File size: 8,935 Bytes
1d3f990 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 | /**
* WebLLM Engine
* Browser-native model inference with WebGPU acceleration
*/
class WebLLMEngine {
constructor(options = {}) {
this.status = 'OFFLINE';
this.model = options.model || 'TinyLlama-1.1B-Chat-v0.4-q4f16_1-1k';
this.temperature = options.temperature || 0.2;
this.topP = options.topP || 0.9;
this.maxTokens = options.maxTokens || 256;
this.engine = null;
this.isLoading = false;
this.isGenerating = false;
this.abortController = null;
this.supportedModels = [
{ id: 'TinyLlama-1.1B-Chat-v0.4-q4f16_1-1k', name: 'TinyLlama 1.1B' },
{ id: 'Mistral-7B-Instruct-v0.2-q4f16_1-MLC', name: 'Mistral 7B' },
{ id: 'Llama-2-7b-chat-hf-q4f32_1-MLC', name: 'Llama 2 7B' },
];
this.conversationHistory = [];
this.maxHistoryLength = 10;
}
/**
* Initialize WebLLM engine
*/
async initialize() {
if (this.status === 'READY' || this.isLoading) {
return;
}
this.isLoading = true;
this.status = 'LOADING';
this.emit('statusChanged', 'LOADING');
try {
// Check WebLLM availability
if (typeof window.webllm === 'undefined') {
throw new Error('WebLLM not loaded. Include @mlc-ai/web-llm script.');
}
// Detect WebGPU support
const hasWebGPU = !!(navigator.gpu);
console.log(`WebGPU available: ${hasWebGPU}`);
// Initialize engine
const webllm = window.webllm;
this.engine = new webllm.Engine({
model: this.model,
useWebGPU: hasWebGPU,
maxSequenceLength: 4096,
});
await this.engine.forward('');
this.status = 'READY';
this.isLoading = false;
this.emit('statusChanged', 'READY');
console.log(`WebLLM engine initialized: ${this.model}`);
} catch (error) {
this.status = 'ERROR';
this.isLoading = false;
this.emit('statusChanged', 'ERROR');
this.emit('error', error.message);
console.error('WebLLM initialization failed:', error);
throw error;
}
}
/**
* Generate response with streaming
*/
async generateResponse(userMessage, systemPrompt = '') {
if (this.status !== 'READY' || !this.engine) {
throw new Error('Engine not ready');
}
if (this.isGenerating) {
throw new Error('Generation already in progress');
}
this.isGenerating = true;
this.status = 'GENERATING';
this.emit('statusChanged', 'GENERATING');
this.emit('generationStart');
this.abortController = new AbortController();
let fullResponse = '';
try {
// Build conversation
const messages = this.buildConversation(userMessage, systemPrompt);
// Stream generation
const generator = await this.engine.generate(messages, {
temperature: this.temperature,
top_p: this.topP,
max_new_tokens: this.maxTokens,
});
for await (const token of generator) {
if (this.abortController.signal.aborted) {
break;
}
fullResponse += token;
this.emit('token', token);
}
// Add to conversation history
this.addToHistory({
role: 'user',
content: userMessage,
});
this.addToHistory({
role: 'assistant',
content: fullResponse,
});
this.status = 'READY';
this.emit('statusChanged', 'READY');
this.emit('generationComplete', fullResponse);
return fullResponse;
} catch (error) {
if (error.name !== 'AbortError') {
this.status = 'ERROR';
this.emit('statusChanged', 'ERROR');
this.emit('error', error.message);
console.error('Generation failed:', error);
throw error;
}
} finally {
this.isGenerating = false;
}
}
/**
* Build conversation history for API
*/
buildConversation(userMessage, systemPrompt) {
const messages = [];
if (systemPrompt) {
messages.push({
role: 'system',
content: systemPrompt,
});
}
// Add conversation history (bounded)
for (const msg of this.conversationHistory.slice(-this.maxHistoryLength)) {
messages.push(msg);
}
// Add current message
messages.push({
role: 'user',
content: userMessage,
});
return messages;
}
/**
* Add message to history
*/
addToHistory(message) {
this.conversationHistory.push(message);
// Trim history to max length
if (this.conversationHistory.length > this.maxHistoryLength * 2) {
this.conversationHistory = this.conversationHistory.slice(-this.maxHistoryLength);
}
}
/**
* Stop current generation
*/
stopGeneration() {
if (this.abortController) {
this.abortController.abort();
this.isGenerating = false;
this.status = 'READY';
this.emit('statusChanged', 'READY');
this.emit('generationStopped');
}
}
/**
* Clear conversation history
*/
clearHistory() {
this.conversationHistory = [];
this.emit('historyCleared');
}
/**
* Change temperature
*/
setTemperature(temp) {
this.temperature = Math.max(0, Math.min(2, temp));
}
/**
* Change top-p
*/
setTopP(p) {
this.topP = Math.max(0, Math.min(1, p));
}
/**
* Change max tokens
*/
setMaxTokens(tokens) {
this.maxTokens = Math.max(1, Math.min(2048, tokens));
}
/**
* Get current settings
*/
getSettings() {
return {
model: this.model,
temperature: this.temperature,
topP: this.topP,
maxTokens: this.maxTokens,
status: this.status,
};
}
/**
* Get supported models
*/
getSupportedModels() {
return this.supportedModels;
}
/**
* Switch model
*/
async switchModel(modelId) {
if (this.isGenerating) {
throw new Error('Cannot switch model while generating');
}
const supported = this.supportedModels.some(m => m.id === modelId);
if (!supported) {
throw new Error(`Model not supported: ${modelId}`);
}
this.model = modelId;
this.status = 'OFFLINE';
// Reset engine to force re-initialization
this.engine = null;
this.emit('statusChanged', 'OFFLINE');
}
/**
* Event emitter
*/
listeners = {};
on(event, callback) {
if (!this.listeners[event]) {
this.listeners[event] = [];
}
this.listeners[event].push(callback);
}
off(event, callback) {
if (this.listeners[event]) {
this.listeners[event] = this.listeners[event].filter(cb => cb !== callback);
}
}
emit(event, data) {
if (this.listeners[event]) {
for (const callback of this.listeners[event]) {
try {
callback(data);
} catch (error) {
console.error(`Error in listener for ${event}:`, error);
}
}
}
}
/**
* Check WebGPU support
*/
static hasWebGPU() {
return !!navigator.gpu;
}
/**
* Check WebLLM availability
*/
static hasWebLLM() {
return typeof window.webllm !== 'undefined';
}
/**
* Estimate download size
*/
estimateModelSize(modelId) {
const sizes = {
'TinyLlama-1.1B-Chat-v0.4-q4f16_1-1k': '600MB',
'Mistral-7B-Instruct-v0.2-q4f16_1-MLC': '4GB',
'Llama-2-7b-chat-hf-q4f32_1-MLC': '8GB',
};
return sizes[modelId] || 'Unknown';
}
}
// Export for use in other modules
if (typeof module !== 'undefined' && module.exports) {
module.exports = WebLLMEngine;
}
|