Spaces:
Sleeping
Sleeping
Upload 58 files
Browse files- components/LiveAssistant.tsx +162 -16
components/LiveAssistant.tsx
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
|
| 2 |
import React, { useState, useRef, useEffect } from 'react';
|
| 3 |
-
import { Mic, X, Power, Loader2, Bot, Volume2, Radio, Activity, RefreshCw, ChevronDown } from 'lucide-react';
|
| 4 |
import { api } from '../services/api';
|
| 5 |
|
| 6 |
// --- Audio Types & Helpers ---
|
|
@@ -23,7 +23,16 @@ export const LiveAssistant: React.FC = () => {
|
|
| 23 |
const [transcript, setTranscript] = useState('');
|
| 24 |
const [volumeLevel, setVolumeLevel] = useState(0);
|
| 25 |
|
| 26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
const mediaStreamRef = useRef<MediaStream | null>(null);
|
| 28 |
const processorRef = useRef<ScriptProcessorNode | null>(null);
|
| 29 |
const sourceNodeRef = useRef<MediaStreamAudioSourceNode | null>(null);
|
|
@@ -33,16 +42,75 @@ export const LiveAssistant: React.FC = () => {
|
|
| 33 |
const nextPlayTimeRef = useRef<number>(0);
|
| 34 |
const analyserRef = useRef<AnalyserNode | null>(null);
|
| 35 |
const volumeIntervalRef = useRef<any>(null);
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
useEffect(() => {
|
| 38 |
if (!isOpen) {
|
| 39 |
handleDisconnect();
|
| 40 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
return () => {
|
| 42 |
handleDisconnect();
|
| 43 |
};
|
| 44 |
}, [isOpen]);
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
useEffect(() => {
|
| 47 |
if (status === 'DISCONNECTED') {
|
| 48 |
setVolumeLevel(0);
|
|
@@ -59,7 +127,48 @@ export const LiveAssistant: React.FC = () => {
|
|
| 59 |
return () => clearInterval(volumeIntervalRef.current);
|
| 60 |
}, [status]);
|
| 61 |
|
| 62 |
-
const
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
if (!audioContextRef.current) {
|
| 64 |
// @ts-ignore
|
| 65 |
const AudioCtor = window.AudioContext || window.webkitAudioContext;
|
|
@@ -88,7 +197,7 @@ export const LiveAssistant: React.FC = () => {
|
|
| 88 |
setTranscript('正在连接服务器...');
|
| 89 |
|
| 90 |
try {
|
| 91 |
-
|
| 92 |
|
| 93 |
const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
|
| 94 |
// Append User Info to ensure backend creates a distinct session context
|
|
@@ -186,18 +295,30 @@ export const LiveAssistant: React.FC = () => {
|
|
| 186 |
setStatus('CONNECTED');
|
| 187 |
}
|
| 188 |
|
|
|
|
|
|
|
| 189 |
const stream = await navigator.mediaDevices.getUserMedia({ audio: {
|
| 190 |
sampleRate: INPUT_SAMPLE_RATE,
|
| 191 |
channelCount: 1,
|
| 192 |
-
echoCancellation: true
|
|
|
|
|
|
|
| 193 |
}});
|
| 194 |
mediaStreamRef.current = stream;
|
| 195 |
|
| 196 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
const source = ctx.createMediaStreamSource(stream);
|
| 198 |
const processor = ctx.createScriptProcessor(4096, 1, 1);
|
| 199 |
|
| 200 |
processor.onaudioprocess = (e) => {
|
|
|
|
|
|
|
|
|
|
| 201 |
const inputData = e.inputBuffer.getChannelData(0);
|
| 202 |
const l = inputData.length;
|
| 203 |
const int16Data = new Int16Array(l);
|
|
@@ -231,14 +352,18 @@ export const LiveAssistant: React.FC = () => {
|
|
| 231 |
|
| 232 |
} catch (e) {
|
| 233 |
console.error(e);
|
|
|
|
| 234 |
setTranscript('无法访问麦克风');
|
| 235 |
}
|
| 236 |
};
|
| 237 |
|
| 238 |
const stopRecording = () => {
|
| 239 |
-
if (
|
|
|
|
| 240 |
|
|
|
|
| 241 |
if (processorRef.current) {
|
|
|
|
| 242 |
processorRef.current.disconnect();
|
| 243 |
processorRef.current = null;
|
| 244 |
}
|
|
@@ -250,6 +375,11 @@ export const LiveAssistant: React.FC = () => {
|
|
| 250 |
mediaStreamRef.current.getTracks().forEach(t => t.stop());
|
| 251 |
mediaStreamRef.current = null;
|
| 252 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 253 |
|
| 254 |
setStatus('THINKING');
|
| 255 |
setTranscript('思考中...');
|
|
@@ -273,26 +403,41 @@ export const LiveAssistant: React.FC = () => {
|
|
| 273 |
if (!api.auth.getCurrentUser()) return null;
|
| 274 |
|
| 275 |
return (
|
| 276 |
-
<div
|
|
|
|
|
|
|
|
|
|
|
|
|
| 277 |
{!isOpen && (
|
| 278 |
-
<
|
| 279 |
-
|
| 280 |
-
|
|
|
|
| 281 |
>
|
| 282 |
-
<
|
| 283 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 284 |
)}
|
| 285 |
|
| 286 |
{isOpen && (
|
| 287 |
<div className="bg-slate-900 w-80 md:w-96 rounded-3xl shadow-2xl border border-slate-700 overflow-hidden flex flex-col animate-in slide-in-from-bottom-5 fade-in duration-300 h-[500px]">
|
| 288 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 289 |
<div className="flex items-center gap-2">
|
| 290 |
<div className={`w-2 h-2 rounded-full ${status === 'DISCONNECTED' ? 'bg-red-500' : 'bg-green-500 animate-pulse'}`}></div>
|
| 291 |
<span className="font-bold text-sm">AI 实时通话 (代理模式)</span>
|
| 292 |
</div>
|
| 293 |
<div className="flex gap-2">
|
| 294 |
-
<button onClick={handleDisconnect} title="重置" className="hover:bg-white/10 p-1.5 rounded-full text-gray-400 hover:text-white transition-colors"><RefreshCw size={16}/></button>
|
| 295 |
-
<button onClick={() => setIsOpen(false)} title="最小化" className="hover:bg-white/10 p-1.5 rounded-full text-gray-400 hover:text-white transition-colors"><ChevronDown size={20}/></button>
|
| 296 |
</div>
|
| 297 |
</div>
|
| 298 |
|
|
@@ -359,6 +504,7 @@ export const LiveAssistant: React.FC = () => {
|
|
| 359 |
<button
|
| 360 |
onMouseDown={startRecording}
|
| 361 |
onMouseUp={stopRecording}
|
|
|
|
| 362 |
onTouchStart={(e) => { e.preventDefault(); startRecording(); }}
|
| 363 |
onTouchEnd={(e) => { e.preventDefault(); stopRecording(); }}
|
| 364 |
className={`w-20 h-20 rounded-full flex items-center justify-center shadow-lg transition-all transform ${
|
|
|
|
| 1 |
|
| 2 |
import React, { useState, useRef, useEffect } from 'react';
|
| 3 |
+
import { Mic, X, Power, Loader2, Bot, Volume2, Radio, Activity, RefreshCw, ChevronDown, Move } from 'lucide-react';
|
| 4 |
import { api } from '../services/api';
|
| 5 |
|
| 6 |
// --- Audio Types & Helpers ---
|
|
|
|
| 23 |
const [transcript, setTranscript] = useState('');
|
| 24 |
const [volumeLevel, setVolumeLevel] = useState(0);
|
| 25 |
|
| 26 |
+
// Dragging State
|
| 27 |
+
const [position, setPosition] = useState<{x: number, y: number} | null>(null);
|
| 28 |
+
const containerRef = useRef<HTMLDivElement>(null);
|
| 29 |
+
const dragRef = useRef({ isDragging: false, startX: 0, startY: 0, initialLeft: 0, initialTop: 0 });
|
| 30 |
+
const hasMovedRef = useRef(false);
|
| 31 |
+
|
| 32 |
+
// Refs
|
| 33 |
+
const audioContextRef = useRef<AudioContext | null>(null); // Output Context
|
| 34 |
+
const inputAudioContextRef = useRef<AudioContext | null>(null); // Input Context (New)
|
| 35 |
+
|
| 36 |
const mediaStreamRef = useRef<MediaStream | null>(null);
|
| 37 |
const processorRef = useRef<ScriptProcessorNode | null>(null);
|
| 38 |
const sourceNodeRef = useRef<MediaStreamAudioSourceNode | null>(null);
|
|
|
|
| 42 |
const nextPlayTimeRef = useRef<number>(0);
|
| 43 |
const analyserRef = useRef<AnalyserNode | null>(null);
|
| 44 |
const volumeIntervalRef = useRef<any>(null);
|
| 45 |
+
|
| 46 |
+
// Track recording state for the processor callback to avoid stale closures or race conditions
|
| 47 |
+
const isRecordingRef = useRef(false);
|
| 48 |
|
| 49 |
useEffect(() => {
|
| 50 |
if (!isOpen) {
|
| 51 |
handleDisconnect();
|
| 52 |
}
|
| 53 |
+
// When opening/closing, ensure we stay within bounds if position exists
|
| 54 |
+
if (position && containerRef.current) {
|
| 55 |
+
const { innerWidth, innerHeight } = window;
|
| 56 |
+
const rect = containerRef.current.getBoundingClientRect();
|
| 57 |
+
const newX = Math.min(Math.max(0, position.x), innerWidth - rect.width);
|
| 58 |
+
const newY = Math.min(Math.max(0, position.y), innerHeight - rect.height);
|
| 59 |
+
if (newX !== position.x || newY !== position.y) {
|
| 60 |
+
setPosition({ x: newX, y: newY });
|
| 61 |
+
}
|
| 62 |
+
}
|
| 63 |
return () => {
|
| 64 |
handleDisconnect();
|
| 65 |
};
|
| 66 |
}, [isOpen]);
|
| 67 |
|
| 68 |
+
useEffect(() => {
|
| 69 |
+
const handleMove = (e: MouseEvent | TouchEvent) => {
|
| 70 |
+
if (!dragRef.current.isDragging) return;
|
| 71 |
+
|
| 72 |
+
const clientX = 'touches' in e ? e.touches[0].clientX : (e as MouseEvent).clientX;
|
| 73 |
+
const clientY = 'touches' in e ? e.touches[0].clientY : (e as MouseEvent).clientY;
|
| 74 |
+
|
| 75 |
+
const deltaX = clientX - dragRef.current.startX;
|
| 76 |
+
const deltaY = clientY - dragRef.current.startY;
|
| 77 |
+
|
| 78 |
+
if (Math.abs(deltaX) > 5 || Math.abs(deltaY) > 5) {
|
| 79 |
+
hasMovedRef.current = true;
|
| 80 |
+
}
|
| 81 |
+
|
| 82 |
+
let newX = dragRef.current.initialLeft + deltaX;
|
| 83 |
+
let newY = dragRef.current.initialTop + deltaY;
|
| 84 |
+
|
| 85 |
+
// Bounds check
|
| 86 |
+
if (containerRef.current) {
|
| 87 |
+
const rect = containerRef.current.getBoundingClientRect();
|
| 88 |
+
const { innerWidth, innerHeight } = window;
|
| 89 |
+
newX = Math.min(Math.max(0, newX), innerWidth - rect.width);
|
| 90 |
+
newY = Math.min(Math.max(0, newY), innerHeight - rect.height);
|
| 91 |
+
}
|
| 92 |
+
|
| 93 |
+
setPosition({ x: newX, y: newY });
|
| 94 |
+
};
|
| 95 |
+
|
| 96 |
+
const handleUp = () => {
|
| 97 |
+
dragRef.current.isDragging = false;
|
| 98 |
+
document.body.style.userSelect = '';
|
| 99 |
+
};
|
| 100 |
+
|
| 101 |
+
window.addEventListener('mousemove', handleMove);
|
| 102 |
+
window.addEventListener('mouseup', handleUp);
|
| 103 |
+
window.addEventListener('touchmove', handleMove, { passive: false });
|
| 104 |
+
window.addEventListener('touchend', handleUp);
|
| 105 |
+
|
| 106 |
+
return () => {
|
| 107 |
+
window.removeEventListener('mousemove', handleMove);
|
| 108 |
+
window.removeEventListener('mouseup', handleUp);
|
| 109 |
+
window.removeEventListener('touchmove', handleMove);
|
| 110 |
+
window.removeEventListener('touchend', handleUp);
|
| 111 |
+
};
|
| 112 |
+
}, []);
|
| 113 |
+
|
| 114 |
useEffect(() => {
|
| 115 |
if (status === 'DISCONNECTED') {
|
| 116 |
setVolumeLevel(0);
|
|
|
|
| 127 |
return () => clearInterval(volumeIntervalRef.current);
|
| 128 |
}, [status]);
|
| 129 |
|
| 130 |
+
const handleDragStart = (e: React.MouseEvent | React.TouchEvent) => {
|
| 131 |
+
if (!containerRef.current) return;
|
| 132 |
+
|
| 133 |
+
// Prevent default to stop scrolling on mobile while dragging
|
| 134 |
+
// e.preventDefault(); // Note: might block click if not careful, handled by checking dragging state
|
| 135 |
+
|
| 136 |
+
const clientX = 'touches' in e ? e.touches[0].clientX : (e as React.MouseEvent).clientX;
|
| 137 |
+
const clientY = 'touches' in e ? e.touches[0].clientY : (e as React.MouseEvent).clientY;
|
| 138 |
+
|
| 139 |
+
const rect = containerRef.current.getBoundingClientRect();
|
| 140 |
+
|
| 141 |
+
// If position is null (initial state), set it to current computed position
|
| 142 |
+
if (!position) {
|
| 143 |
+
setPosition({ x: rect.left, y: rect.top });
|
| 144 |
+
dragRef.current = {
|
| 145 |
+
isDragging: true,
|
| 146 |
+
startX: clientX,
|
| 147 |
+
startY: clientY,
|
| 148 |
+
initialLeft: rect.left,
|
| 149 |
+
initialTop: rect.top
|
| 150 |
+
};
|
| 151 |
+
} else {
|
| 152 |
+
dragRef.current = {
|
| 153 |
+
isDragging: true,
|
| 154 |
+
startX: clientX,
|
| 155 |
+
startY: clientY,
|
| 156 |
+
initialLeft: position.x,
|
| 157 |
+
initialTop: position.y
|
| 158 |
+
};
|
| 159 |
+
}
|
| 160 |
+
|
| 161 |
+
hasMovedRef.current = false;
|
| 162 |
+
document.body.style.userSelect = 'none';
|
| 163 |
+
};
|
| 164 |
+
|
| 165 |
+
const handleToggleOpen = () => {
|
| 166 |
+
if (!hasMovedRef.current) {
|
| 167 |
+
setIsOpen(prev => !prev);
|
| 168 |
+
}
|
| 169 |
+
};
|
| 170 |
+
|
| 171 |
+
const initOutputAudioContext = () => {
|
| 172 |
if (!audioContextRef.current) {
|
| 173 |
// @ts-ignore
|
| 174 |
const AudioCtor = window.AudioContext || window.webkitAudioContext;
|
|
|
|
| 197 |
setTranscript('正在连接服务器...');
|
| 198 |
|
| 199 |
try {
|
| 200 |
+
initOutputAudioContext();
|
| 201 |
|
| 202 |
const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
|
| 203 |
// Append User Info to ensure backend creates a distinct session context
|
|
|
|
| 295 |
setStatus('CONNECTED');
|
| 296 |
}
|
| 297 |
|
| 298 |
+
isRecordingRef.current = true; // Mark as recording
|
| 299 |
+
|
| 300 |
const stream = await navigator.mediaDevices.getUserMedia({ audio: {
|
| 301 |
sampleRate: INPUT_SAMPLE_RATE,
|
| 302 |
channelCount: 1,
|
| 303 |
+
echoCancellation: true,
|
| 304 |
+
autoGainControl: true,
|
| 305 |
+
noiseSuppression: true
|
| 306 |
}});
|
| 307 |
mediaStreamRef.current = stream;
|
| 308 |
|
| 309 |
+
// Use a new context for input to ensure 16k rate if browser supports specific ctx rate
|
| 310 |
+
// @ts-ignore
|
| 311 |
+
const AudioCtor = window.AudioContext || window.webkitAudioContext;
|
| 312 |
+
const ctx = new AudioCtor({ sampleRate: INPUT_SAMPLE_RATE });
|
| 313 |
+
inputAudioContextRef.current = ctx;
|
| 314 |
+
|
| 315 |
const source = ctx.createMediaStreamSource(stream);
|
| 316 |
const processor = ctx.createScriptProcessor(4096, 1, 1);
|
| 317 |
|
| 318 |
processor.onaudioprocess = (e) => {
|
| 319 |
+
// Crucial check: If we decided to stop recording, do not process even if event fires
|
| 320 |
+
if (!isRecordingRef.current) return;
|
| 321 |
+
|
| 322 |
const inputData = e.inputBuffer.getChannelData(0);
|
| 323 |
const l = inputData.length;
|
| 324 |
const int16Data = new Int16Array(l);
|
|
|
|
| 352 |
|
| 353 |
} catch (e) {
|
| 354 |
console.error(e);
|
| 355 |
+
isRecordingRef.current = false;
|
| 356 |
setTranscript('无法访问麦克风');
|
| 357 |
}
|
| 358 |
};
|
| 359 |
|
| 360 |
const stopRecording = () => {
|
| 361 |
+
if (!isRecordingRef.current) return;
|
| 362 |
+
isRecordingRef.current = false; // Stop processing loop immediately
|
| 363 |
|
| 364 |
+
// Cleanup Mic Processing
|
| 365 |
if (processorRef.current) {
|
| 366 |
+
processorRef.current.onaudioprocess = null; // Detach handler
|
| 367 |
processorRef.current.disconnect();
|
| 368 |
processorRef.current = null;
|
| 369 |
}
|
|
|
|
| 375 |
mediaStreamRef.current.getTracks().forEach(t => t.stop());
|
| 376 |
mediaStreamRef.current = null;
|
| 377 |
}
|
| 378 |
+
// Force close input context to ensure hardware releases
|
| 379 |
+
if (inputAudioContextRef.current) {
|
| 380 |
+
inputAudioContextRef.current.close().catch(()=>{});
|
| 381 |
+
inputAudioContextRef.current = null;
|
| 382 |
+
}
|
| 383 |
|
| 384 |
setStatus('THINKING');
|
| 385 |
setTranscript('思考中...');
|
|
|
|
| 403 |
if (!api.auth.getCurrentUser()) return null;
|
| 404 |
|
| 405 |
return (
|
| 406 |
+
<div
|
| 407 |
+
ref={containerRef}
|
| 408 |
+
className={`fixed z-[9999] touch-none ${position ? '' : 'bottom-6 right-6'}`}
|
| 409 |
+
style={position ? { left: position.x, top: position.y } : undefined}
|
| 410 |
+
>
|
| 411 |
{!isOpen && (
|
| 412 |
+
<div
|
| 413 |
+
className="cursor-move"
|
| 414 |
+
onMouseDown={handleDragStart}
|
| 415 |
+
onTouchStart={handleDragStart}
|
| 416 |
>
|
| 417 |
+
<button
|
| 418 |
+
onClick={handleToggleOpen}
|
| 419 |
+
className="w-14 h-14 rounded-full bg-gradient-to-br from-indigo-600 to-purple-600 text-white shadow-2xl flex items-center justify-center hover:scale-110 transition-transform cursor-pointer border-2 border-white/20 animate-in zoom-in"
|
| 420 |
+
>
|
| 421 |
+
<Bot size={28} />
|
| 422 |
+
</button>
|
| 423 |
+
</div>
|
| 424 |
)}
|
| 425 |
|
| 426 |
{isOpen && (
|
| 427 |
<div className="bg-slate-900 w-80 md:w-96 rounded-3xl shadow-2xl border border-slate-700 overflow-hidden flex flex-col animate-in slide-in-from-bottom-5 fade-in duration-300 h-[500px]">
|
| 428 |
+
{/* Draggable Header */}
|
| 429 |
+
<div
|
| 430 |
+
className="bg-slate-800/50 p-4 flex justify-between items-center text-white shrink-0 backdrop-blur-md cursor-move select-none"
|
| 431 |
+
onMouseDown={handleDragStart}
|
| 432 |
+
onTouchStart={handleDragStart}
|
| 433 |
+
>
|
| 434 |
<div className="flex items-center gap-2">
|
| 435 |
<div className={`w-2 h-2 rounded-full ${status === 'DISCONNECTED' ? 'bg-red-500' : 'bg-green-500 animate-pulse'}`}></div>
|
| 436 |
<span className="font-bold text-sm">AI 实时通话 (代理模式)</span>
|
| 437 |
</div>
|
| 438 |
<div className="flex gap-2">
|
| 439 |
+
<button onClick={handleDisconnect} title="重置" className="hover:bg-white/10 p-1.5 rounded-full text-gray-400 hover:text-white transition-colors cursor-pointer" onMouseDown={e=>e.stopPropagation()}><RefreshCw size={16}/></button>
|
| 440 |
+
<button onClick={() => setIsOpen(false)} title="最小化" className="hover:bg-white/10 p-1.5 rounded-full text-gray-400 hover:text-white transition-colors cursor-pointer" onMouseDown={e=>e.stopPropagation()}><ChevronDown size={20}/></button>
|
| 441 |
</div>
|
| 442 |
</div>
|
| 443 |
|
|
|
|
| 504 |
<button
|
| 505 |
onMouseDown={startRecording}
|
| 506 |
onMouseUp={stopRecording}
|
| 507 |
+
onMouseLeave={stopRecording}
|
| 508 |
onTouchStart={(e) => { e.preventDefault(); startRecording(); }}
|
| 509 |
onTouchEnd={(e) => { e.preventDefault(); stopRecording(); }}
|
| 510 |
className={`w-20 h-20 rounded-full flex items-center justify-center shadow-lg transition-all transform ${
|