ai.interactive.fiction/public/js/openai-tts-module.js

/**
 * OpenAITTSModule
 * Provides TTS via OpenAI API
 */
import { ApiTTSModuleBase } from './api-tts-module-base.js';

export class OpenAITTSModule extends ApiTTSModuleBase {
    constructor() {
        super('openai', 'OpenAI TTS');
        
        // Voice options specific to OpenAI
        this.voiceOptions = {
            voice: 'alloy', // Default voice for OpenAI
            model: 'tts-1', // Standard model
            speed: 1.0,
            response_format: 'mp3' // OpenAI supports mp3, opus, aac, and flac (not wav)
        };
        
        // Predefined voices - OpenAI has a fixed set
        this.voices = [
            { id: 'alloy', name: 'Alloy', language: 'en' },
            { id: 'echo', name: 'Echo', language: 'en' },
            { id: 'fable', name: 'Fable', language: 'en' },
            { id: 'onyx', name: 'Onyx', language: 'en' },
            { id: 'nova', name: 'Nova', language: 'en' },
            { id: 'shimmer', name: 'Shimmer', language: 'en' }
        ];
    }
    
    /**
     * Get the default API base URL for OpenAI
     * @returns {string} - Default API base URL
     */
    getDefaultApiBaseUrl() {
        return 'https://api.openai.com/v1';
    }
    
    /**
     * Initialize the module
     * @returns {Promise<boolean>} - Resolves with success status
     */
    async initialize() {
        try {
            this.reportProgress(10, 'Initializing OpenAI TTS');
            
            // Initialize parent
            const parentInit = await super.initialize();
            if (!parentInit) {
                console.error('OpenAI TTS: Parent initialization failed');
                return false;
            }
            
            // Get required dependencies
            const persistenceManager = this.getModule('persistence-manager');
            if (!persistenceManager) {
                console.error('OpenAI TTS: Required dependency persistence-manager not found');
                return false;
            }
            
            // Check for API key
            const apiKey = persistenceManager.getPreference('openai', 'api_key', '');
            if (!apiKey) {
                console.error('OpenAI TTS: API key not configured');
                return false;
            }
            
            // Set API key
            this.apiKey = apiKey;
            
            // Load preferences
            const preferredVoice = persistenceManager.getPreference('openai', 'voice', this.voiceOptions.voice);
            if (preferredVoice) {
                this.voiceOptions.voice = preferredVoice;
            }
            
            const preferredModel = persistenceManager.getPreference('openai', 'model', this.voiceOptions.model);
            if (preferredModel) {
                this.voiceOptions.model = preferredModel;
            }
            
            const preferredSpeed = persistenceManager.getPreference('openai', 'speed', this.voiceOptions.speed);
            if (typeof preferredSpeed === 'number') {
                this.voiceOptions.speed = preferredSpeed;
            }
            
            // Setup available voices
            this.voices = this.getAvailableVoices();
            
            this.isReady = true;
            this.reportProgress(100, 'OpenAI TTS initialized');
            return true;
        } catch (error) {
            console.error('OpenAI TTS: Initialization error:', error);
            this.isReady = false;
            return false;
        }
    }
    
    /**
     * Load available voices
     * @returns {Promise<boolean>} - Resolves with success status
     */
    async loadVoices() {
        // OpenAI has a fixed set of voices, no need to fetch them
        return true;
    }
    
    /**
     * Select a voice for the given locale
     * @param {string} locale - Locale code
     * @returns {boolean} - Success status
     */
    selectVoiceForLocale(locale) {
        // Extract language code from locale (e.g., 'en-US' -> 'en')
        const langCode = locale.split('-')[0].toLowerCase();
        
        // All OpenAI voices are English-based
        // For English locales, we could customize the voice selection
        // For non-English locales, we'll just use the default
        
        // In this simple implementation, we'll just use the default voice
        return this.selectDefaultVoice();
    }
    
    /**
     * Select a default voice
     * @returns {boolean} - Success status
     */
    selectDefaultVoice() {
        this.voiceOptions.voice = 'alloy';
        return true;
    }
    
    /**
     * Generate speech audio data using OpenAI API
     * @param {string} text - Text to generate speech for
     * @returns {Promise<Object>} - Audio data object
     */
    async generateSpeechAudio(text) {
        if (!text || !this.apiKey) {
            return { 
                success: false, 
                reason: 'missing_api_key_or_text' 
            };
        }
        
        try {
            // Process the text
            const processedText = this.preprocessText(text);
            
            // Create request payload
            const payload = {
                model: this.voiceOptions.model || 'tts-1',
                input: processedText,
                voice: this.voiceOptions.voice || 'alloy',
                response_format: this.voiceOptions.response_format || 'mp3',
                speed: this.voiceOptions.speed || 1.0
            };
            
            // Make API request
            const response = await fetch(`${this.apiBaseUrl}/audio/speech`, {
                method: 'POST',
                headers: {
                    'Content-Type': 'application/json',
                    'Authorization': `Bearer ${this.apiKey}`
                },
                body: JSON.stringify(payload)
            });
            
            if (!response.ok) {
                const errorText = await response.text();
                throw new Error(`API error: ${response.status} ${response.statusText} - ${errorText}`);
            }
            
            // Get audio blob from response
            const audioBlob = await response.blob();
            
            // Convert to array buffer for consistency with other modules
            const arrayBuffer = await audioBlob.arrayBuffer();
            
            return {
                success: true,
                audioData: arrayBuffer
            };
        } catch (error) {
            console.error('OpenAI TTS: Error generating speech:', error);
            return { 
                success: false, 
                reason: 'api_error', 
                error: error.message 
            };
        }
    }
    
    /**
     * Set voice options
     * @param {Object} options - Voice options
     */
    setVoiceOptions(options = {}) {
        // Handle common options
        if (options.voice) {
            this.voiceOptions.voice = options.voice;
            
            // Save voice preference
            const persistenceManager = this.getModule('persistence-manager');
            if (persistenceManager) {
                persistenceManager.updatePreference('tts', 'openai_voice', options.voice);
            }
        }
        
        if (typeof options.speed === 'number') {
            this.voiceOptions.speed = Math.max(0.5, Math.min(2.0, options.speed));
        }
        
        // Handle OpenAI-specific options
        if (options.model) {
            this.voiceOptions.model = options.model;
            
            // Save the model preference
            const persistenceManager = this.getModule('persistence-manager');
            if (persistenceManager) {
                persistenceManager.updatePreference('tts', 'openai_model', options.model);
            }
        }
        
        if (options.response_format) {
            // Ensure valid format: mp3, opus, aac, or flac
            const validFormats = ['mp3', 'opus', 'aac', 'flac'];
            if (validFormats.includes(options.response_format)) {
                this.voiceOptions.response_format = options.response_format;
                
                // Save the format preference
                const persistenceManager = this.getModule('persistence-manager');
                if (persistenceManager) {
                    persistenceManager.updatePreference('tts', 'openai_format', options.response_format);
                }
            }
        }
    }
}

// Register the module with the module registry
// Module registry MUST be accessed via window, not direct import
if (window.moduleRegistry) {
    try {
        // Create instance first, then register it
        const openAITTSModule = new OpenAITTSModule();
        window.moduleRegistry.register(openAITTSModule);
        console.log('OpenAI TTS Module registered successfully');
    } catch (err) {
        console.error('Failed to register OpenAI TTS Module:', err);
    }
} else {
    console.error('Module registry not available when attempting to register OpenAI TTS Module');
}