voice-router-dev 0.5.4 → 0.6.0

package/CHANGELOG.md

@@ -5,6 +5,374 @@ All notable changes to this project will be documented in this file.
 The format is based on [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).
 
+## [0.6.0] - 2026-01-11
+
+### Added
+
+#### OpenAI Official Spec Integration
+
+OpenAI types now auto-generated from the official [Stainless-hosted OpenAPI spec](https://app.stainless.com/api/spec/documented/openai/openapi.documented.yml):
+
+```typescript
+import { OpenAIModel, OpenAIResponseFormat } from 'voice-router-dev/constants'
+import type {
+  RealtimeSessionCreateRequest,
+  RealtimeTranscriptionSessionCreateRequest,
+  CreateTranscriptionResponseDiarizedJson
+} from 'voice-router-dev'
+
+// All models from official spec
+const model = OpenAIModel["gpt-4o-transcribe-diarize"]
+
+// Response formats including diarization
+const format = OpenAIResponseFormat.diarized_json
+```
+
+**What changed:**
+- **Single source of truth**: Stainless live spec (auto-updated by OpenAI)
+- **54 schemas** generated (up from 15 manual types)
+- **7 endpoints** included: batch audio + realtime streaming
+- **Diarization types** now from official spec (`CreateTranscriptionResponseDiarizedJson`)
+- **Realtime API types**: `RealtimeSessionCreateRequest`, `RealtimeTranscriptionSessionCreateRequest`, `VadConfig`, etc.
+
+**New models in `OpenAIModel`:**
+- `whisper-1` - Open source Whisper V2
+- `gpt-4o-transcribe` - GPT-4o based transcription
+- `gpt-4o-mini-transcribe` - Faster, cost-effective
+- `gpt-4o-mini-transcribe-2025-12-15` - Dated version
+- `gpt-4o-transcribe-diarize` - With speaker diarization
+
+**New response formats in `OpenAIResponseFormat`:**
+- `diarized_json` - JSON with speaker annotations (requires `gpt-4o-transcribe-diarize`)
+
+#### OpenAI Realtime Streaming Types
+
+WebSocket event types for OpenAI Realtime API:
+
+```typescript
+import { OpenAIStreamingTypes } from 'voice-router-dev'
+
+// Session creation
+const session: OpenAIStreamingTypes.RealtimeSessionConfig = {
+  modalities: ['text', 'audio'],
+  voice: 'ash',
+  input_audio_format: 'pcm16',
+  input_audio_transcription: { model: 'whisper-1' },
+  turn_detection: { type: 'server_vad', threshold: 0.6 }
+}
+
+// WebSocket event handling
+type ServerEvent = OpenAIStreamingTypes.RealtimeServerEvent
+type ClientEvent = OpenAIStreamingTypes.RealtimeClientEvent
+```
+
+**Endpoints:**
+- OpenAI: `wss://api.openai.com/v1/realtime?model=gpt-4o-realtime-preview`
+- Azure OpenAI: `wss://{endpoint}/openai/realtime?deployment={model}&api-version={version}`
+
+#### Soniox Provider (8th Provider)
+
+New adapter for [Soniox](https://soniox.com) speech-to-text with batch and streaming support:
+
+```typescript
+import { createSonioxAdapter, SonioxLanguages } from 'voice-router-dev'
+
+const adapter = createSonioxAdapter({
+  apiKey: process.env.SONIOX_API_KEY
+})
+
+// Batch transcription
+const result = await adapter.transcribe({
+  type: 'url',
+  url: 'https://example.com/audio.mp3'
+}, {
+  language: 'en',
+  diarization: true
+})
+
+// Real-time streaming
+const session = await adapter.transcribeStream({
+  language: 'en',
+  sampleRate: 16000
+}, {
+  onTranscript: (event) => console.log(event.text),
+  onError: (error) => console.error(error)
+})
+
+// Dynamic model/language discovery
+const models = await adapter.getModels()
+const languages = await adapter.getLanguagesForModel('stt-rt-preview')
+```
+
+**Features:**
+- Batch transcription via URL or file upload
+- Real-time WebSocket streaming with endpoint detection
+- Speaker diarization
+- Language identification (auto-detect)
+- Translation support (one-way and bidirectional)
+- Custom vocabulary via structured context
+- 60+ supported languages
+
+**Generated types from OpenAPI spec (`api.soniox.com/v1/openapi.json`):**
+- `SonioxLanguages` - Array of `{code, name}` for all 60 languages
+- `SonioxLanguageCodes` - ISO 639-1 language codes
+- `SonioxLanguageLabels` - Code-to-name mapping
+- 90+ schema types via Orval (Transcription, Model, Language, etc.)
+
+#### Speechmatics Batch API Type Generation
+
+Full type generation from Speechmatics SDK batch spec (`speechmatics-batch.yml`):
+
+```typescript
+import type { JobConfig, RetrieveTranscriptResponse } from 'voice-router-dev'
+import { OperatingPoint, TranscriptionConfigDiarization } from 'voice-router-dev'
+
+// Use generated enums instead of hardcoded strings
+const config: JobConfig = {
+  type: 'transcription',
+  transcription_config: {
+    language: 'en',
+    operating_point: OperatingPoint.enhanced,
+    diarization: TranscriptionConfigDiarization.speaker
+  }
+}
+```
+
+**Generated from SDK spec:**
+- 100+ TypeScript types from `speechmatics-batch.yml`
+- Enums: `OperatingPoint`, `TranscriptionConfigDiarization`, `SummarizationConfigSummaryType`, `SummarizationConfigSummaryLength`, `JobDetailsStatus`
+- Removed manual `src/types/speechmatics.ts` (replaced by generated types)
+
+#### Soniox Field Configs
+
+Field config functions for Soniox now available:
+
+```typescript
+import {
+  getSonioxTranscriptionFields,
+  getSonioxStreamingFields,
+  getSonioxListFilterFields,
+  getSonioxFieldConfigs
+} from 'voice-router-dev/field-configs'
+
+const fields = getSonioxTranscriptionFields()
+// → [{ name: 'model', type: 'string', ... }, { name: 'language_hints', ... }, ...]
+```
+
+#### Field Config Coverage (All Providers)
+
+| Provider     | Transcription | Streaming | List Filters | Update Config |
+|--------------|:-------------:|:---------:|:------------:|:-------------:|
+| Gladia       | ✓             | ✓         | ✓            | -             |
+| Deepgram     | ✓             | ✓         | ✓            | -             |
+| AssemblyAI   | ✓             | ✓         | ✓            | ✓             |
+| OpenAI       | ✓             | -         | -            | -             |
+| Speechmatics | ✓             | ✓         | ✓            | ✓             |
+| Soniox       | ✓             | ✓         | ✓            | -             |
+| Azure        | -             | -         | -            | -             |
+
+> **Note:** Azure field configs not yet implemented (no OpenAPI spec available).
+
+#### Zod Schema Exports Reference
+
+All generated Zod schemas are exported for direct use with `zodToFieldConfigs()`:
+
+| Export Name               | Provider     | Source                          |
+|---------------------------|--------------|----------------------------------|
+| `GladiaZodSchemas`        | Gladia       | OpenAPI spec                     |
+| `DeepgramZodSchemas`      | Deepgram     | OpenAPI spec                     |
+| `AssemblyAIZodSchemas`    | AssemblyAI   | OpenAPI spec                     |
+| `OpenAIZodSchemas`        | OpenAI       | OpenAPI spec                     |
+| `SpeechmaticsZodSchemas`  | Speechmatics | OpenAPI spec (batch)             |
+| `SonioxApiZodSchemas`     | Soniox       | OpenAPI spec (batch)             |
+| `SonioxStreamingZodSchemas` | Soniox     | Manual spec (real-time WebSocket)|
+
+> **Note on manual specs:** Soniox and Deepgram streaming types are manually maintained because
+> these providers do not publish AsyncAPI specs for their WebSocket APIs. Types were extracted
+> from their official SDKs (`@soniox/speech-to-text-web` and `@deepgram/sdk`). The REST API
+> types are auto-synced from their OpenAPI specs. If these providers publish AsyncAPI specs
+> in the future, we will switch to auto-generation.
+
+```typescript
+import { zodToFieldConfigs, SonioxApiZodSchemas } from 'voice-router-dev'
+
+// Extract fields from any Zod schema
+const transcriptionFields = zodToFieldConfigs(SonioxApiZodSchemas.createTranscriptionBody)
+```
+
+#### SDK Generation Pipeline Diagram
+
+New auto-generated Mermaid diagram showing the SDK generation flow:
+
+```bash
+pnpm openapi:diagram
+```
+
+Generates `docs/sdk-generation-pipeline.mmd` from codebase analysis:
+- Analyzes `sync-specs.js` for remote/manual spec sources
+- Extracts orval config for API/Zod generation
+- Maps streaming type sync scripts
+- Includes consumer layer (router, webhooks, adapters)
+- Shows public API exports
+
+### Changed
+
+- **OpenAI spec source**: Now uses Stainless live spec instead of manual `openai-whisper-openapi.yml`
+- **`fix-openai-spec.js`**: Filters full OpenAI API to audio + realtime endpoints only
+- **OpenAI adapter**: Uses `OpenAIModel` constants instead of hardcoded strings
+- **Provider capabilities**: OpenAI now shows `streaming: true` (via Realtime API)
+- **Azure adapter**: Uses generated enums instead of hardcoded strings, removed `any` type casts
+- **Speechmatics adapter** now uses generated enums instead of hardcoded string values
+- **Speechmatics adapter** fixed API structure: `sentiment_analysis_config` and `summarization_config` moved to job level (was incorrectly in `transcription_config`)
+- **Speechmatics adapter** fixed `additional_vocab` format: now uses `{content: string}[]` per spec
+- **Speechmatics adapter** fixed `speaker_diarization_config`: uses `speaker_sensitivity` (not `max_speakers`)
+- **Soniox language codes** now generated from OpenAPI spec (60 languages vs 28 hardcoded)
+- OpenAPI sync scripts now include Speechmatics batch spec and Soniox specs
+- Added `openapi:generate:speechmatics`, `openapi:generate:soniox`, `openapi:clean:speechmatics`, `openapi:clean:soniox` scripts
+- Added `openapi:sync-soniox-languages` to generate flow
+
+### Fixed
+
+- OpenAI model values now stay in sync with official spec
+- `OpenAIResponseFormat` now includes `diarized_json` from official spec
+- OpenAI `languageDetection` capability is now `true` (language is optional in request)
+- Azure `languageDetection` capability fixed (was incorrectly `false`)
+- Azure `customVocabulary` capability fixed
+- AssemblyAI/Speechmatics streaming types now survive `openapi:clean` (stored in `specs/`)
+- Speechmatics batch field configs now work (was returning empty array)
+- Speechmatics webhook handler now uses generated `RetrieveTranscriptResponse` type
+- **AssemblyAI streaming field configs** now include SDK v3 fields (`keyterms`, `keytermsPrompt`, `speechModel`, `languageDetection`, etc.) - sync script parses both AsyncAPI spec and SDK TypeScript types
+
+#### Soniox Regional Endpoints (Sovereign Cloud)
+
+Regional endpoint support for Soniox data residency:
+
+```typescript
+import { createSonioxAdapter, SonioxRegion } from 'voice-router-dev'
+
+const adapter = createSonioxAdapter({
+  apiKey: process.env.SONIOX_EU_API_KEY,
+  region: SonioxRegion.eu  // EU data residency
+})
+```
+
+| Region | REST API | WebSocket |
+|--------|----------|-----------|
+| `us` (default) | `api.soniox.com` | `stt-rt.soniox.com` |
+| `eu` | `api.eu.soniox.com` | `stt-rt.eu.soniox.com` |
+| `jp` | `api.jp.soniox.com` | `stt-rt.jp.soniox.com` |
+
+**Note:** Soniox API keys are region-specific. Each project is created with a specific region, and the API key only works with that region's endpoint.
+
+---
+
+## [0.5.5] - 2026-01-09
+
+### Changed
+
+- Dynamic streaming types synced from AsyncAPI/SDK specs for all providers
+- Deepgram streaming params derived from official SDK (`TranscriptionSchema.ts`)
+- AssemblyAI streaming Zod auto-generated from SDK types
+- Speechmatics streaming types from AsyncAPI spec
+
+---
+
+## [0.5.0] - 2026-01-09
+
+### Added
+
+#### Zero-Hardcoding Field Configs
+
+All field configs are now derived from Zod schemas at runtime - zero hardcoded field definitions:
+
+```typescript
+import { zodToFieldConfigs, DeepgramZodSchemas } from 'voice-router-dev'
+
+// Extract fields directly from generated Zod schemas
+const fields = zodToFieldConfigs(DeepgramZodSchemas.listenV1MediaTranscribeQueryParams)
+// → [{ name, type, description, options, default, min, max, ... }]
+
+// Or use pre-built helpers
+import { getDeepgramTranscriptionFields } from 'voice-router-dev'
+const deepgramFields = getDeepgramTranscriptionFields() // 36 fields from Zod
+```
+
+**Exports:**
+- `zodToFieldConfigs(schema)` - Extract field configs from any Zod schema
+- `filterFields(fields, names)` - Include only specified fields
+- `excludeFields(fields, names)` - Exclude specified fields
+- `GladiaZodSchemas`, `DeepgramZodSchemas`, `AssemblyAIZodSchemas`, etc.
+
+#### 100% Streaming Field Coverage
+
+| Provider   | Fields | Source |
+|------------|--------|--------|
+| Gladia     | 10     | OpenAPI Zod |
+| Deepgram   | 30     | OpenAPI Zod |
+| AssemblyAI | 13     | SDK Zod |
+
+### Changed
+
+- Deleted `streaming-field-schemas.ts` (was 461 lines of hardcoding)
+- Rewrote `field-configs.ts`: 890 → 205 lines (zero hardcoded fields)
+- All field configs now derived from Zod schemas at runtime
+
+---
+
+## [0.4.1] - 2026-01-09
+
+### Added
+
+#### Provider Metadata Exports for UI Rendering
+
+Static runtime data derived from OpenAPI specs and adapter definitions:
+
+```typescript
+import {
+  ProviderCapabilitiesMap,
+  CapabilityLabels,
+  LanguageLabels,
+  AllLanguageCodes,
+  ProviderDisplayNames,
+  StreamingProviders,
+  BatchOnlyProviders
+} from 'voice-router-dev/provider-metadata'
+
+// Capability matrix for all providers
+const capabilities = ProviderCapabilitiesMap['deepgram']
+// → { streaming: true, diarization: true, ... }
+
+// Language dropdown data
+const languages = AllLanguageCodes['gladia']
+// → ['en', 'es', 'fr', ...]
+const label = LanguageLabels['en'] // → 'English'
+```
+
+#### Browser-Safe Subpath Exports
+
+New subpath exports with no `node:crypto` dependency:
+
+```typescript
+// Browser-safe imports
+import { AllFieldConfigs } from 'voice-router-dev/field-configs'
+import { ProviderCapabilitiesMap } from 'voice-router-dev/provider-metadata'
+
+// Full SDK (server-side only)
+import { VoiceRouter } from 'voice-router-dev'
+```
+
+**Exports:**
+- `voice-router-dev/constants` - Enums only (existing)
+- `voice-router-dev/field-configs` - Field configurations
+- `voice-router-dev/provider-metadata` - Capabilities, languages, display names
+
+### Changed
+
+- Types refactored to shared `src/types/core.ts` for browser compatibility
+- `router/types.ts` re-exports from `core.ts` (no duplication)
+
+---
+
 ## [0.3.7] - 2026-01-09
 
 ### Added

package/README.md

@@ -1,6 +1,6 @@
 # Voice Router SDK
 
-> Universal speech-to-text router for 6+ transcription providers with a single, unified API.
+> Universal speech-to-text router for 8 transcription providers with a single, unified API.
 
 [![npm version](https://badge.fury.io/js/voice-router-dev.svg)](https://www.npmjs.com/package/voice-router-dev)
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
@@ -8,7 +8,7 @@
 
 ## Why Voice Router?
 
-Switch between speech-to-text providers **without changing your code**. One API for Gladia, AssemblyAI, Deepgram, Azure, OpenAI Whisper, and Speechmatics.
+Switch between speech-to-text providers **without changing your code**. One API for Gladia, AssemblyAI, Deepgram, Azure, OpenAI Whisper, Speechmatics, and Soniox.
 
 ```typescript
 import { VoiceRouter } from 'voice-router-dev';
@@ -31,7 +31,7 @@ const result = await router.transcribe(audio, {
 - **Provider-Agnostic** - Switch providers with one line
 - **Unified API** - Same interface for all providers
 - **Webhook Normalization** - Auto-detect and parse webhooks
-- **Real-time Streaming** - WebSocket support (Gladia, AssemblyAI, Deepgram)
+- **Real-time Streaming** - WebSocket support (Gladia, AssemblyAI, Deepgram, Soniox, OpenAI Realtime)
 - **Advanced Features** - Diarization, sentiment, summarization, chapters, entities
 - **Type-Safe** - Full TypeScript support with OpenAPI-generated types
 - **Typed Extended Data** - Access provider-specific features with full autocomplete
@@ -46,8 +46,9 @@ const result = await router.transcribe(audio, {
 | **AssemblyAI** | Yes | Real-time | HMAC | Chapters, entities, content moderation |
 | **Deepgram** | Sync | WebSocket | Yes | PII redaction, keyword boosting |
 | **Azure STT** | Async | No | HMAC | Custom models, language ID |
-| **OpenAI Whisper** | Sync | No | No | gpt-4o, diarization |
+| **OpenAI** | Sync | Realtime | No | gpt-4o, diarization, Realtime API |
 | **Speechmatics** | Async | No | Query params | High accuracy, summarization |
+| **Soniox** | Yes | WebSocket | No | 60+ languages, translation, regions |
 
 ## Installation
 
@@ -371,8 +372,9 @@ Provider-specific implementations:
 - `AssemblyAIAdapter` - AssemblyAI transcription
 - `DeepgramAdapter` - Deepgram transcription
 - `AzureSTTAdapter` - Azure Speech-to-Text
-- `OpenAIWhisperAdapter` - OpenAI Whisper
+- `OpenAIWhisperAdapter` - OpenAI Whisper + Realtime API
 - `SpeechmaticsAdapter` - Speechmatics transcription
+- `SonioxAdapter` - Soniox transcription (batch + streaming)
 
 ## TypeScript Support
 
@@ -651,6 +653,23 @@ router.registerAdapter(new SpeechmaticsAdapter());
 
 Get your API key: https://speechmatics.com
 
+### Soniox
+```typescript
+import { VoiceRouter, SonioxAdapter, SonioxRegion } from 'voice-router-dev';
+
+const router = new VoiceRouter({
+  providers: {
+    soniox: {
+      apiKey: 'YOUR_KEY',
+      region: SonioxRegion.us  // or 'eu', 'jp'
+    }
+  }
+});
+router.registerAdapter(new SonioxAdapter());
+```
+
+Get your API key: https://soniox.com
+
 ## Contributing
 
 Contributions welcome! Please read our [Contributing Guide](CONTRIBUTING.md).

package/dist/constants.d.mts

@@ -736,10 +736,46 @@ declare const DeepgramRegion: {
     /** European Union endpoint */
     readonly eu: "eu";
 };
+/**
+ * Soniox regional endpoints (Sovereign Cloud)
+ *
+ * Soniox offers regional endpoints for data residency compliance.
+ * All audio, transcripts, and logs stay fully in-region.
+ *
+ * | Region | REST API | WebSocket (Real-time) |
+ * |--------|----------|----------------------|
+ * | US (default) | api.soniox.com | stt-rt.soniox.com |
+ * | EU | api.eu.soniox.com | stt-rt.eu.soniox.com |
+ * | Japan | api.jp.soniox.com | stt-rt.jp.soniox.com |
+ *
+ * **Coming soon:** Korea, Australia, India, Canada, Saudi Arabia, UK, Brazil
+ *
+ * @example
+ * ```typescript
+ * import { SonioxRegion } from 'voice-router-dev/constants'
+ *
+ * const adapter = createSonioxAdapter({
+ *   apiKey: process.env.SONIOX_API_KEY,
+ *   region: SonioxRegion.eu
+ * })
+ * ```
+ *
+ * @see https://soniox.com/docs/stt/data-residency - Official data residency docs
+ */
+declare const SonioxRegion: {
+    /** United States (default) */
+    readonly us: "us";
+    /** European Union */
+    readonly eu: "eu";
+    /** Japan */
+    readonly jp: "jp";
+};
 /** Speechmatics region type derived from const object */
 type SpeechmaticsRegionType = (typeof SpeechmaticsRegion)[keyof typeof SpeechmaticsRegion];
 /** Deepgram region type derived from const object */
 type DeepgramRegionType = (typeof DeepgramRegion)[keyof typeof DeepgramRegion];
+/** Soniox region type derived from const object */
+type SonioxRegionType = (typeof SonioxRegion)[keyof typeof SonioxRegion];
 /**
  * Deepgram TTS voice models
  *
@@ -890,7 +926,12 @@ type DeepgramTTSSampleRateType = (typeof DeepgramTTSSampleRate)[keyof typeof Dee
 /**
  * OpenAI Whisper transcription models
  *
- * Values: `whisper-1`, `gpt-4o-transcribe`, `gpt-4o-mini-transcribe`, `gpt-4o-transcribe-diarize`
+ * Values from official spec (auto-synced from Stainless):
+ * - `whisper-1`: Open source Whisper V2 model
+ * - `gpt-4o-transcribe`: GPT-4o based transcription (more accurate)
+ * - `gpt-4o-mini-transcribe`: Faster, cost-effective GPT-4o mini
+ * - `gpt-4o-mini-transcribe-2025-12-15`: Dated version of GPT-4o mini
+ * - `gpt-4o-transcribe-diarize`: GPT-4o with speaker diarization
  *
  * @example
  * ```typescript
@@ -898,28 +939,36 @@ type DeepgramTTSSampleRateType = (typeof DeepgramTTSSampleRate)[keyof typeof Dee
  *
  * { model: OpenAIModel["whisper-1"] }
  * { model: OpenAIModel["gpt-4o-transcribe"] }
+ * { model: OpenAIModel["gpt-4o-transcribe-diarize"] }
  * ```
  */
 declare const OpenAIModel: {
     readonly "whisper-1": "whisper-1";
-    readonly "gpt-4o-mini-transcribe": "gpt-4o-mini-transcribe";
     readonly "gpt-4o-transcribe": "gpt-4o-transcribe";
+    readonly "gpt-4o-mini-transcribe": "gpt-4o-mini-transcribe";
+    readonly "gpt-4o-mini-transcribe-2025-12-15": "gpt-4o-mini-transcribe-2025-12-15";
     readonly "gpt-4o-transcribe-diarize": "gpt-4o-transcribe-diarize";
 };
 /**
  * OpenAI transcription response formats
  *
- * Values: `json`, `text`, `srt`, `verbose_json`, `vtt`, `diarized_json`
+ * Values from official spec (auto-synced from Stainless):
+ * - `json`: Basic JSON response
+ * - `text`: Plain text
+ * - `srt`: SRT subtitle format
+ * - `verbose_json`: Detailed JSON with timestamps
+ * - `vtt`: VTT subtitle format
+ * - `diarized_json`: JSON with speaker annotations (gpt-4o-transcribe-diarize only)
  *
- * Note: `diarized_json` is only available with `gpt-4o-transcribe-diarize` model.
- * GPT-4o transcribe models only support `json` format.
+ * Note: GPT-4o transcribe models only support `json` format.
+ * For diarization, use `diarized_json` with `gpt-4o-transcribe-diarize` model.
  *
  * @example
  * ```typescript
  * import { OpenAIResponseFormat } from 'voice-router-dev/constants'
  *
  * { responseFormat: OpenAIResponseFormat.verbose_json }
- * { responseFormat: OpenAIResponseFormat.srt }
+ * { responseFormat: OpenAIResponseFormat.diarized_json }
  * ```
  */
 declare const OpenAIResponseFormat: {
@@ -935,4 +984,4 @@ type OpenAIModelType = (typeof OpenAIModel)[keyof typeof OpenAIModel];
 /** OpenAI response format type derived from const object */
 type OpenAIResponseFormatType = (typeof OpenAIResponseFormat)[keyof typeof OpenAIResponseFormat];
 
-export { AssemblyAIEncoding, type AssemblyAIEncodingType, AssemblyAISampleRate, type AssemblyAISampleRateType, AssemblyAISpeechModel, type AssemblyAISpeechModelType, AssemblyAIStatus, type AssemblyAIStatusType, AzureStatus, type AzureStatusType, DeepgramCallbackMethod, type DeepgramCallbackMethodType, DeepgramEncoding, type DeepgramEncodingType, DeepgramIntentMode, type DeepgramIntentModeType, DeepgramModel, type DeepgramModelType, DeepgramRedact, type DeepgramRedactType, DeepgramRegion, type DeepgramRegionType, DeepgramSampleRate, type DeepgramSampleRateType, DeepgramStatus, type DeepgramStatusType, DeepgramTTSContainer, type DeepgramTTSContainerType, DeepgramTTSEncoding, type DeepgramTTSEncodingType, DeepgramTTSModel, type DeepgramTTSModelType, DeepgramTTSSampleRate, type DeepgramTTSSampleRateType, DeepgramTopicMode, type DeepgramTopicModeType, GladiaBitDepth, type GladiaBitDepthType, GladiaEncoding, type GladiaEncodingType, GladiaLanguage, type GladiaLanguageType, GladiaModel, type GladiaModelType, GladiaRegion, type GladiaRegionType, GladiaSampleRate, type GladiaSampleRateType, GladiaStatus, type GladiaStatusType, GladiaTranslationLanguage, type GladiaTranslationLanguageType, OpenAIModel, type OpenAIModelType, OpenAIResponseFormat, type OpenAIResponseFormatType, SpeechmaticsRegion, type SpeechmaticsRegionType };
+export { AssemblyAIEncoding, type AssemblyAIEncodingType, AssemblyAISampleRate, type AssemblyAISampleRateType, AssemblyAISpeechModel, type AssemblyAISpeechModelType, AssemblyAIStatus, type AssemblyAIStatusType, AzureStatus, type AzureStatusType, DeepgramCallbackMethod, type DeepgramCallbackMethodType, DeepgramEncoding, type DeepgramEncodingType, DeepgramIntentMode, type DeepgramIntentModeType, DeepgramModel, type DeepgramModelType, DeepgramRedact, type DeepgramRedactType, DeepgramRegion, type DeepgramRegionType, DeepgramSampleRate, type DeepgramSampleRateType, DeepgramStatus, type DeepgramStatusType, DeepgramTTSContainer, type DeepgramTTSContainerType, DeepgramTTSEncoding, type DeepgramTTSEncodingType, DeepgramTTSModel, type DeepgramTTSModelType, DeepgramTTSSampleRate, type DeepgramTTSSampleRateType, DeepgramTopicMode, type DeepgramTopicModeType, GladiaBitDepth, type GladiaBitDepthType, GladiaEncoding, type GladiaEncodingType, GladiaLanguage, type GladiaLanguageType, GladiaModel, type GladiaModelType, GladiaRegion, type GladiaRegionType, GladiaSampleRate, type GladiaSampleRateType, GladiaStatus, type GladiaStatusType, GladiaTranslationLanguage, type GladiaTranslationLanguageType, OpenAIModel, type OpenAIModelType, OpenAIResponseFormat, type OpenAIResponseFormatType, SonioxRegion, type SonioxRegionType, SpeechmaticsRegion, type SpeechmaticsRegionType };

package/dist/constants.d.ts

@@ -736,10 +736,46 @@ declare const DeepgramRegion: {
     /** European Union endpoint */
     readonly eu: "eu";
 };
+/**
+ * Soniox regional endpoints (Sovereign Cloud)
+ *
+ * Soniox offers regional endpoints for data residency compliance.
+ * All audio, transcripts, and logs stay fully in-region.
+ *
+ * | Region | REST API | WebSocket (Real-time) |
+ * |--------|----------|----------------------|
+ * | US (default) | api.soniox.com | stt-rt.soniox.com |
+ * | EU | api.eu.soniox.com | stt-rt.eu.soniox.com |
+ * | Japan | api.jp.soniox.com | stt-rt.jp.soniox.com |
+ *
+ * **Coming soon:** Korea, Australia, India, Canada, Saudi Arabia, UK, Brazil
+ *
+ * @example
+ * ```typescript
+ * import { SonioxRegion } from 'voice-router-dev/constants'
+ *
+ * const adapter = createSonioxAdapter({
+ *   apiKey: process.env.SONIOX_API_KEY,
+ *   region: SonioxRegion.eu
+ * })
+ * ```
+ *
+ * @see https://soniox.com/docs/stt/data-residency - Official data residency docs
+ */
+declare const SonioxRegion: {
+    /** United States (default) */
+    readonly us: "us";
+    /** European Union */
+    readonly eu: "eu";
+    /** Japan */
+    readonly jp: "jp";
+};
 /** Speechmatics region type derived from const object */
 type SpeechmaticsRegionType = (typeof SpeechmaticsRegion)[keyof typeof SpeechmaticsRegion];
 /** Deepgram region type derived from const object */
 type DeepgramRegionType = (typeof DeepgramRegion)[keyof typeof DeepgramRegion];
+/** Soniox region type derived from const object */
+type SonioxRegionType = (typeof SonioxRegion)[keyof typeof SonioxRegion];
 /**
  * Deepgram TTS voice models
  *
@@ -890,7 +926,12 @@ type DeepgramTTSSampleRateType = (typeof DeepgramTTSSampleRate)[keyof typeof Dee
 /**
  * OpenAI Whisper transcription models
  *
- * Values: `whisper-1`, `gpt-4o-transcribe`, `gpt-4o-mini-transcribe`, `gpt-4o-transcribe-diarize`
+ * Values from official spec (auto-synced from Stainless):
+ * - `whisper-1`: Open source Whisper V2 model
+ * - `gpt-4o-transcribe`: GPT-4o based transcription (more accurate)
+ * - `gpt-4o-mini-transcribe`: Faster, cost-effective GPT-4o mini
+ * - `gpt-4o-mini-transcribe-2025-12-15`: Dated version of GPT-4o mini
+ * - `gpt-4o-transcribe-diarize`: GPT-4o with speaker diarization
  *
  * @example
  * ```typescript
@@ -898,28 +939,36 @@ type DeepgramTTSSampleRateType = (typeof DeepgramTTSSampleRate)[keyof typeof Dee
  *
  * { model: OpenAIModel["whisper-1"] }
  * { model: OpenAIModel["gpt-4o-transcribe"] }
+ * { model: OpenAIModel["gpt-4o-transcribe-diarize"] }
  * ```
  */
 declare const OpenAIModel: {
     readonly "whisper-1": "whisper-1";
-    readonly "gpt-4o-mini-transcribe": "gpt-4o-mini-transcribe";
     readonly "gpt-4o-transcribe": "gpt-4o-transcribe";
+    readonly "gpt-4o-mini-transcribe": "gpt-4o-mini-transcribe";
+    readonly "gpt-4o-mini-transcribe-2025-12-15": "gpt-4o-mini-transcribe-2025-12-15";
     readonly "gpt-4o-transcribe-diarize": "gpt-4o-transcribe-diarize";
 };
 /**
  * OpenAI transcription response formats
  *
- * Values: `json`, `text`, `srt`, `verbose_json`, `vtt`, `diarized_json`
+ * Values from official spec (auto-synced from Stainless):
+ * - `json`: Basic JSON response
+ * - `text`: Plain text
+ * - `srt`: SRT subtitle format
+ * - `verbose_json`: Detailed JSON with timestamps
+ * - `vtt`: VTT subtitle format
+ * - `diarized_json`: JSON with speaker annotations (gpt-4o-transcribe-diarize only)
  *
- * Note: `diarized_json` is only available with `gpt-4o-transcribe-diarize` model.
- * GPT-4o transcribe models only support `json` format.
+ * Note: GPT-4o transcribe models only support `json` format.
+ * For diarization, use `diarized_json` with `gpt-4o-transcribe-diarize` model.
  *
  * @example
  * ```typescript
  * import { OpenAIResponseFormat } from 'voice-router-dev/constants'
  *
  * { responseFormat: OpenAIResponseFormat.verbose_json }
- * { responseFormat: OpenAIResponseFormat.srt }
+ * { responseFormat: OpenAIResponseFormat.diarized_json }
  * ```
  */
 declare const OpenAIResponseFormat: {
@@ -935,4 +984,4 @@ type OpenAIModelType = (typeof OpenAIModel)[keyof typeof OpenAIModel];
 /** OpenAI response format type derived from const object */
 type OpenAIResponseFormatType = (typeof OpenAIResponseFormat)[keyof typeof OpenAIResponseFormat];
 
-export { AssemblyAIEncoding, type AssemblyAIEncodingType, AssemblyAISampleRate, type AssemblyAISampleRateType, AssemblyAISpeechModel, type AssemblyAISpeechModelType, AssemblyAIStatus, type AssemblyAIStatusType, AzureStatus, type AzureStatusType, DeepgramCallbackMethod, type DeepgramCallbackMethodType, DeepgramEncoding, type DeepgramEncodingType, DeepgramIntentMode, type DeepgramIntentModeType, DeepgramModel, type DeepgramModelType, DeepgramRedact, type DeepgramRedactType, DeepgramRegion, type DeepgramRegionType, DeepgramSampleRate, type DeepgramSampleRateType, DeepgramStatus, type DeepgramStatusType, DeepgramTTSContainer, type DeepgramTTSContainerType, DeepgramTTSEncoding, type DeepgramTTSEncodingType, DeepgramTTSModel, type DeepgramTTSModelType, DeepgramTTSSampleRate, type DeepgramTTSSampleRateType, DeepgramTopicMode, type DeepgramTopicModeType, GladiaBitDepth, type GladiaBitDepthType, GladiaEncoding, type GladiaEncodingType, GladiaLanguage, type GladiaLanguageType, GladiaModel, type GladiaModelType, GladiaRegion, type GladiaRegionType, GladiaSampleRate, type GladiaSampleRateType, GladiaStatus, type GladiaStatusType, GladiaTranslationLanguage, type GladiaTranslationLanguageType, OpenAIModel, type OpenAIModelType, OpenAIResponseFormat, type OpenAIResponseFormatType, SpeechmaticsRegion, type SpeechmaticsRegionType };
+export { AssemblyAIEncoding, type AssemblyAIEncodingType, AssemblyAISampleRate, type AssemblyAISampleRateType, AssemblyAISpeechModel, type AssemblyAISpeechModelType, AssemblyAIStatus, type AssemblyAIStatusType, AzureStatus, type AzureStatusType, DeepgramCallbackMethod, type DeepgramCallbackMethodType, DeepgramEncoding, type DeepgramEncodingType, DeepgramIntentMode, type DeepgramIntentModeType, DeepgramModel, type DeepgramModelType, DeepgramRedact, type DeepgramRedactType, DeepgramRegion, type DeepgramRegionType, DeepgramSampleRate, type DeepgramSampleRateType, DeepgramStatus, type DeepgramStatusType, DeepgramTTSContainer, type DeepgramTTSContainerType, DeepgramTTSEncoding, type DeepgramTTSEncodingType, DeepgramTTSModel, type DeepgramTTSModelType, DeepgramTTSSampleRate, type DeepgramTTSSampleRateType, DeepgramTopicMode, type DeepgramTopicModeType, GladiaBitDepth, type GladiaBitDepthType, GladiaEncoding, type GladiaEncodingType, GladiaLanguage, type GladiaLanguageType, GladiaModel, type GladiaModelType, GladiaRegion, type GladiaRegionType, GladiaSampleRate, type GladiaSampleRateType, GladiaStatus, type GladiaStatusType, GladiaTranslationLanguage, type GladiaTranslationLanguageType, OpenAIModel, type OpenAIModelType, OpenAIResponseFormat, type OpenAIResponseFormatType, SonioxRegion, type SonioxRegionType, SpeechmaticsRegion, type SpeechmaticsRegionType };