build: sync deps for audio module skeleton

Author: wizenheimer

package.json

@@ -33,6 +33,7 @@
     "typescript": "^5.7.3"
   },
   "dependencies": {
-    "@huggingface/transformers": "^3.3.3"
+    "@huggingface/transformers": "^3.3.3",
+    "phonemizer": "^1.2.1"
   }
 }

pnpm-lock.yaml

@@ -7,6 +7,7 @@ import { ProgressTracker } from './ProgressTracker';
 import { WebGPUChecker } from './WebGPUChecker';
 import { GenerationModule } from './GenerationModule';
 import { EmbeddingsModule } from './EmbeddingsModule';
+import { AudioModule } from './AudioModule';
 
 import {
   TinyLMOptions,
@@ -17,7 +18,9 @@ import {
   CompletionChunk,
   ModelLoadOptions,
   EmbeddingCreateOptions,
-  EmbeddingResult
+  EmbeddingResult,
+  SpeechCreateOptions,
+  SpeechResult
 } from './types';
 
 /**
@@ -33,6 +36,7 @@ export class TinyLM {
   // Modules
   private generationModule: GenerationModule;
   private embeddingsModule: EmbeddingsModule;
+  private audioModule: AudioModule;
 
   // API structure
   readonly chat: {
@@ -46,13 +50,22 @@ export class TinyLM {
     create: (options: EmbeddingCreateOptions) => Promise<EmbeddingResult>;
   };
 
+  readonly audio: {
+    speech: {
+      create: (options: SpeechCreateOptions) => Promise<SpeechResult>;
+    };
+  };
+
   readonly models: {
     load: (options: ModelLoadOptions) => Promise<any>;
     offload: (options: { model: string }) => Promise<boolean>;
     interrupt: () => boolean;
     reset: () => void;
     check: () => Promise<CapabilityInfo>;
     list: () => string[];
+    loadTTS: (options: { model: string }) => Promise<boolean>;
+    offloadTTS: (options: { model: string }) => Promise<boolean>;
+    listTTS: () => string[];
   };
 
   /**
@@ -79,6 +92,7 @@ export class TinyLM {
     // Initialize modules
     this.generationModule = new GenerationModule(this);
     this.embeddingsModule = new EmbeddingsModule(this);
+    this.audioModule = new AudioModule(this);
 
     // Create API structure similar to OpenAI
     this.chat = {
@@ -92,14 +106,23 @@ export class TinyLM {
       create: this.embeddingsModule.create.bind(this.embeddingsModule)
     };
 
+    this.audio = {
+      speech: {
+        create: this.audioModule.createSpeech.bind(this.audioModule)
+      }
+    };
+
     // Model management API
     this.models = {
       load: this.generationModule.loadModel.bind(this.generationModule),
       offload: this.generationModule.offloadModel.bind(this.generationModule),
       interrupt: this.generationModule.interrupt.bind(this.generationModule),
       reset: this.generationModule.reset.bind(this.generationModule),
       check: this.checkCapabilities.bind(this),
-      list: () => Array.from(this.generationModule.getModelRegistry().keys())
+      list: () => Array.from(this.generationModule.getModelRegistry().keys()),
+      loadTTS: this.audioModule.loadModel.bind(this.audioModule),
+      offloadTTS: this.audioModule.offloadModel.bind(this.audioModule),
+      listTTS: this.audioModule.getLoadedModels.bind(this.audioModule)
     };
   }
 
@@ -173,6 +196,11 @@ export class TinyLM {
           : undefined
       });
 
+      await this.audioModule.init({
+        ttsModels: options.ttsModels || [],
+        lazyLoad: options.lazyLoad
+      });
+
       this.initialized = true;
     }
     return this;

src/TinyLM.ts

@@ -8,6 +8,7 @@ export { TinyLM } from './TinyLM';
 // Export modules
 export { GenerationModule } from './GenerationModule';
 export { EmbeddingsModule } from './EmbeddingsModule';
+export { AudioModule } from './AudioModule';
 
 // Export utilities
 export { tryGarbageCollection, detectEnvironment } from './utils';
@@ -17,6 +18,7 @@ export { FileProgressTracker } from './FileProgressTracker';
 export { WebGPUChecker } from './WebGPUChecker';
 export { GenerationController, InterruptableStoppingCriteria } from './GenerationController';
 export { createStreamer } from './TextStreamHandler';
+export { TTSEngine } from './TTSEngine';
 
 // Export type definitions
 export type {
@@ -40,6 +42,10 @@ export type {
   EmbeddingCreateOptions,
   EmbeddingResult,
 
+  // Audio types
+  SpeechCreateOptions,
+  SpeechResult,
+
   // Model types
   ModelLoadOptions,
   InitOptions,

src/index.ts

@@ -254,11 +254,12 @@ export interface ModelLoadOptions {
 }
 
 /**
- * Interface for model initialization options
+ * Update InitOptions to include ttsModels
  */
 export interface InitOptions {
   models?: string[];
   embeddingModels?: string[];
+  ttsModels?: string[];
   lazyLoad?: boolean;
   [key: string]: any;
 }
@@ -290,3 +291,31 @@ export interface EmbeddingResult {
     total_tokens: number;
   };
 }
+
+
+/**
+ * Interface for speech creation options
+ */
+export interface SpeechCreateOptions {
+  model?: string;
+  input: string;
+  voice?: string;
+  response_format?: 'mp3' | 'wav';
+  speed?: number;
+  [key: string]: any;
+}
+
+/**
+ * Interface for speech result
+ */
+export interface SpeechResult {
+  id: string;
+  object: string;
+  created: number;
+  model: string;
+  audio: ArrayBuffer;
+  content_type: string;
+  _tinylm?: {
+    time_ms: number;
+  };
+}