@contentgrowth/llm-service 1.3.0 → 1.3.2

package/dist/index.d.cts

@@ -61,6 +61,13 @@ declare class BaseLLMProvider {
         imageData: string;
         mimeType: string;
     }>;
+    /**
+     * Text-to-Speech synthesis
+     * @param {string} text - Text to synthesize
+     * @param {Object} options - TTS options (voice, model, etc.)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, options?: any): Promise<Buffer>;
     /**
      * Start video generation (returns operation name for polling)
      * @param {string} prompt
@@ -344,6 +351,14 @@ declare class LLMService {
      * @param {Object} options - Generation options (aspectRatio, images, etc.)
      */
     imageGeneration(prompt: string, tenantId: string, systemPrompt: string, options?: any): Promise<any>;
+    /**
+     * Text-to-Speech synthesis
+     * @param {string} text - Text to synthesize
+     * @param {string} tenantId - Tenant identifier
+     * @param {Object} options - TTS options (voice, model, etc.)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, tenantId: string, options?: any): Promise<Buffer>;
     /**
      * Start a Deep Research task
      * @param {string} prompt - Research prompt/query
@@ -631,6 +646,18 @@ declare class GoogleProvider extends BaseLLMProvider {
         error: any;
     }>;
     extractWithLLM(fileData: any, mimeType: any, prompt: any, schema?: any, options?: {}): Promise<any>;
+    /**
+     * Text-to-Speech synthesis using Gemini TTS models
+     * @param {string} text - Text to synthesize
+     * @param {Object} options - TTS options
+     * @param {string} options.voice - Voice name (default: 'Aoede')
+     * @param {string} options.model - Model to use (default: from config)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, options?: {
+        voice: string;
+        model: string;
+    }): Promise<Buffer>;
 }
 
 /**

package/dist/index.d.ts

@@ -61,6 +61,13 @@ declare class BaseLLMProvider {
         imageData: string;
         mimeType: string;
     }>;
+    /**
+     * Text-to-Speech synthesis
+     * @param {string} text - Text to synthesize
+     * @param {Object} options - TTS options (voice, model, etc.)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, options?: any): Promise<Buffer>;
     /**
      * Start video generation (returns operation name for polling)
      * @param {string} prompt
@@ -344,6 +351,14 @@ declare class LLMService {
      * @param {Object} options - Generation options (aspectRatio, images, etc.)
      */
     imageGeneration(prompt: string, tenantId: string, systemPrompt: string, options?: any): Promise<any>;
+    /**
+     * Text-to-Speech synthesis
+     * @param {string} text - Text to synthesize
+     * @param {string} tenantId - Tenant identifier
+     * @param {Object} options - TTS options (voice, model, etc.)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, tenantId: string, options?: any): Promise<Buffer>;
     /**
      * Start a Deep Research task
      * @param {string} prompt - Research prompt/query
@@ -631,6 +646,18 @@ declare class GoogleProvider extends BaseLLMProvider {
         error: any;
     }>;
     extractWithLLM(fileData: any, mimeType: any, prompt: any, schema?: any, options?: {}): Promise<any>;
+    /**
+     * Text-to-Speech synthesis using Gemini TTS models
+     * @param {string} text - Text to synthesize
+     * @param {Object} options - TTS options
+     * @param {string} options.voice - Voice name (default: 'Aoede')
+     * @param {string} options.model - Model to use (default: from config)
+     * @returns {Promise<Buffer>} Audio data as Buffer
+     */
+    textToSpeech(text: string, options?: {
+        voice: string;
+        model: string;
+    }): Promise<Buffer>;
 }
 
 /**

package/dist/index.js

@@ -291,6 +291,15 @@ var BaseLLMProvider = class {
   async imageGeneration(prompt, systemPrompt, options = {}) {
     throw new Error("Image generation not supported by this provider");
   }
+  /**
+   * Text-to-Speech synthesis
+   * @param {string} text - Text to synthesize
+   * @param {Object} options - TTS options (voice, model, etc.)
+   * @returns {Promise<Buffer>} Audio data as Buffer
+   */
+  async textToSpeech(text, options = {}) {
+    throw new Error("Text-to-speech not supported by this provider");
+  }
   /**
    * Start video generation (returns operation name for polling)
    * @param {string} prompt
@@ -1171,6 +1180,52 @@ ${prompt}` : prompt;
       throw error;
     }
   }
+  /**
+   * Text-to-Speech synthesis using Gemini TTS models
+   * @param {string} text - Text to synthesize
+   * @param {Object} options - TTS options
+   * @param {string} options.voice - Voice name (default: 'Aoede')
+   * @param {string} options.model - Model to use (default: from config)
+   * @returns {Promise<Buffer>} Audio data as Buffer
+   */
+  async textToSpeech(text, options = {}) {
+    var _a, _b, _c, _d, _e;
+    const voice = options.voice || "Aoede";
+    const modelName = options.model || this.models.tts || "gemini-3.1-flash-tts-preview";
+    const requestOptions = {
+      model: modelName,
+      contents: [{
+        role: "user",
+        parts: [{ text }]
+      }],
+      config: {
+        responseModalities: ["AUDIO"],
+        speechConfig: {
+          voiceConfig: {
+            prebuiltVoiceConfig: {
+              voiceName: voice
+            }
+          }
+        }
+      }
+    };
+    try {
+      const response = await this._generateContent(requestOptions);
+      const candidate = (_a = response.candidates) == null ? void 0 : _a[0];
+      if (!candidate) {
+        throw new LLMServiceException("No candidates returned from TTS model", 500);
+      }
+      const audioPart = (_c = (_b = candidate.content) == null ? void 0 : _b.parts) == null ? void 0 : _c.find((p) => p.inlineData || p.inline_data);
+      const audioB64 = ((_d = audioPart == null ? void 0 : audioPart.inlineData) == null ? void 0 : _d.data) || ((_e = audioPart == null ? void 0 : audioPart.inline_data) == null ? void 0 : _e.data);
+      if (!audioB64) {
+        throw new LLMServiceException("No audio data returned from TTS model", 500);
+      }
+      return Buffer.from(audioB64, "base64");
+    } catch (error) {
+      console.error(`[${this.constructor.name}] textToSpeech failed (API Key: ${this._getMaskedApiKey()}):`, error);
+      throw error;
+    }
+  }
 };
 
 // src/llm/providers/anthropic-provider.js
@@ -1641,6 +1696,17 @@ var LLMService = class {
     const provider = await this._getProvider(tenantId);
     return provider.imageGeneration(prompt, systemPrompt, options);
   }
+  /**
+   * Text-to-Speech synthesis
+   * @param {string} text - Text to synthesize
+   * @param {string} tenantId - Tenant identifier
+   * @param {Object} options - TTS options (voice, model, etc.)
+   * @returns {Promise<Buffer>} Audio data as Buffer
+   */
+  async textToSpeech(text, tenantId, options = {}) {
+    const provider = await this._getProvider(tenantId);
+    return provider.textToSpeech(text, options);
+  }
   /**
    * Start a Deep Research task
    * @param {string} prompt - Research prompt/query