ea52f422a9
- Introduce a new user script `copilot-native-element.js` that implements a native-track bridge for audio and video input selection in WebRTC. - Implement functionality for managing remote audio and video tracks, including recording capabilities for AI input. - Enhance chat logging features to provide real-time feedback on AI interactions and recording status. - Integrate mechanisms for handling media streams, improving overall media management in the WebRTC environment.
1007 lines
43 KiB
JavaScript
1007 lines
43 KiB
JavaScript
// ==UserScript==
|
||
// @name 12345 AI Native Element Copilot
|
||
// @namespace http://tampermonkey.net/
|
||
// @version 1.0
|
||
// @description Native-track bridge with fixed element selection: element-2:audio input, element-0:video vision.
|
||
// @author You
|
||
// @match *://yhy-yw.22sj.cn:*/*
|
||
// @grant none
|
||
// @run-at document-start
|
||
// ==/UserScript==
|
||
|
||
(function () {
|
||
'use strict';
|
||
|
||
const DEFAULT_SIGNALING_URL = 'ws://182.92.86.220:8000/ws/voice';
|
||
const STATIC_AUDIO_ELEMENT_REASON = 'element-2:audio';
|
||
const STATIC_VIDEO_ELEMENT_REASON = 'element-0:video';
|
||
|
||
const originalAddEventListener = RTCPeerConnection.prototype.addEventListener;
|
||
const originalAddTrack = RTCPeerConnection.prototype.addTrack;
|
||
const originalAddTransceiver = RTCPeerConnection.prototype.addTransceiver;
|
||
const originalSetRemoteDescription = RTCPeerConnection.prototype.setRemoteDescription;
|
||
const mediaSrcObjectDescriptor = Object.getOwnPropertyDescriptor(HTMLMediaElement.prototype, 'srcObject');
|
||
|
||
const remoteAudio = new Map();
|
||
const remoteVideo = new Map();
|
||
const businessSenders = new Map();
|
||
const aiOutputTrackIds = new Set();
|
||
const inboundMemory = new Map();
|
||
const outboundMemory = new Map();
|
||
const internalPcs = new WeakSet();
|
||
const AUDIO_UPLINK_STORAGE_KEY = 'element-copilot-send-audio';
|
||
const VIDEO_UPLINK_STORAGE_KEY = 'element-copilot-send-video';
|
||
|
||
let selectedRemoteAudio = null;
|
||
let selectedRemoteVideo = null;
|
||
let selectedBusinessSender = null;
|
||
let originalBusinessTrack = null;
|
||
let aiAudioTrack = null;
|
||
|
||
let aiPc = null;
|
||
let aiWs = null;
|
||
let aiDc = null;
|
||
let aiAudioSender = null;
|
||
let aiVideoSender = null;
|
||
let isAiTakeover = false;
|
||
let aiAttached = false;
|
||
let inputVu = 0;
|
||
let outputVu = 0;
|
||
let inputHealthTimer = null;
|
||
let outputHealthTimer = null;
|
||
let assistantLine = null;
|
||
let assistantBuffer = '';
|
||
let inputRecorder = null;
|
||
let inputRecordingChunks = [];
|
||
let inputRecordingTrack = null;
|
||
let inputRecordingStartedAt = 0;
|
||
|
||
function log(message) {
|
||
console.log(`[AI Native Element] ${message}`);
|
||
}
|
||
|
||
function readBoolSetting(key, fallback) {
|
||
try {
|
||
const value = localStorage.getItem(key);
|
||
if (value === null) return fallback;
|
||
return value === 'true';
|
||
} catch {
|
||
return fallback;
|
||
}
|
||
}
|
||
|
||
function writeBoolSetting(key, value) {
|
||
try {
|
||
localStorage.setItem(key, value ? 'true' : 'false');
|
||
} catch {
|
||
// Ignore storage failures; the current checkbox state still applies.
|
||
}
|
||
}
|
||
|
||
function shouldSendAudioToAi() {
|
||
return readBoolSetting(AUDIO_UPLINK_STORAGE_KEY, true);
|
||
}
|
||
|
||
function shouldSendVideoToAi() {
|
||
return readBoolSetting(VIDEO_UPLINK_STORAGE_KEY, true);
|
||
}
|
||
|
||
function appendChatLine(role, text, important = false) {
|
||
const box = document.getElementById('element-copilot-log');
|
||
if (!box) return;
|
||
if (box.children.length === 1 && box.children[0].dataset.placeholder) box.innerHTML = '';
|
||
const row = document.createElement('div');
|
||
row.textContent = `[${new Date().toLocaleTimeString([], { hour12: false })}] ${role}: ${text}`;
|
||
row.style.marginBottom = '4px';
|
||
if (important) row.style.color = '#fcd34d';
|
||
box.appendChild(row);
|
||
while (box.children.length > 120) box.removeChild(box.firstChild);
|
||
box.scrollTop = box.scrollHeight;
|
||
return row;
|
||
}
|
||
|
||
function startAssistantLine() {
|
||
assistantBuffer = '';
|
||
assistantLine = null;
|
||
}
|
||
|
||
function appendAssistantDelta(delta) {
|
||
if (!delta) return;
|
||
const box = document.getElementById('element-copilot-log');
|
||
if (!box) {
|
||
console.log(`[AI Native Element] AI: ${delta}`);
|
||
return;
|
||
}
|
||
if (box.children.length === 1 && box.children[0].dataset.placeholder) box.innerHTML = '';
|
||
if (!assistantLine) assistantLine = appendChatLine('AI', '');
|
||
assistantBuffer += delta;
|
||
assistantLine.textContent = `[${new Date().toLocaleTimeString([], { hour12: false })}] AI: ${assistantBuffer}`;
|
||
box.scrollTop = box.scrollHeight;
|
||
}
|
||
|
||
function finishAssistantLine() {
|
||
assistantLine = null;
|
||
assistantBuffer = '';
|
||
}
|
||
|
||
function logUserTranscript(text) {
|
||
if (!text?.trim()) return;
|
||
finishAssistantLine();
|
||
appendChatLine('用户', text.trim(), true);
|
||
}
|
||
|
||
function recorderMimeType() {
|
||
if (window.MediaRecorder?.isTypeSupported?.('audio/webm;codecs=opus')) return 'audio/webm;codecs=opus';
|
||
if (window.MediaRecorder?.isTypeSupported?.('audio/webm')) return 'audio/webm';
|
||
return '';
|
||
}
|
||
|
||
function recordingFileName(track) {
|
||
const stamp = new Date().toISOString().replace(/[:.]/g, '-');
|
||
const trackPart = track?.id ? track.id.slice(0, 8) : 'no-track';
|
||
return `ai-input-${trackPart}-${stamp}.webm`;
|
||
}
|
||
|
||
async function startInputRecording() {
|
||
if (inputRecorder?.state === 'recording') return;
|
||
selectRemoteAudio('record');
|
||
if (!isLive(selectedRemoteAudio)) {
|
||
appendChatLine('录音', `没有可录制的 ${STATIC_AUDIO_ELEMENT_REASON} 音轨`, true);
|
||
return;
|
||
}
|
||
if (!window.MediaRecorder) {
|
||
appendChatLine('录音', '当前浏览器不支持 MediaRecorder', true);
|
||
return;
|
||
}
|
||
|
||
inputRecordingChunks = [];
|
||
inputRecordingTrack = selectedRemoteAudio;
|
||
inputRecordingStartedAt = Date.now();
|
||
const stream = new MediaStream([selectedRemoteAudio]);
|
||
const mimeType = recorderMimeType();
|
||
inputRecorder = new MediaRecorder(stream, mimeType ? { mimeType } : undefined);
|
||
inputRecorder.ondataavailable = (event) => {
|
||
if (event.data?.size > 0) inputRecordingChunks.push(event.data);
|
||
};
|
||
inputRecorder.onstop = () => {
|
||
const blob = new Blob(inputRecordingChunks, { type: inputRecorder?.mimeType || 'audio/webm' });
|
||
const url = URL.createObjectURL(blob);
|
||
const link = document.createElement('a');
|
||
link.href = url;
|
||
link.download = recordingFileName(inputRecordingTrack);
|
||
document.body.appendChild(link);
|
||
link.click();
|
||
link.remove();
|
||
setTimeout(() => URL.revokeObjectURL(url), 30000);
|
||
|
||
const seconds = ((Date.now() - inputRecordingStartedAt) / 1000).toFixed(1);
|
||
appendChatLine('录音', `已保存 ${link.download}（${seconds}s，${(blob.size / 1024).toFixed(1)} KB）`, true);
|
||
inputRecorder = null;
|
||
inputRecordingChunks = [];
|
||
inputRecordingTrack = null;
|
||
updateRecordButton();
|
||
};
|
||
inputRecorder.start(250);
|
||
appendChatLine('录音', `开始录制 AI 输入：${trackLabel(inputRecordingTrack)}`, true);
|
||
updateRecordButton();
|
||
}
|
||
|
||
function stopInputRecording() {
|
||
if (inputRecorder?.state === 'recording') inputRecorder.stop();
|
||
}
|
||
|
||
function toggleInputRecording() {
|
||
if (inputRecorder?.state === 'recording') stopInputRecording();
|
||
else startInputRecording();
|
||
}
|
||
|
||
function updateRecordButton() {
|
||
const button = document.getElementById('element-copilot-record');
|
||
if (!button) return;
|
||
const recording = inputRecorder?.state === 'recording';
|
||
button.textContent = recording ? '停止录制 AI 输入' : '录制 AI 输入';
|
||
button.style.background = recording ? '#dc2626' : '#2563eb';
|
||
}
|
||
|
||
function pcId() {
|
||
const bytes = new Uint8Array(16);
|
||
crypto.getRandomValues(bytes);
|
||
return 'PC-' + Array.from(bytes).map((b) => b.toString(16).padStart(2, '0')).join('');
|
||
}
|
||
|
||
function isInternalPc(pc) {
|
||
return Boolean(pc && (pc.__aiNativeElementInternal || internalPcs.has(pc) || pc === aiPc));
|
||
}
|
||
|
||
function isLive(track) {
|
||
return track && track.readyState === 'live';
|
||
}
|
||
|
||
function trackLabel(track) {
|
||
if (!track) return 'none';
|
||
return `${track.id}${track.muted ? ' muted' : ''}`;
|
||
}
|
||
|
||
function markInternalPc(pc) {
|
||
if (!pc) return;
|
||
pc.__aiNativeElementInternal = true;
|
||
internalPcs.add(pc);
|
||
}
|
||
|
||
function elementReasonFor(element) {
|
||
if (!element) return null;
|
||
const tag = element.tagName?.toLowerCase();
|
||
if (tag !== 'audio' && tag !== 'video') return null;
|
||
const elements = Array.from(document.querySelectorAll(tag));
|
||
const index = elements.indexOf(element);
|
||
if (index < 0) return `${tag}-srcObject`;
|
||
return `element-${index}:${tag}`;
|
||
}
|
||
|
||
function hasElementReason(item, reason) {
|
||
return Boolean(item?.reasons?.has(reason));
|
||
}
|
||
|
||
function rememberRemoteReceiver(receiver, reason) {
|
||
const track = receiver?.track;
|
||
if (!isLive(track) || isInternalPc(receiver?.transport)) return;
|
||
|
||
if (track.kind === 'audio' && !aiOutputTrackIds.has(track.id)) {
|
||
let item = remoteAudio.get(track.id);
|
||
if (!item) {
|
||
item = { track, receiver, reasons: new Set(), bytesDelta: 0, packetsDelta: 0 };
|
||
remoteAudio.set(track.id, item);
|
||
log(`audio candidate ${trackLabel(track)} (${reason})`, true);
|
||
}
|
||
item.track = track;
|
||
item.receiver = receiver || item.receiver;
|
||
item.reasons.add(reason);
|
||
selectRemoteAudio(reason);
|
||
}
|
||
|
||
if (track.kind === 'video') {
|
||
let item = remoteVideo.get(track.id);
|
||
if (!item) {
|
||
item = { track, receiver, reasons: new Set() };
|
||
remoteVideo.set(track.id, item);
|
||
log(`video candidate ${trackLabel(track)} (${reason})`, true);
|
||
}
|
||
item.track = track;
|
||
item.receiver = receiver || item.receiver;
|
||
item.reasons.add(reason);
|
||
selectRemoteVideo(reason);
|
||
}
|
||
}
|
||
|
||
function rememberRemoteStream(stream, reason) {
|
||
if (!stream || !(stream instanceof MediaStream)) return;
|
||
for (const track of stream.getAudioTracks()) {
|
||
if (aiOutputTrackIds.has(track.id)) continue;
|
||
let item = remoteAudio.get(track.id);
|
||
if (!item) {
|
||
item = { track, receiver: null, reasons: new Set(), bytesDelta: 0, packetsDelta: 0 };
|
||
remoteAudio.set(track.id, item);
|
||
log(`audio stream candidate ${trackLabel(track)} (${reason})`, true);
|
||
}
|
||
item.track = track;
|
||
item.reasons.add(reason);
|
||
}
|
||
for (const track of stream.getVideoTracks()) {
|
||
let item = remoteVideo.get(track.id);
|
||
if (!item) {
|
||
item = { track, receiver: null, reasons: new Set() };
|
||
remoteVideo.set(track.id, item);
|
||
log(`video stream candidate ${trackLabel(track)} (${reason})`, true);
|
||
}
|
||
item.track = track;
|
||
item.reasons.add(reason);
|
||
}
|
||
selectRemoteAudio(reason);
|
||
selectRemoteVideo(reason);
|
||
}
|
||
|
||
function selectRemoteAudio(reason) {
|
||
const current = selectedRemoteAudio && remoteAudio.get(selectedRemoteAudio.id);
|
||
if (isLive(selectedRemoteAudio) && !selectedRemoteAudio.muted && hasElementReason(current, STATIC_AUDIO_ELEMENT_REASON)) {
|
||
return selectedRemoteAudio;
|
||
}
|
||
|
||
const item = Array.from(remoteAudio.values()).find((candidate) => (
|
||
isLive(candidate.track)
|
||
&& !candidate.track.muted
|
||
&& hasElementReason(candidate, STATIC_AUDIO_ELEMENT_REASON)
|
||
));
|
||
if (!item) return null;
|
||
|
||
if (selectedRemoteAudio !== item.track) {
|
||
selectedRemoteAudio = item.track;
|
||
log(`selected input audio (${reason}) from ${STATIC_AUDIO_ELEMENT_REASON}: ${trackLabel(selectedRemoteAudio)}`, true);
|
||
attachAudioToAi();
|
||
}
|
||
return selectedRemoteAudio;
|
||
}
|
||
|
||
function selectRemoteVideo(reason) {
|
||
const current = selectedRemoteVideo && remoteVideo.get(selectedRemoteVideo.id);
|
||
if (isLive(selectedRemoteVideo) && !selectedRemoteVideo.muted && hasElementReason(current, STATIC_VIDEO_ELEMENT_REASON)) {
|
||
return selectedRemoteVideo;
|
||
}
|
||
|
||
const item = Array.from(remoteVideo.values()).find((candidate) => (
|
||
isLive(candidate.track)
|
||
&& !candidate.track.muted
|
||
&& hasElementReason(candidate, STATIC_VIDEO_ELEMENT_REASON)
|
||
));
|
||
if (!item) return null;
|
||
|
||
if (selectedRemoteVideo !== item.track) {
|
||
selectedRemoteVideo = item.track;
|
||
log(`selected input video (${reason}) from ${STATIC_VIDEO_ELEMENT_REASON}: ${trackLabel(selectedRemoteVideo)}`, true);
|
||
attachVideoToAi();
|
||
}
|
||
return selectedRemoteVideo;
|
||
}
|
||
|
||
function rememberBusinessSender(sender, track, reason) {
|
||
if (!sender) return;
|
||
const senderTrack = track || sender.track;
|
||
if (senderTrack && senderTrack.kind !== 'audio') return;
|
||
|
||
const id = sender.__aiNativeElementSenderId || `sender-${businessSenders.size + 1}`;
|
||
sender.__aiNativeElementSenderId = id;
|
||
|
||
let item = businessSenders.get(id);
|
||
if (!item) {
|
||
item = { id, sender, track: senderTrack || null, originalTrack: senderTrack || null, sources: new Set(), bytesDelta: 0, packetsDelta: 0 };
|
||
businessSenders.set(id, item);
|
||
log(`business sender candidate ${id} (${reason}): ${trackLabel(senderTrack)}`, true);
|
||
}
|
||
item.sender = sender;
|
||
item.track = sender.track || senderTrack || item.track;
|
||
if (!isAiTakeover && senderTrack && senderTrack !== aiAudioTrack) item.originalTrack = senderTrack;
|
||
item.sources.add(reason);
|
||
wrapBusinessSender(sender);
|
||
selectBusinessSender(reason);
|
||
}
|
||
|
||
function wrapBusinessSender(sender) {
|
||
if (!sender || sender.__aiNativeElementWrapped) return;
|
||
const originalReplaceTrack = sender.replaceTrack?.bind(sender);
|
||
if (!originalReplaceTrack) return;
|
||
|
||
sender.replaceTrack = async function (track) {
|
||
const result = await originalReplaceTrack(track);
|
||
rememberBusinessSender(sender, track, 'replaceTrack');
|
||
return result;
|
||
};
|
||
sender.__aiNativeElementWrapped = true;
|
||
}
|
||
|
||
function isRemoteInputTrack(track) {
|
||
return Boolean(track && remoteAudio.has(track.id) && !aiOutputTrackIds.has(track.id));
|
||
}
|
||
|
||
function selectBusinessSender(reason) {
|
||
if (selectedBusinessSender && !isRemoteInputTrack(selectedBusinessSender.track)) return selectedBusinessSender;
|
||
|
||
const item = Array.from(businessSenders.values()).find((candidate) => {
|
||
const track = candidate.sender.track || candidate.track || candidate.originalTrack;
|
||
return candidate.sources.has('pc-addTrack') && !isRemoteInputTrack(track);
|
||
});
|
||
if (!item) return null;
|
||
|
||
selectedBusinessSender = item.sender;
|
||
originalBusinessTrack = item.originalTrack || item.sender.track || originalBusinessTrack;
|
||
log(`selected output sender (${reason}): ${item.id}, track=${trackLabel(item.sender.track)}`, true);
|
||
attachAiOutputToBusiness();
|
||
return selectedBusinessSender;
|
||
}
|
||
|
||
function setVu(id, level) {
|
||
const el = document.getElementById(id);
|
||
if (el) el.style.width = `${Math.min(100, level * 2500)}%`;
|
||
}
|
||
|
||
function renderVu() {
|
||
inputVu *= 0.88;
|
||
outputVu *= 0.88;
|
||
setVu('element-copilot-input-vu', inputVu);
|
||
setVu('element-copilot-output-vu', outputVu);
|
||
requestAnimationFrame(renderVu);
|
||
}
|
||
|
||
async function refreshStats() {
|
||
await refreshInputStats();
|
||
await refreshOutputStats();
|
||
}
|
||
|
||
async function refreshInputStats() {
|
||
for (const item of remoteAudio.values()) {
|
||
if (!item.receiver?.getStats) continue;
|
||
const report = await item.receiver.getStats();
|
||
report.forEach((stat) => {
|
||
const kind = stat.kind || stat.mediaType;
|
||
if (stat.type !== 'inbound-rtp' || kind !== 'audio') return;
|
||
const prev = inboundMemory.get(stat.id);
|
||
const bytes = stat.bytesReceived || 0;
|
||
const packets = stat.packetsReceived || 0;
|
||
item.bytesReceived = bytes;
|
||
item.packetsReceived = packets;
|
||
item.bytesDelta = prev ? Math.max(0, bytes - prev.bytes) : 0;
|
||
item.packetsDelta = prev ? Math.max(0, packets - prev.packets) : 0;
|
||
inboundMemory.set(stat.id, { bytes, packets });
|
||
if (item.track === selectedRemoteAudio) {
|
||
inputVu = Math.max(inputVu, Math.min(0.12, item.bytesDelta / 12000), Math.min(0.04, item.packetsDelta / 500));
|
||
}
|
||
});
|
||
}
|
||
}
|
||
|
||
async function refreshOutputStats() {
|
||
for (const item of businessSenders.values()) {
|
||
if (!item.sender?.getStats) continue;
|
||
item.track = item.sender.track || item.track;
|
||
const report = await item.sender.getStats();
|
||
report.forEach((stat) => {
|
||
const kind = stat.kind || stat.mediaType;
|
||
if (stat.type !== 'outbound-rtp' || kind !== 'audio') return;
|
||
const key = `${item.id}:${stat.id}`;
|
||
const prev = outboundMemory.get(key);
|
||
const bytes = stat.bytesSent || 0;
|
||
const packets = stat.packetsSent || 0;
|
||
item.bytesSent = bytes;
|
||
item.packetsSent = packets;
|
||
item.bytesDelta = prev ? Math.max(0, bytes - prev.bytes) : 0;
|
||
item.packetsDelta = prev ? Math.max(0, packets - prev.packets) : 0;
|
||
outboundMemory.set(key, { bytes, packets });
|
||
if (item.sender === selectedBusinessSender) {
|
||
outputVu = Math.max(outputVu, Math.min(0.12, item.bytesDelta / 12000), Math.min(0.04, item.packetsDelta / 500));
|
||
}
|
||
});
|
||
}
|
||
}
|
||
|
||
async function outboundBytes(sender) {
|
||
if (!sender?.getStats) return null;
|
||
let total = 0;
|
||
const report = await sender.getStats();
|
||
report.forEach((stat) => {
|
||
const kind = stat.kind || stat.mediaType;
|
||
if (stat.type === 'outbound-rtp' && kind === 'audio') total += stat.bytesSent || 0;
|
||
});
|
||
return total;
|
||
}
|
||
|
||
function aiState() {
|
||
if (!aiPc) return 'pc=null';
|
||
return `conn=${aiPc.connectionState},ice=${aiPc.iceConnectionState},signaling=${aiPc.signalingState},remote=${aiPc.remoteDescription?.type || 'none'}`;
|
||
}
|
||
|
||
function aiReady() {
|
||
return Boolean(aiPc?.remoteDescription && (
|
||
aiPc.connectionState === 'connected'
|
||
|| aiPc.iceConnectionState === 'connected'
|
||
|| aiPc.iceConnectionState === 'completed'
|
||
));
|
||
}
|
||
|
||
function waitForAiReady(timeoutMs = 10000) {
|
||
if (aiReady()) return Promise.resolve(true);
|
||
return new Promise((resolve) => {
|
||
const startedAt = Date.now();
|
||
const timer = setInterval(() => {
|
||
if (aiReady()) {
|
||
clearInterval(timer);
|
||
resolve(true);
|
||
} else if (Date.now() - startedAt >= timeoutMs) {
|
||
clearInterval(timer);
|
||
resolve(false);
|
||
}
|
||
}, 250);
|
||
});
|
||
}
|
||
|
||
function checkAiInput(track) {
|
||
if (inputHealthTimer) clearTimeout(inputHealthTimer);
|
||
inputHealthTimer = setTimeout(async () => {
|
||
if (!aiAudioSender || aiAudioSender.track !== track) return;
|
||
await waitForAiReady();
|
||
const before = await outboundBytes(aiAudioSender);
|
||
await new Promise((resolve) => setTimeout(resolve, 2500));
|
||
const after = await outboundBytes(aiAudioSender);
|
||
log(`AI input packet check: before=${before}, after=${after}, ${aiState()}`, true);
|
||
}, 500);
|
||
}
|
||
|
||
function checkAiOutput(track) {
|
||
if (outputHealthTimer) clearTimeout(outputHealthTimer);
|
||
outputHealthTimer = setTimeout(async () => {
|
||
if (!selectedBusinessSender || selectedBusinessSender.track !== track) return;
|
||
const before = await outboundBytes(selectedBusinessSender);
|
||
await new Promise((resolve) => setTimeout(resolve, 2500));
|
||
const after = await outboundBytes(selectedBusinessSender);
|
||
log(`AI output packet check: before=${before}, after=${after}, sender=${trackLabel(selectedBusinessSender.track)}`, true);
|
||
}, 800);
|
||
}
|
||
|
||
async function attachAudioToAi() {
|
||
if (!shouldSendAudioToAi()) return false;
|
||
if (!aiAudioSender || !isLive(selectedRemoteAudio)) return false;
|
||
if (aiAudioSender.track === selectedRemoteAudio) return true;
|
||
await aiAudioSender.replaceTrack(selectedRemoteAudio);
|
||
log(`AI input uses ${trackLabel(selectedRemoteAudio)} from ${STATIC_AUDIO_ELEMENT_REASON}`, true);
|
||
checkAiInput(selectedRemoteAudio);
|
||
return true;
|
||
}
|
||
|
||
async function attachVideoToAi() {
|
||
if (!shouldSendVideoToAi()) return false;
|
||
if (!aiVideoSender || !isLive(selectedRemoteVideo)) return false;
|
||
if (aiVideoSender.track === selectedRemoteVideo) return true;
|
||
await aiVideoSender.replaceTrack(selectedRemoteVideo);
|
||
log(`AI video uses ${trackLabel(selectedRemoteVideo)} from ${STATIC_VIDEO_ELEMENT_REASON}`, true);
|
||
return true;
|
||
}
|
||
|
||
async function attachAiOutputToBusiness() {
|
||
selectBusinessSender('before-output');
|
||
if (!isAiTakeover || !selectedBusinessSender || !isLive(aiAudioTrack)) {
|
||
log(`AI output waiting: takeover=${isAiTakeover}, sender=${Boolean(selectedBusinessSender)}, aiTrack=${trackLabel(aiAudioTrack)}`);
|
||
return false;
|
||
}
|
||
if (selectedBusinessSender.track === aiAudioTrack) return true;
|
||
await selectedBusinessSender.replaceTrack(aiAudioTrack);
|
||
aiAttached = true;
|
||
log(`business sender uses AI audio ${trackLabel(aiAudioTrack)}`, true);
|
||
checkAiOutput(aiAudioTrack);
|
||
return true;
|
||
}
|
||
|
||
async function restoreBusinessAudio() {
|
||
if (!selectedBusinessSender || !aiAttached) return;
|
||
await selectedBusinessSender.replaceTrack(originalBusinessTrack || null);
|
||
aiAttached = false;
|
||
log(`business sender restored ${trackLabel(originalBusinessTrack)}`, true);
|
||
}
|
||
|
||
function wrapTrackListener(listener) {
|
||
if (!listener || listener.__aiNativeElementWrapped) return listener;
|
||
const wrapped = function (event) {
|
||
const ownerPc = this instanceof RTCPeerConnection ? this : event?.currentTarget;
|
||
if (!isInternalPc(ownerPc) && event?.track) {
|
||
rememberRemoteReceiver(event.receiver, 'pc-track');
|
||
for (const stream of event.streams || []) rememberRemoteStream(stream, 'pc-track-stream');
|
||
}
|
||
if (typeof listener === 'function') return listener.apply(this, arguments);
|
||
if (listener?.handleEvent) return listener.handleEvent(event);
|
||
return undefined;
|
||
};
|
||
wrapped.__aiNativeElementWrapped = true;
|
||
return wrapped;
|
||
}
|
||
|
||
RTCPeerConnection.prototype.addEventListener = function (type, listener, options) {
|
||
if (type === 'track') return originalAddEventListener.call(this, type, wrapTrackListener(listener), options);
|
||
return originalAddEventListener.call(this, type, listener, options);
|
||
};
|
||
|
||
try {
|
||
const descriptor = Object.getOwnPropertyDescriptor(RTCPeerConnection.prototype, 'ontrack');
|
||
Object.defineProperty(RTCPeerConnection.prototype, 'ontrack', {
|
||
configurable: true,
|
||
enumerable: true,
|
||
get() {
|
||
return descriptor?.get ? descriptor.get.call(this) : this.__aiNativeElementOnTrack || null;
|
||
},
|
||
set(listener) {
|
||
const wrapped = wrapTrackListener(listener);
|
||
this.__aiNativeElementOnTrack = wrapped;
|
||
if (descriptor?.set) descriptor.set.call(this, wrapped);
|
||
else this.addEventListener('track', wrapped);
|
||
},
|
||
});
|
||
} catch (error) {
|
||
console.warn('[AI Native Element] failed to wrap ontrack', error);
|
||
}
|
||
|
||
RTCPeerConnection.prototype.addTrack = function (track, ...streams) {
|
||
const sender = originalAddTrack.call(this, track, ...streams);
|
||
if (!isInternalPc(this) && track?.kind === 'audio') rememberBusinessSender(sender, track, 'pc-addTrack');
|
||
return sender;
|
||
};
|
||
|
||
if (originalAddTransceiver) {
|
||
RTCPeerConnection.prototype.addTransceiver = function (trackOrKind, init) {
|
||
const transceiver = originalAddTransceiver.call(this, trackOrKind, init);
|
||
if (!isInternalPc(this)) {
|
||
const kind = trackOrKind instanceof MediaStreamTrack ? trackOrKind.kind : trackOrKind;
|
||
if (kind === 'audio') {
|
||
const track = trackOrKind instanceof MediaStreamTrack ? trackOrKind : transceiver.sender?.track;
|
||
rememberBusinessSender(transceiver.sender, track, 'pc-addTransceiver');
|
||
}
|
||
}
|
||
return transceiver;
|
||
};
|
||
}
|
||
|
||
RTCPeerConnection.prototype.setRemoteDescription = function () {
|
||
const result = originalSetRemoteDescription.apply(this, arguments);
|
||
if (!isInternalPc(this)) {
|
||
Promise.resolve(result).then(() => {
|
||
for (const receiver of this.getReceivers?.() || []) rememberRemoteReceiver(receiver, 'receiver-after-srd');
|
||
for (const sender of this.getSenders?.() || []) rememberBusinessSender(sender, sender.track, 'sender-after-srd');
|
||
}).catch(() => {});
|
||
}
|
||
return result;
|
||
};
|
||
|
||
if (mediaSrcObjectDescriptor?.set) {
|
||
Object.defineProperty(HTMLMediaElement.prototype, 'srcObject', {
|
||
configurable: true,
|
||
enumerable: mediaSrcObjectDescriptor.enumerable,
|
||
get() {
|
||
return mediaSrcObjectDescriptor.get.call(this);
|
||
},
|
||
set(stream) {
|
||
if (!this.__aiNativeElementCopilotElement) {
|
||
const reason = elementReasonFor(this) || `${this.tagName.toLowerCase()}-srcObject`;
|
||
rememberRemoteStream(stream, reason);
|
||
}
|
||
return mediaSrcObjectDescriptor.set.call(this, stream);
|
||
},
|
||
});
|
||
}
|
||
|
||
function scanMediaElements() {
|
||
document.querySelectorAll('audio').forEach((element, index) => {
|
||
if (element.__aiNativeElementCopilotElement) return;
|
||
rememberRemoteStream(element.srcObject, `element-${index}:audio`);
|
||
});
|
||
document.querySelectorAll('video').forEach((element, index) => {
|
||
if (element.__aiNativeElementCopilotElement) return;
|
||
rememberRemoteStream(element.srcObject, `element-${index}:video`);
|
||
});
|
||
selectRemoteAudio('media-scan');
|
||
selectRemoteVideo('media-scan');
|
||
}
|
||
|
||
setInterval(scanMediaElements, 1000);
|
||
|
||
async function connectAi(signalingUrl, assistantId) {
|
||
disconnectAi();
|
||
|
||
const id = pcId();
|
||
aiPc = new RTCPeerConnection({ iceServers: [{ urls: 'stun:stun.l.google.com:19302' }] });
|
||
markInternalPc(aiPc);
|
||
aiPc.onconnectionstatechange = () => log(`AI PC connectionState=${aiPc.connectionState}`);
|
||
aiPc.oniceconnectionstatechange = () => log(`AI PC iceConnectionState=${aiPc.iceConnectionState}`);
|
||
|
||
aiDc = aiPc.createDataChannel('chat');
|
||
aiDc.onopen = () => {
|
||
setStatus('connected');
|
||
log('AI data channel open', true);
|
||
aiDc.send(JSON.stringify({ type: 'client-ready' }));
|
||
};
|
||
aiDc.onclose = () => setStatus('disconnected');
|
||
aiDc.onmessage = (event) => {
|
||
try {
|
||
const msg = JSON.parse(event.data);
|
||
const content = msg.delta || msg.content;
|
||
if (msg.type === 'transcript') {
|
||
logUserTranscript(content);
|
||
} else if (msg.type === 'assistant-text-start') {
|
||
startAssistantLine();
|
||
} else if (msg.type === 'assistant-text-delta') {
|
||
appendAssistantDelta(content);
|
||
} else if (msg.type === 'assistant-text-end') {
|
||
finishAssistantLine();
|
||
} else {
|
||
log(`data channel message: ${msg.type}`);
|
||
}
|
||
} catch {
|
||
appendChatLine('DATA', event.data);
|
||
}
|
||
};
|
||
|
||
scanMediaElements();
|
||
selectRemoteAudio('connect');
|
||
selectRemoteVideo('connect');
|
||
selectBusinessSender('connect');
|
||
|
||
const sendAudio = shouldSendAudioToAi();
|
||
const audioTransceiver = sendAudio && isLive(selectedRemoteAudio)
|
||
? aiPc.addTransceiver(selectedRemoteAudio, { direction: 'sendrecv' })
|
||
: aiPc.addTransceiver('audio', { direction: sendAudio ? 'sendrecv' : 'recvonly' });
|
||
aiAudioSender = audioTransceiver.sender;
|
||
log(!sendAudio
|
||
? 'AI audio uplink disabled by panel setting'
|
||
: isLive(selectedRemoteAudio)
|
||
? `AI input starts with ${trackLabel(selectedRemoteAudio)} (${STATIC_AUDIO_ELEMENT_REASON})`
|
||
: `AI input waiting for ${STATIC_AUDIO_ELEMENT_REASON}`, true);
|
||
|
||
const sendVideo = shouldSendVideoToAi();
|
||
const shouldSendVideo = sendVideo && isLive(selectedRemoteVideo) && !selectedRemoteVideo.muted;
|
||
if (shouldSendVideo) {
|
||
const videoTransceiver = aiPc.addTransceiver(selectedRemoteVideo, { direction: 'sendonly' });
|
||
aiVideoSender = videoTransceiver.sender;
|
||
log(`AI video starts with ${trackLabel(selectedRemoteVideo)} (${STATIC_VIDEO_ELEMENT_REASON})`);
|
||
} else {
|
||
aiVideoSender = null;
|
||
log(sendVideo
|
||
? `AI video disabled: waiting for live track at ${STATIC_VIDEO_ELEMENT_REASON}`
|
||
: 'AI video uplink disabled by panel setting');
|
||
}
|
||
|
||
aiPc.ontrack = (event) => {
|
||
if (event.track.kind !== 'audio') return;
|
||
aiOutputTrackIds.add(event.track.id);
|
||
remoteAudio.delete(event.track.id);
|
||
aiAudioTrack = event.track;
|
||
log(`received AI audio ${trackLabel(aiAudioTrack)}`, true);
|
||
aiAudioTrack.onunmute = () => {
|
||
log(`AI audio unmuted ${trackLabel(aiAudioTrack)}`);
|
||
attachAiOutputToBusiness();
|
||
};
|
||
attachAiOutputToBusiness();
|
||
};
|
||
|
||
aiPc.onicecandidate = (event) => {
|
||
if (aiWs?.readyState !== WebSocket.OPEN) return;
|
||
aiWs.send(JSON.stringify({
|
||
type: 'ice-candidate',
|
||
payload: {
|
||
pc_id: id,
|
||
candidate: event.candidate ? {
|
||
candidate: event.candidate.candidate,
|
||
sdpMid: event.candidate.sdpMid,
|
||
sdpMLineIndex: event.candidate.sdpMLineIndex,
|
||
} : null,
|
||
},
|
||
}));
|
||
};
|
||
|
||
setStatus('connecting');
|
||
aiWs = new WebSocket(signalingUrl);
|
||
aiWs.onopen = async () => {
|
||
const offer = await aiPc.createOffer();
|
||
await aiPc.setLocalDescription(offer);
|
||
aiWs.send(JSON.stringify({
|
||
type: 'offer',
|
||
payload: {
|
||
pc_id: id,
|
||
sdp: aiPc.localDescription.sdp,
|
||
type: aiPc.localDescription.type,
|
||
assistant_id: assistantId,
|
||
vision_enabled: shouldSendVideo,
|
||
},
|
||
}));
|
||
log('sent AI WebRTC offer', true);
|
||
};
|
||
aiWs.onmessage = async (event) => {
|
||
try {
|
||
const msg = JSON.parse(event.data);
|
||
if (msg.type === 'answer') {
|
||
await aiPc.setRemoteDescription({ type: 'answer', sdp: msg.payload.sdp });
|
||
log(`AI WebRTC answer applied; ${aiState()}`, true);
|
||
} else if (msg.type === 'ice-candidate' && msg.payload?.candidate) {
|
||
await aiPc.addIceCandidate(msg.payload.candidate).catch(() => {});
|
||
} else if (msg.type === 'error') {
|
||
throw new Error(msg.payload?.message || 'server error');
|
||
}
|
||
} catch (error) {
|
||
log(`signaling error: ${error.message || error}`);
|
||
}
|
||
};
|
||
aiWs.onerror = () => {
|
||
setStatus('failed');
|
||
log('AI signaling websocket error');
|
||
};
|
||
aiWs.onclose = () => setStatus('disconnected');
|
||
|
||
attachAudioToAi();
|
||
attachVideoToAi();
|
||
}
|
||
|
||
function disconnectAi() {
|
||
try { aiDc?.close(); } catch {}
|
||
try { aiPc?.close(); } catch {}
|
||
try { aiWs?.close(); } catch {}
|
||
aiDc = null;
|
||
aiPc = null;
|
||
aiWs = null;
|
||
aiAudioSender = null;
|
||
aiVideoSender = null;
|
||
aiAudioTrack = null;
|
||
if (inputRecorder?.state === 'recording') stopInputRecording();
|
||
setStatus('disconnected');
|
||
}
|
||
|
||
function setStatus(status) {
|
||
const el = document.getElementById('element-copilot-status');
|
||
if (!el) return;
|
||
el.textContent = {
|
||
disconnected: '未连接',
|
||
connecting: '连接中',
|
||
connected: '已连接',
|
||
failed: '连接失败',
|
||
}[status] || status;
|
||
}
|
||
|
||
function updateModeButton() {
|
||
const button = document.getElementById('element-copilot-toggle');
|
||
if (!button) return;
|
||
if (isAiTakeover) {
|
||
button.textContent = '当前：AI 接管｜点击切回人工';
|
||
button.style.background = '#8b5cf6';
|
||
} else {
|
||
button.textContent = '当前：人工接管｜点击切换 AI';
|
||
button.style.background = '#10b981';
|
||
}
|
||
}
|
||
|
||
window.__aiNativeElementDump = function () {
|
||
return {
|
||
staticAudioElement: STATIC_AUDIO_ELEMENT_REASON,
|
||
staticVideoElement: STATIC_VIDEO_ELEMENT_REASON,
|
||
selectedInput: trackLabel(selectedRemoteAudio),
|
||
selectedVideo: trackLabel(selectedRemoteVideo),
|
||
aiVideoSenderTrack: trackLabel(aiVideoSender?.track),
|
||
selectedOutput: trackLabel(selectedBusinessSender?.track),
|
||
originalOutput: trackLabel(originalBusinessTrack),
|
||
aiAudio: trackLabel(aiAudioTrack),
|
||
aiAttached,
|
||
aiState: aiState(),
|
||
recording: inputRecorder?.state === 'recording' ? {
|
||
track: trackLabel(inputRecordingTrack),
|
||
seconds: ((Date.now() - inputRecordingStartedAt) / 1000).toFixed(1),
|
||
chunks: inputRecordingChunks.length,
|
||
} : null,
|
||
inputs: Array.from(remoteAudio.values()).map((item) => ({
|
||
selected: item.track === selectedRemoteAudio,
|
||
id: item.track.id,
|
||
muted: item.track.muted,
|
||
live: item.track.readyState,
|
||
hasReceiver: Boolean(item.receiver),
|
||
matchesStaticElement: hasElementReason(item, STATIC_AUDIO_ELEMENT_REASON),
|
||
bytesReceived: item.bytesReceived || 0,
|
||
packetsReceived: item.packetsReceived || 0,
|
||
bytesDelta: item.bytesDelta,
|
||
packetsDelta: item.packetsDelta,
|
||
reasons: Array.from(item.reasons),
|
||
})),
|
||
videos: Array.from(remoteVideo.values()).map((item) => ({
|
||
selected: item.track === selectedRemoteVideo,
|
||
id: item.track.id,
|
||
muted: item.track.muted,
|
||
live: item.track.readyState,
|
||
hasReceiver: Boolean(item.receiver),
|
||
matchesStaticElement: hasElementReason(item, STATIC_VIDEO_ELEMENT_REASON),
|
||
reasons: Array.from(item.reasons),
|
||
})),
|
||
outputs: Array.from(businessSenders.values()).map((item) => ({
|
||
selected: item.sender === selectedBusinessSender,
|
||
id: item.id,
|
||
track: trackLabel(item.sender.track || item.track),
|
||
originalTrack: trackLabel(item.originalTrack),
|
||
pcAddTrack: item.sources.has('pc-addTrack'),
|
||
isRemoteInput: isRemoteInputTrack(item.sender.track || item.track),
|
||
bytesSent: item.bytesSent,
|
||
packetsSent: item.packetsSent,
|
||
bytesDelta: item.bytesDelta,
|
||
packetsDelta: item.packetsDelta,
|
||
sources: Array.from(item.sources),
|
||
})),
|
||
};
|
||
};
|
||
|
||
function renderPanel() {
|
||
if (!document.body || document.getElementById('element-copilot-panel')) return;
|
||
const panel = document.createElement('div');
|
||
panel.id = 'element-copilot-panel';
|
||
panel.style.cssText = [
|
||
'position:fixed',
|
||
'top:20px',
|
||
'left:20px',
|
||
'width:360px',
|
||
'z-index:999999',
|
||
'box-sizing:border-box',
|
||
'padding:16px',
|
||
'border-radius:10px',
|
||
'background:rgba(17,24,39,.96)',
|
||
'color:white',
|
||
'font-family:system-ui,-apple-system,BlinkMacSystemFont,sans-serif',
|
||
'box-shadow:0 12px 32px rgba(0,0,0,.35)',
|
||
].join(';');
|
||
panel.innerHTML = `
|
||
<div style="display:flex;justify-content:space-between;gap:12px;align-items:center;margin-bottom:12px;">
|
||
<strong style="font-size:15px;color:#f9a8d4;">AI 元素选轨</strong>
|
||
<span id="element-copilot-status" style="font-size:12px;color:#d1d5db;">未连接</span>
|
||
</div>
|
||
<div style="font-size:11px;color:#9ca3af;margin-bottom:10px;line-height:1.5;">
|
||
音频输入：${STATIC_AUDIO_ELEMENT_REASON}<br>
|
||
视频输入：${STATIC_VIDEO_ELEMENT_REASON}
|
||
</div>
|
||
<label style="display:block;font-size:12px;color:#9ca3af;margin-bottom:4px;">信令地址</label>
|
||
<input id="element-copilot-url" value="${DEFAULT_SIGNALING_URL}" style="width:100%;box-sizing:border-box;margin-bottom:10px;padding:8px;border:1px solid #4b5563;border-radius:6px;background:#111827;color:white;">
|
||
<label style="display:block;font-size:12px;color:#9ca3af;margin-bottom:4px;">助手 ID</label>
|
||
<input id="element-copilot-assistant" placeholder="请输入助手 ID" style="width:100%;box-sizing:border-box;margin-bottom:12px;padding:8px;border:1px solid #4b5563;border-radius:6px;background:#111827;color:white;">
|
||
<div style="display:grid;grid-template-columns:1fr 1fr;gap:8px;margin-bottom:12px;">
|
||
<label style="display:flex;align-items:center;gap:6px;padding:8px;border:1px solid #374151;border-radius:6px;background:#111827;font-size:12px;color:#e5e7eb;cursor:pointer;">
|
||
<input id="element-copilot-send-audio" type="checkbox" style="margin:0;">
|
||
<span>传音频给 AI</span>
|
||
</label>
|
||
<label style="display:flex;align-items:center;gap:6px;padding:8px;border:1px solid #374151;border-radius:6px;background:#111827;font-size:12px;color:#e5e7eb;cursor:pointer;">
|
||
<input id="element-copilot-send-video" type="checkbox" style="margin:0;">
|
||
<span>传视频给 AI</span>
|
||
</label>
|
||
</div>
|
||
<div style="margin-bottom:10px;">
|
||
<div style="display:flex;justify-content:space-between;font-size:12px;color:#d1fae5;margin-bottom:4px;"><span>${STATIC_AUDIO_ELEMENT_REASON}</span><span>原生音轨</span></div>
|
||
<div style="height:8px;border-radius:999px;overflow:hidden;background:#374151;"><div id="element-copilot-input-vu" style="height:100%;width:0%;background:linear-gradient(90deg,#10b981,#fcd34d);transition:width .06s linear;"></div></div>
|
||
</div>
|
||
<div style="margin-bottom:12px;">
|
||
<div style="display:flex;justify-content:space-between;font-size:12px;color:#bfdbfe;margin-bottom:4px;"><span>AI 输出</span><span>业务上行</span></div>
|
||
<div style="height:8px;border-radius:999px;overflow:hidden;background:#374151;"><div id="element-copilot-output-vu" style="height:100%;width:0%;background:linear-gradient(90deg,#60a5fa,#c084fc);transition:width .06s linear;"></div></div>
|
||
</div>
|
||
<button id="element-copilot-toggle" style="width:100%;padding:12px;border:0;border-radius:6px;background:#10b981;color:white;font-weight:700;cursor:pointer;">当前：人工接管｜点击切换 AI</button>
|
||
<button id="element-copilot-record" style="width:100%;margin-top:8px;padding:10px;border:0;border-radius:6px;background:#2563eb;color:white;font-weight:700;cursor:pointer;">录制 AI 输入</button>
|
||
<div id="element-copilot-log" style="height:170px;margin-top:12px;padding:8px;overflow:auto;border:1px solid #374151;border-radius:6px;background:#030712;color:#a7f3d0;font:11px ui-monospace,Menlo,monospace;">
|
||
<span data-placeholder="1" style="color:#6b7280;">等待交互...</span>
|
||
</div>
|
||
`;
|
||
document.body.appendChild(panel);
|
||
|
||
const toggle = document.getElementById('element-copilot-toggle');
|
||
const recordButton = document.getElementById('element-copilot-record');
|
||
const urlInput = document.getElementById('element-copilot-url');
|
||
const assistantInput = document.getElementById('element-copilot-assistant');
|
||
const sendAudioInput = document.getElementById('element-copilot-send-audio');
|
||
const sendVideoInput = document.getElementById('element-copilot-send-video');
|
||
sendAudioInput.checked = shouldSendAudioToAi();
|
||
sendVideoInput.checked = shouldSendVideoToAi();
|
||
sendAudioInput.addEventListener('change', () => {
|
||
writeBoolSetting(AUDIO_UPLINK_STORAGE_KEY, sendAudioInput.checked);
|
||
log(`AI audio uplink setting: ${sendAudioInput.checked ? 'on' : 'off'}`);
|
||
});
|
||
sendVideoInput.addEventListener('change', () => {
|
||
writeBoolSetting(VIDEO_UPLINK_STORAGE_KEY, sendVideoInput.checked);
|
||
log(`AI video uplink setting: ${sendVideoInput.checked ? 'on' : 'off'}`);
|
||
});
|
||
recordButton.addEventListener('click', toggleInputRecording);
|
||
toggle.addEventListener('click', async () => {
|
||
if (!isAiTakeover && !assistantInput.value.trim()) {
|
||
alert('请先填写助手 ID。');
|
||
return;
|
||
}
|
||
isAiTakeover = !isAiTakeover;
|
||
updateModeButton();
|
||
if (isAiTakeover) {
|
||
await connectAi(urlInput.value.trim(), assistantInput.value.trim());
|
||
await attachAiOutputToBusiness();
|
||
} else {
|
||
await restoreBusinessAudio();
|
||
disconnectAi();
|
||
}
|
||
});
|
||
updateModeButton();
|
||
updateRecordButton();
|
||
}
|
||
|
||
function renderPanelSoon() {
|
||
try { renderPanel(); } catch (error) { console.warn('[AI Native Element] panel error', error); }
|
||
}
|
||
|
||
if (document.readyState === 'loading') {
|
||
document.addEventListener('DOMContentLoaded', renderPanelSoon, { once: true });
|
||
} else {
|
||
renderPanelSoon();
|
||
}
|
||
requestAnimationFrame(renderVu);
|
||
setInterval(refreshStats, 250);
|
||
setTimeout(renderPanelSoon, 1500);
|
||
setTimeout(renderPanelSoon, 3000);
|
||
window.addEventListener('beforeunload', () => {
|
||
if (inputRecorder?.state === 'recording') stopInputRecording();
|
||
});
|
||
})();
|