openreader/src/contexts/TTSContext.tsx
2025-01-26 16:49:05 -07:00

497 lines
No EOL
14 KiB
TypeScript

'use client';
import React, {
createContext,
useContext,
useState,
useCallback,
useEffect,
useRef,
} from 'react';
import OpenAI from 'openai';
import { LRUCache } from 'lru-cache'; // Import LRUCache directly
import { Howl } from 'howler';
import { useConfig } from '@/contexts/ConfigContext';
import { splitIntoSentences, preprocessSentenceForAudio } from '@/services/nlp';
import { audioBufferToURL } from '@/services/audio';
// Media globals
declare global {
interface Window {
webkitAudioContext: typeof AudioContext;
}
}
type AudioContextType = typeof window extends undefined
? never
: (AudioContext);
interface TTSContextType {
isPlaying: boolean;
currentText: string;
togglePlay: () => void;
skipForward: () => void;
skipBackward: () => void;
setText: (text: string) => void;
currentSentence: string;
audioQueue: AudioBuffer[];
stop: () => void;
stopAndPlayFromIndex: (index: number) => void;
sentences: string[];
isProcessing: boolean;
setIsProcessing: (value: boolean) => void;
setIsPlaying: (value: boolean) => void;
speed: number;
setSpeed: (speed: number) => void;
setSpeedAndRestart: (speed: number) => void;
voice: string;
setVoice: (voice: string) => void;
setVoiceAndRestart: (voice: string) => void;
availableVoices: string[];
currentIndex: number;
setCurrentIndex: (index: number) => void;
currDocPage: number;
currDocPages: number | undefined;
setCurrDocPages: (num: number | undefined) => void;
incrementPage: (num?: number) => void;
skipToPage: (page: number) => void;
}
const TTSContext = createContext<TTSContextType | undefined>(undefined);
export function TTSProvider({ children }: { children: React.ReactNode }) {
const { apiKey: openApiKey, baseUrl: openApiBaseUrl, isLoading: configIsLoading } = useConfig();
// Move openai initialization to a ref to avoid breaking hooks rules
const openaiRef = useRef<OpenAI | null>(null);
// All existing state declarations and refs stay at the top
const [isPlaying, setIsPlaying] = useState(false);
const [currentText, setCurrentText] = useState('');
const [sentences, setSentences] = useState<string[]>([]);
const [currentIndex, setCurrentIndex] = useState(0);
const [audioContext, setAudioContext] = useState<AudioContextType>();
const [activeHowl, setActiveHowl] = useState<Howl | null>(null);
const [audioQueue] = useState<AudioBuffer[]>([]);
const [isProcessing, setIsProcessing] = useState(false);
const [speed, setSpeed] = useState(1);
const [voice, setVoice] = useState('alloy');
const [availableVoices, setAvailableVoices] = useState<string[]>([]);
const [currDocPage, setCurrDocPage] = useState<number>(1);
const [currDocPages, setCurrDocPages] = useState<number>();
const [nextPageLoading, setNextPageLoading] = useState(false);
// Audio cache using LRUCache with a maximum size of 50 entries
const audioCacheRef = useRef(new LRUCache<string, AudioBuffer>({ max: 50 }));
const setText = useCallback((text: string) => {
setCurrentText(text);
console.log('Setting page text:', text);
const newSentences = splitIntoSentences(text);
setSentences(newSentences);
// Clear audio cache
//audioCacheRef.current.clear();
setNextPageLoading(false);
}, []);
const abortAudio = useCallback(() => {
if (activeHowl) {
activeHowl.stop();
setActiveHowl(null);
}
}, [activeHowl]);
const togglePlay = useCallback(() => {
setIsPlaying((prev) => {
if (!prev) {
return true;
} else {
abortAudio();
return false;
}
});
}, [abortAudio]);
const skipToPage = useCallback((page: number) => {
abortAudio();
setIsPlaying(false);
setNextPageLoading(true);
setCurrentIndex(0);
setCurrDocPage(page);
}, [abortAudio]);
const incrementPage = useCallback((num = 1) => {
setNextPageLoading(true);
setCurrDocPage(currDocPage + num);
}, [currDocPage]);
const advance = useCallback(async (backwards = false) => {
setCurrentIndex((prev) => {
const nextIndex = prev + (backwards ? -1 : 1);
if (nextIndex < sentences.length && nextIndex >= 0) {
console.log('Advancing to next sentence:', sentences[nextIndex]);
return nextIndex;
} else if (nextIndex >= sentences.length && currDocPage < currDocPages!) {
console.log('Advancing to next page:', currDocPage + 1);
incrementPage();
return 0;
} else if (nextIndex < 0 && currDocPage > 1) {
console.log('Advancing to previous page:', currDocPage - 1);
incrementPage(-1);
return 0;
} else if (nextIndex >= sentences.length && currDocPage >= currDocPages!) {
console.log('Reached end of document');
setIsPlaying(false);
return prev;
}
return prev;
});
}, [sentences, currDocPage, currDocPages, incrementPage]);
const skipForward = useCallback(() => {
setIsProcessing(true);
abortAudio();
advance();
setIsProcessing(false);
}, [abortAudio, advance]);
const skipBackward = useCallback(() => {
setIsProcessing(true);
abortAudio();
advance(true); // Pass true to go backwards
setIsProcessing(false);
}, [abortAudio, advance]);
// Initialize OpenAI instance when config loads
useEffect(() => {
const fetchVoices = async () => {
try {
const response = await fetch(`${openApiBaseUrl}/audio/voices`, {
headers: {
'Authorization': `Bearer ${openApiKey}`,
'Content-Type': 'application/json',
},
});
if (!response.ok) throw new Error('Failed to fetch voices');
const data = await response.json();
setAvailableVoices(data.voices || []);
} catch (error) {
console.error('Error fetching voices:', error);
// Set available voices to default openai voices
// Supported voices are alloy, ash, coral, echo, fable, onyx, nova, sage and shimmer
setAvailableVoices(['alloy', 'ash', 'coral', 'echo', 'fable', 'onyx', 'nova', 'sage', 'shimmer']);
}
};
if (!configIsLoading && openApiKey && openApiBaseUrl) {
openaiRef.current = new OpenAI({
apiKey: openApiKey,
baseURL: openApiBaseUrl,
dangerouslyAllowBrowser: true,
});
fetchVoices();
}
}, [configIsLoading, openApiKey, openApiBaseUrl]);
useEffect(() => {
/*
* Initializes the AudioContext for text-to-speech playback.
* Creates a new AudioContext instance if one doesn't exist.
* Only runs on the client side to avoid SSR issues.
*
* Dependencies:
* - audioContext: Re-runs if the audioContext is null or changes
*/
if (typeof window !== 'undefined' && !audioContext) {
const AudioContextClass = window.AudioContext || window.webkitAudioContext;
if (AudioContextClass) {
try {
setAudioContext(new AudioContextClass());
} catch (error) {
console.error('Failed to initialize AudioContext:', error);
}
}
}
return () => {
if (audioContext) {
audioContext.close().catch((error) => {
console.error('Error closing AudioContext:', error);
});
}
}
}, [audioContext]);
// Now the MediaSession effect can use these functions
useEffect(() => {
if ('mediaSession' in navigator) {
navigator.mediaSession.metadata = new MediaMetadata({
title: 'Text-to-Speech',
artist: 'OpenReader WebUI',
album: 'Current Document',
});
navigator.mediaSession.setActionHandler('play', () => togglePlay());
navigator.mediaSession.setActionHandler('pause', () => togglePlay());
navigator.mediaSession.setActionHandler('nexttrack', () => skipForward());
navigator.mediaSession.setActionHandler('previoustrack', () => skipBackward());
}
}, [togglePlay, skipForward, skipBackward]);
//new function to return audio buffer with caching
const getAudio = useCallback(async (sentence: string): Promise<AudioBuffer | undefined> => {
// Check if the audio is already cached
const cachedAudio = audioCacheRef.current.get(sentence);
if (cachedAudio) {
console.log('Using cached audio for sentence:', sentence.substring(0, 20));
return cachedAudio;
}
// If not cached, fetch the audio from OpenAI API
if (openaiRef.current) {
console.log('Requesting audio for sentence:', sentence);
const response = await openaiRef.current.audio.speech.create({
model: 'tts-1',
voice: voice as "alloy",
input: sentence,
speed: speed,
});
const arrayBuffer = await response.arrayBuffer();
const audioBuffer = await audioContext!.decodeAudioData(arrayBuffer);
// Cache the audio buffer
audioCacheRef.current.set(sentence, audioBuffer);
return audioBuffer;
}
}, [audioContext, voice, speed]);
const processSentence = useCallback(async (sentence: string, preload = false): Promise<string> => {
if (isProcessing && !preload) throw new Error('Audio is already being processed');
if (!audioContext || !openaiRef.current) throw new Error('Audio context not initialized');
// Only set processing state if not preloading
if (!preload) setIsProcessing(true);
const cleanedSentence = preprocessSentenceForAudio(sentence);
const audioBuffer = await getAudio(cleanedSentence);
return audioBufferToURL(audioBuffer!);
}, [isProcessing, audioContext, getAudio]);
const playSentenceWithHowl = useCallback(async (sentence: string) => {
try {
const audioUrl = await processSentence(sentence);
if (!audioUrl) {
throw new Error('No audio URL generated');
}
const howl = new Howl({
src: [audioUrl],
format: ['wav'],
html5: true,
onplay: () => {
if ('mediaSession' in navigator) {
navigator.mediaSession.playbackState = 'playing';
}
},
onpause: () => {
if ('mediaSession' in navigator) {
navigator.mediaSession.playbackState = 'paused';
}
},
onend: () => {
URL.revokeObjectURL(audioUrl);
setActiveHowl(null);
if (isPlaying) {
advance();
}
},
onloaderror: (id, error) => {
console.error('Error loading audio:', error);
setIsProcessing(false);
setActiveHowl(null);
URL.revokeObjectURL(audioUrl);
// Don't auto-advance on load error
setIsPlaying(false);
},
});
setActiveHowl(howl);
howl.play();
setIsProcessing(false);
} catch (error) {
console.error('Error playing TTS:', error);
setActiveHowl(null);
setIsProcessing(false);
//setIsPlaying(false); // Stop playback on error
if (error instanceof Error &&
(error.message.includes('Unable to decode audio data') ||
error.message.includes('EncodingError'))) {
console.log('Skipping problematic sentence:', sentence);
advance(); // Skip problematic sentence
}
}
}, [isPlaying, processSentence, advance]);
const preloadNextAudio = useCallback(() => {
try {
if (sentences[currentIndex + 1] && !audioCacheRef.current.has(sentences[currentIndex + 1])) {
processSentence(sentences[currentIndex + 1], true); // True indicates preloading
}
} catch (error) {
console.error('Error preloading next sentence:', error);
}
}, [currentIndex, sentences, audioCacheRef, processSentence]);
const playAudio = useCallback(async () => {
await playSentenceWithHowl(sentences[currentIndex]);
}, [sentences, currentIndex, playSentenceWithHowl]);
// main driver useEffect
useEffect(() => {
if (!isPlaying) return; // Don't proceed if stopped
if (isProcessing) return; // Don't proceed if processing audio
if (!sentences[currentIndex]) return; // Don't proceed if no sentence to play
if (nextPageLoading) return; // Don't proceed if loading next page
if (activeHowl) return; // Don't proceed if audio is already playing
// Play the current sentence and preload the next one if available
playAudio();
if (sentences[currentIndex + 1]) {
preloadNextAudio();
}
return () => {
abortAudio();
};
}, [
isPlaying,
isProcessing,
currentIndex,
sentences,
activeHowl,
nextPageLoading,
playAudio,
preloadNextAudio,
abortAudio
]);
const stop = useCallback(() => {
// Cancel any ongoing request
abortAudio();
setIsPlaying(false);
setCurrentIndex(0);
setCurrentText('');
setIsProcessing(false);
}, [abortAudio]);
const stopAndPlayFromIndex = useCallback((index: number) => {
abortAudio();
setCurrentIndex(index);
setIsPlaying(true);
}, [abortAudio]);
const setCurrentIndexWithoutPlay = useCallback((index: number) => {
abortAudio();
setCurrentIndex(index);
}, [abortAudio]);
const setSpeedAndRestart = useCallback((newSpeed: number) => {
setSpeed(newSpeed);
// Clear the audio cache since it contains audio at the old speed
audioCacheRef.current.clear();
if (isPlaying) {
const currentIdx = currentIndex;
stop();
setCurrentIndex(currentIdx);
setIsPlaying(true);
}
}, [isPlaying, currentIndex, stop]);
const setVoiceAndRestart = useCallback((newVoice: string) => {
setVoice(newVoice);
// Clear the audio cache since it contains audio with the old voice
audioCacheRef.current.clear();
if (isPlaying) {
const currentIdx = currentIndex;
stop();
setCurrentIndex(currentIdx);
setIsPlaying(true);
}
}, [isPlaying, currentIndex, stop]);
const value = {
isPlaying,
currentText,
togglePlay,
skipForward,
skipBackward,
setText,
currentSentence: sentences[currentIndex] || '',
audioQueue,
stop,
setCurrentIndex: setCurrentIndexWithoutPlay,
stopAndPlayFromIndex,
sentences,
isProcessing,
setIsProcessing,
setIsPlaying,
speed,
setSpeed,
setSpeedAndRestart,
voice,
setVoice,
setVoiceAndRestart,
availableVoices,
currentIndex,
currDocPage,
currDocPages,
setCurrDocPages,
incrementPage,
skipToPage,
};
if (configIsLoading) {
return null;
}
return (
<TTSContext.Provider value={value}>
{children}
</TTSContext.Provider>
);
}
export function useTTS() {
const context = useContext(TTSContext);
if (context === undefined) {
throw new Error('useTTS must be used within a TTSProvider');
}
return context;
}