, cause? }`) instead of strings
- **CSS** — replaced Sass variables with CSS custom properties (`--vts-primary`, `--vts-bg`, etc.)
- **Node.js** — minimum version raised from 14 to **18**

### Added

- **Multi-provider TTS** — OpenAI, ElevenLabs, and Azure Cognitive Services providers alongside native Web Speech
- **`useStreamingTTS()`** — pipe `AsyncIterable<string>` LLM token streams to speech with sentence-boundary buffering
- **`useVoiceQueue()`** — FIFO queue composable for sequential utterance playback
- **TypeScript** — full declarations shipped in the package; no `@types/*` needed
- **SSR-safe** — all browser API access guarded by `typeof window !== 'undefined'`
- **`baseURL` option** — all AI providers accept a `baseURL` to route through a server-side proxy
- **189 tests** — unit, component and integration tests via Vitest + `@vue/test-utils`
- **Provenance publishing** — NPM publish includes SLSA provenance via GitHub Actions OIDC
- **VitePress docs** — full documentation site at `apps/docs/`

### Changed

- Rewritten in TypeScript with Vite 6 library mode (dual ESM + CJS output)
- Package sub-path exports: `vue-text-to-speech/providers/openai`, `.../elevenlabs`, `.../azure`
- Components use CSS custom properties for theming (no UI framework dependency)

---

## [1.x] — Legacy

See [GitHub releases](https://github.com/VasuLaravel/vue-text-to-speech/releases) for v1 history.