Spaces:
Running
Running
// script1.js | |
import { client, handle_file } from 'https://cdn.jsdelivr.net/npm/@gradio/client/+esm'; | |
const video = document.getElementById('webcam'); | |
let gradioApp; // Declare gradioApp outside the function | |
const GRADIO_CLIENTS = [ | |
"multimodalart/Florence-2-l4", | |
"gokaygokay/Florence-2", | |
"multimodalart/Florence-2-l4-2", | |
"gokaygokay/Florence-2", | |
]; | |
async function startWebcam() { | |
try { | |
const stream = await navigator.mediaDevices.getUserMedia({ video: true }); | |
video.srcObject = stream; | |
} catch (error) { | |
console.error("Error accessing webcam:", error); | |
} | |
} | |
async function getCaption() { | |
if (!gradioApp) { | |
try { | |
const randomClient = GRADIO_CLIENTS[Math.floor(Math.random() * GRADIO_CLIENTS.length)]; | |
gradioApp = await client(randomClient); | |
} catch (error) { | |
console.error("Error loading Gradio client:", error); | |
return "Error getting caption"; // Or some other default | |
} | |
} | |
try { | |
const canvas = document.createElement('canvas'); | |
canvas.width = video.videoWidth; | |
canvas.height = video.videoHeight; | |
const context = canvas.getContext('2d'); | |
context.drawImage(video, 0, 0, canvas.width, canvas.height); | |
const blob = await new Promise(resolve => canvas.toBlob(resolve, 'image/png')); | |
const handledFile = await handle_file(blob); | |
const result = await gradioApp.predict("/process_image", [handledFile, "More Detailed Caption"]); | |
return result.data[0]; | |
} catch (error) { | |
console.error("Error getting caption:", error); | |
return "Error getting caption"; // Or handle the error differently | |
} | |
} | |
// Constants and Configuration | |
const USER_SPEECH_INTERRUPT_DELAY = 500; | |
const TEXT_TO_SPEECH_API_ENDPOINT = "https://api.streamelements.com/kappa/v2/speech"; | |
const CHUNK_SIZE = 300; | |
const MAX_PREFETCH_REQUESTS = 10; | |
const PREFETCH_CACHE_EXPIRATION = 60000; // 1 minute | |
const AUDIO_CACHE_EXPIRATION = 3600000; // 1 hour | |
// DOM Elements | |
const startStopButton = document.getElementById('startStopButton'); | |
const voiceSelectionDropdown = document.getElementById('voiceSelect'); | |
const modelSelectionDropdown = document.getElementById('modelSelect'); | |
const noiseSuppressionCheckbox = document.getElementById('noiseSuppression'); | |
const responseTimeDisplay = document.getElementById('responseTime'); | |
const userActivityIndicator = document.getElementById('userIndicator'); | |
const aiActivityIndicator = document.getElementById('aiIndicator'); | |
const transcriptDiv = document.getElementById('transcript'); | |
// Speech Recognition | |
let speechRecognizer; | |
let isSpeechRecognitionActive = false; | |
// AI Interaction State | |
let activeQuery = null; | |
let queryStartTime = 0; | |
let isRequestInProgress = false; | |
let isUserSpeaking = false; | |
let requestAbortController = null; | |
let firstResponseTextTimestamp = null; | |
// Audio Management | |
let currentAudio = null; | |
let audioPlaybackQueue = []; | |
// Prefetching and Caching | |
const prefetchCache = new Map(); | |
const pendingPrefetchRequests = new Map(); | |
const prefetchQueue = []; | |
let prefetchTextQuery = ""; | |
// Conversation History | |
let conversationHistory = []; | |
// Audio Caching | |
const audioCache = new Map(); | |
// Utility Functions | |
// Normalize query text | |
const normalizeQueryText = query => query.trim().toLowerCase().replace(/[^\w\s]/g, ''); | |
// Generate a cache key | |
const generateCacheKey = (normalizedQuery, voice, history, modelName) => | |
`${normalizedQuery}-${voice}-${JSON.stringify(history)}-${modelName}`; | |
// Update activity indicators | |
const updateActivityIndicators = (state = null) => { | |
userActivityIndicator.textContent = isUserSpeaking ? "User: Speaking" : "User: Idle"; | |
if (isRequestInProgress && !currentAudio) { | |
aiActivityIndicator.textContent = "AI: Processing..."; | |
} else if (currentAudio && !isUserSpeaking) { | |
aiActivityIndicator.textContent = state || "AI: Speaking"; | |
} else if (isUserSpeaking) { | |
aiActivityIndicator.textContent = "AI: Listening"; | |
} else { | |
aiActivityIndicator.textContent = "AI: Idle"; | |
} | |
}; | |
// Update latency display | |
const updateLatency = () => { | |
if (firstResponseTextTimestamp) { | |
const latency = firstResponseTextTimestamp - queryStartTime; | |
responseTimeDisplay.textContent = `Latency: ${latency}ms`; | |
} else { | |
responseTimeDisplay.textContent = "Latency: 0ms"; | |
} | |
}; | |
// Add to conversation history | |
const addToConversationHistory = (role, content) => { | |
if (conversationHistory.length > 0 && | |
conversationHistory[conversationHistory.length - 1].role === 'assistant' && | |
conversationHistory[conversationHistory.length - 1].content === "") { | |
conversationHistory.pop(); | |
} | |
conversationHistory.push({ role, content }); | |
if (conversationHistory.length > 6) conversationHistory.splice(0, 2); | |
}; | |
// Check if audio playback should be interrupted | |
const shouldInterruptAudioPlayback = (interimTranscript) => | |
Date.now() - lastUserSpeechTimestamp > USER_SPEECH_INTERRUPT_DELAY || interimTranscript.length > 5; | |
// Audio Management Functions | |
// Play audio from the queue | |
const playNextAudio = async () => { | |
if (audioPlaybackQueue.length > 0) { | |
const audioData = audioPlaybackQueue.shift(); | |
const audio = new Audio(audioData.url); | |
updateActivityIndicators(); | |
const audioPromise = new Promise(resolve => { | |
audio.onended = resolve; | |
audio.onerror = resolve; | |
}); | |
if (currentAudio) { | |
currentAudio.pause(); | |
currentAudio.currentTime = 0; | |
} | |
currentAudio = audio; | |
await audio.play(); | |
await audioPromise; | |
playNextAudio(); | |
} else { | |
updateActivityIndicators(); | |
} | |
}; | |
// Interrupt audio playback | |
const interruptAudioPlayback = (reason = 'unknown') => { | |
console.log(`Interrupting audio (reason: ${reason})...`); | |
if (currentAudio) { | |
currentAudio.pause(); | |
currentAudio.currentTime = 0; | |
currentAudio = null; | |
} | |
audioPlaybackQueue.length = 0; | |
isRequestInProgress = false; | |
if (requestAbortController) { | |
requestAbortController.abort(); | |
requestAbortController = null; | |
} | |
prefetchCache.clear(); | |
prefetchQueue.length = 0; | |
updateActivityIndicators(); | |
}; | |
// Prefetching and Caching Functions | |
// Prefetch and cache the first TTS audio chunk | |
const prefetchFirstAudioChunk = (query, voice) => { | |
const normalizedQuery = normalizeQueryText(query); | |
const cacheKey = generateCacheKey(normalizedQuery, voice, conversationHistory, modelSelectionDropdown.value); | |
if (pendingPrefetchRequests.has(cacheKey) || prefetchCache.has(cacheKey)) return; | |
prefetchQueue.push({ query: query.trim(), voice, cacheKey }); | |
processPrefetchQueue(); | |
}; | |
// Process the prefetch queue | |
const processPrefetchQueue = async () => { | |
while (prefetchQueue.length > 0 && pendingPrefetchRequests.size < MAX_PREFETCH_REQUESTS) { | |
const { query, voice, cacheKey } = prefetchQueue.shift(); | |
const abortController = new AbortController(); | |
pendingPrefetchRequests.set(cacheKey, abortController); | |
try { | |
const firstAudioUrl = await streamAndPrefetchAudio(query, voice, abortController.signal); | |
if (firstAudioUrl) prefetchCache.set(cacheKey, { url: firstAudioUrl, timestamp: Date.now() }); | |
} catch (error) { | |
if (error.name !== 'AbortError') console.error("Error prefetching audio:", error); | |
} finally { | |
pendingPrefetchRequests.delete(cacheKey); | |
processPrefetchQueue(); | |
} | |
} | |
}; | |
// Cancel pending prefetch requests | |
const cancelPrefetchRequests = (query) => { | |
const normalizedQuery = normalizeQueryText(query); | |
for (const [cacheKey, abortController] of pendingPrefetchRequests) { | |
if (cacheKey.startsWith(normalizedQuery)) { | |
abortController.abort(); | |
pendingPrefetchRequests.delete(cacheKey); | |
} | |
} | |
}; | |
// AI Interaction Functions | |
// Modify sendQueryToAI to include the caption | |
async function sendQueryToAI(query) { | |
const caption = await getCaption(); | |
const modifiedQuery = JSON.stringify({ USER: query, CAPTION: caption }); | |
console.log("Sending query to AI:", modifiedQuery); | |
isRequestInProgress = true; | |
updateActivityIndicators(); | |
firstResponseTextTimestamp = null; | |
const normalizedQuery = normalizeQueryText(query); | |
const cacheKey = generateCacheKey(normalizedQuery, modelSelectionDropdown.value, conversationHistory, modelSelectionDropdown.value); | |
queryStartTime = Date.now(); | |
// Check prefetch cache | |
if (prefetchCache.has(cacheKey)) { | |
const cachedData = prefetchCache.get(cacheKey); | |
if (Date.now() - cachedData.timestamp < PREFETCH_CACHE_EXPIRATION) { | |
audioPlaybackQueue.push({ url: cachedData.url, isPrefetched: true }); | |
playNextAudio(); | |
} else { | |
prefetchCache.delete(cacheKey); | |
} | |
} | |
requestAbortController = new AbortController(); | |
try { | |
await streamAndHandleAudioResponse(modifiedQuery, voiceSelectionDropdown.value, requestAbortController.signal); | |
} catch (error) { | |
if (error.name !== 'AbortError') { | |
console.error("Error sending query to AI:", error); | |
} | |
} finally { | |
isRequestInProgress = false; | |
updateActivityIndicators(); | |
} | |
}; | |
// Process the final speech transcript | |
const processSpeechTranscript = (transcript) => { | |
const trimmedTranscript = transcript.trimStart(); | |
if (trimmedTranscript !== '' && !isRequestInProgress) { | |
activeQuery = trimmedTranscript; | |
sendQueryToAI(activeQuery); | |
addToConversationHistory('user', activeQuery); | |
} | |
}; | |
// Network and Streaming Functions | |
// Stream AI response and handle audio | |
const streamAndHandleAudioResponse = async (query, voice, abortSignal) => { | |
const response = await fetchAIResponse(query, abortSignal); | |
if (!response.ok) { | |
if (response.status === 429) { | |
console.log("Rate limit hit, retrying in 1 second..."); | |
await new Promise(resolve => setTimeout(resolve, 1000)); | |
await sendQueryToAI(query); | |
return; | |
} | |
throw new Error(`Network response was not ok: ${response.status}`); | |
} | |
console.log("Streaming audio response received"); | |
await handleStreamingResponse(response.body, voice, abortSignal); | |
}; | |
// Stream AI response for prefetching | |
const streamAndPrefetchAudio = async (query, voice, abortSignal) => { | |
const response = await fetchAIResponse(query, abortSignal); | |
if (!response.ok) throw new Error('Network response was not ok'); | |
return handleStreamingResponseForPrefetch(response.body, voice, abortSignal); | |
}; | |
// Fetch AI response | |
const fetchAIResponse = async (query, abortSignal) => { | |
const userSambanovaKey = document.getElementById('apiKey').value.trim() !== '' ? document.getElementById('apiKey').value.trim() : 'none'; | |
const url = '/stream_text'; | |
const requestBody = { | |
query: query, | |
history: JSON.stringify(conversationHistory), | |
model: modelSelectionDropdown.value, | |
api_key: userSambanovaKey | |
}; | |
return fetch(url, { | |
method: 'POST', | |
headers: { | |
'Accept': 'text/event-stream', | |
'Content-Type': 'application/json' | |
}, | |
body: JSON.stringify(requestBody), | |
signal: abortSignal | |
}); | |
}; | |
// Handle the streaming response for prefetching | |
const handleStreamingResponseForPrefetch = async (responseStream, voice, abortSignal) => { | |
const reader = responseStream.getReader(); | |
const decoder = new TextDecoder("utf-8"); | |
let buffer = ""; | |
try { | |
while (true) { | |
const { done, value } = await reader.read(); | |
if (done) break; | |
if (abortSignal.aborted) throw new DOMException('Request aborted', 'AbortError'); | |
const chunk = decoder.decode(value, { stream: true }); | |
buffer += chunk; | |
const lines = buffer.split('\n'); | |
for (let i = 0; i < lines.length - 1; i++) { | |
const line = lines[i]; | |
if (line.startsWith('data: ')) { | |
const textContent = line.substring(6).trim(); | |
if (textContent) { | |
return await generateTextToSpeechAudio(textContent, voice); | |
} | |
} | |
} | |
buffer = lines[lines.length - 1]; | |
} | |
} catch (error) { | |
console.error("Error in handleStreamingResponseForPrefetch:", error); | |
} finally { | |
reader.releaseLock(); | |
} | |
return null; | |
}; | |
// Handle the streaming audio response | |
const handleStreamingResponse = async (responseStream, voice, abortSignal) => { | |
const reader = responseStream.getReader(); | |
const decoder = new TextDecoder("utf-8"); | |
let buffer = ""; | |
let initialChunksSent = 0; | |
let fullResponseText = ""; | |
let fullResponseText2 = ""; | |
let textChunk = ""; | |
let sentText = ""; | |
try { | |
while (true) { | |
const { done, value } = await reader.read(); | |
if (done) break; | |
if (abortSignal.aborted) throw new DOMException('Request aborted', 'AbortError'); | |
if (isUserSpeaking) { | |
interruptAudioPlayback('user is speaking'); | |
break; | |
} | |
const chunk = decoder.decode(value, { stream: true }); | |
buffer += chunk; | |
const lines = buffer.split('\n'); | |
for (let i = 0; i < lines.length - 1; i++) { | |
const line = lines[i]; | |
if (line.startsWith('data: ')) { | |
const textContent = line.substring(6).trim(); | |
if (textContent) { | |
if (!firstResponseTextTimestamp) firstResponseTextTimestamp = Date.now(); | |
fullResponseText += textContent + " "; | |
fullResponseText2 += textContent + " "; | |
textChunk += textContent + " "; | |
transcriptDiv.textContent = fullResponseText2; | |
if (initialChunksSent < 2) { | |
const audioUrl = await generateTextToSpeechAudio(textContent, voice); | |
if (audioUrl) { | |
audioPlaybackQueue.push({ url: audioUrl, isPrefetched: false }); | |
if (!currentAudio) playNextAudio(); | |
} | |
sentText += textContent + " "; | |
initialChunksSent++; | |
} else { | |
let unsentTextChunk = textChunk.replace(sentText, '').trim(); | |
if (unsentTextChunk.length >= CHUNK_SIZE) { | |
const audioUrl = await generateTextToSpeechAudio(unsentTextChunk, voice); | |
if (audioUrl) { | |
audioPlaybackQueue.push({ url: audioUrl, isPrefetched: false }); | |
if (!currentAudio) playNextAudio(); | |
} | |
textChunk = ""; | |
} | |
} | |
if (fullResponseText !== '') { | |
fullResponseText = ''; | |
} | |
} | |
} | |
} | |
buffer = lines[lines.length - 1]; | |
} | |
} catch (error) { | |
console.error("Error in handleStreamingResponse:", error); | |
} finally { | |
reader.releaseLock(); | |
let unsentTextChunk = textChunk.replace(sentText, '').trim(); | |
if (unsentTextChunk !== "") { | |
const audioUrl = await generateTextToSpeechAudio(unsentTextChunk, voice); | |
if (audioUrl) { | |
audioPlaybackQueue.push({ url: audioUrl, isPrefetched: false }); | |
if (!currentAudio) playNextAudio(); | |
} | |
} | |
if (fullResponseText !== '') { | |
fullResponseText = ''; | |
} | |
if (fullResponseText2 !== '') { | |
addToConversationHistory('assistant', fullResponseText2); | |
fullResponseText2 = ''; | |
} | |
} | |
}; | |
// Generate Text-to-Speech audio with caching | |
const generateTextToSpeechAudio = async (text, voice) => { | |
const normalizedText = normalizeQueryText(text); | |
const cacheKey = `${normalizedText}-${voice}`; | |
if (audioCache.has(cacheKey)) { | |
const cachedData = audioCache.get(cacheKey); | |
if (Date.now() - cachedData.timestamp < AUDIO_CACHE_EXPIRATION) { | |
return cachedData.url; | |
} else { | |
audioCache.delete(cacheKey); | |
} | |
} | |
try { | |
const response = await fetch(`${TEXT_TO_SPEECH_API_ENDPOINT}?voice=${voice}&text=${encodeURIComponent(text)}`, { method: 'GET' }); | |
if (!response.ok) throw new Error('Network response was not ok'); | |
const audioBlob = await response.blob(); | |
const audioUrl = URL.createObjectURL(audioBlob); | |
audioCache.set(cacheKey, { url: audioUrl, timestamp: Date.now() }); | |
return audioUrl; | |
} catch (error) { | |
console.error("Error generating TTS audio:", error); | |
return null; | |
} | |
}; | |
// Speech Recognition Initialization | |
if ('webkitSpeechRecognition' in window) { | |
speechRecognizer = new webkitSpeechRecognition(); | |
Object.assign(speechRecognizer, { | |
continuous: true, | |
interimResults: true, | |
language: 'en-US', | |
maxAlternatives: 3 | |
}); | |
speechRecognizer.onstart = () => { | |
console.log("Speech recognition started"); | |
isUserSpeaking = true; | |
lastUserSpeechTimestamp = Date.now(); | |
updateActivityIndicators(); | |
startStopButton.innerHTML = '<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M9 9h6v6h-6z"></path><path d="M12 1a3 3 0 0 0-3 3v8a3 3 0 0 0 6 0V4a3 3 0 0 0-3-3z"></path><path d="M19 10v2a7 7 0 0 1-14 0v-2"></path><line x1="12" y1="19" x2="12" y2="23"></line><line x1="8" y1="23" x2="16" y2="23"></line></svg> Stop Listening'; | |
}; | |
speechRecognizer.onresult = (event) => { | |
let interimTranscript = ''; | |
for (let i = event.resultIndex; i < event.results.length; i++) { | |
const transcript = event.results[i][0].transcript; | |
if (event.results[i].isFinal) { | |
interruptAudioPlayback('final'); | |
processSpeechTranscript(transcript); | |
isUserSpeaking = false; | |
updateActivityIndicators(); | |
queryStartTime = Date.now(); | |
} else { | |
interimTranscript += transcript; | |
isUserSpeaking = true; | |
lastUserSpeechTimestamp = Date.now(); | |
updateActivityIndicators(); | |
if (interimTranscript.length > prefetchTextQuery.length + 5) { | |
cancelPrefetchRequests(prefetchTextQuery); | |
} | |
prefetchTextQuery = interimTranscript; | |
prefetchFirstAudioChunk(interimTranscript, voiceSelectionDropdown.value); | |
if (isRequestInProgress && shouldInterruptAudioPlayback(interimTranscript)) { | |
interruptAudioPlayback('interim'); | |
} | |
} | |
} | |
}; | |
speechRecognizer.onerror = (event) => { | |
console.error('Speech recognition error:', event.error); | |
if (isSpeechRecognitionActive) speechRecognizer.start(); | |
}; | |
speechRecognizer.onend = () => { | |
isUserSpeaking = false; | |
updateActivityIndicators(); | |
if (isSpeechRecognitionActive) speechRecognizer.start(); | |
}; | |
startStopButton.addEventListener('click', () => { | |
if (isSpeechRecognitionActive) { | |
speechRecognizer.stop(); | |
isSpeechRecognitionActive = false; | |
startStopButton.innerHTML = '<svg id="microphoneIcon" xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M12 1a3 3 0 0 0-3 3v8a3 3 0 0 0 6 0V4a3 3 0 0 0-3-3z"></path><path d="M19 10v2a7 7 0 0 1-14 0v-2"></path><line x1="12" y1="19" x2="12" y2="23"></line><line x1="8" y1="23" x2="16" y2="23"></line></svg> Start Listening'; | |
} else { | |
speechRecognizer.start(); | |
isSpeechRecognitionActive = true; | |
startStopButton.innerHTML = '<svg xmlns="http://www.w3.org/2000/svg" width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M9 9h6v6h-6z"></path><path d="M12 1a3 3 0 0 0-3 3v8a3 3 0 0 0 6 0V4a3 3 0 0 0-3-3z"></path><path d="M19 10v2a7 7 0 0 1-14 0v-2"></path><line x1="12" y1="19" x2="12" y2="23"></line><line x1="8" y1="23" x2="16" y2="23"></line></svg> Stop Listening'; | |
} | |
}); | |
} else { | |
alert('Your browser does not support the Web Speech API.'); | |
} | |
setInterval(updateLatency, 100); | |
window.onload = startWebcam; |