ISIS MCP

import { getLlama, LlamaChatSession, Llama, LlamaModel, LlamaContext } from 'node-llama-cpp'; import { existsSync, mkdirSync, createWriteStream } from 'fs'; import { homedir } from 'os'; import { join } from 'path'; interface LlamaEngineConfig { modelName?: string; cacheDir?: string; maxTokens?: number; } export class LlamaEngine { private static instance: LlamaEngine; private llama: Llama | null = null; private model: LlamaModel | null = null; private modelPath: string; private cacheDir: string; private maxTokens: number; private modelName: string; private isDownloading: boolean = false; private constructor(config: LlamaEngineConfig = {}) { this.modelName = config.modelName || 'Llama-3.2-1B-Instruct-Q4_K_M.gguf'; this.cacheDir = config.cacheDir || join(homedir(), '.cache', 'isis-mcp'); this.modelPath = join(this.cacheDir, this.modelName); this.maxTokens = config.maxTokens || 250; if (!existsSync(this.cacheDir)) { mkdirSync(this.cacheDir, { recursive: true }); } } static getInstance(config?: LlamaEngineConfig): LlamaEngine { if (!LlamaEngine.instance) { LlamaEngine.instance = new LlamaEngine(config); } return LlamaEngine.instance; } private async downloadModel(): Promise<void> { if (this.isDownloading) { console.log('⏳ Model download already in progress...'); return; } this.isDownloading = true; console.log(`📥 Downloading ${this.modelName}...`); try { const modelUrl = `https://huggingface.co/lmstudio-community/Llama-3.2-1B-Instruct-GGUF/resolve/main/${this.modelName}`; const response = await fetch(modelUrl); if (!response.ok) { throw new Error(`Failed to download: ${response.statusText}`); } const totalSize = parseInt(response.headers.get('content-length') || '0'); let downloaded = 0; let lastPercent = 0; const body = response.body; if (!body) throw new Error('No response body'); const reader = body.getReader(); const writer = createWriteStream(this.modelPath); while (true) { const { done, value } = await reader.read(); if (done) break; writer.write(Buffer.from(value)); downloaded += value.length; const percent = Math.floor((downloaded / totalSize) * 100); if (percent > lastPercent && percent % 10 === 0) { console.log(`📥 Preparando sumarização inteligente... ${percent}%`); lastPercent = percent; } } writer.end(); console.log(`✅ Modelo ${this.modelName} pronto! Sumarização inteligente ativada.`); } catch (error) { console.error(`❌ Failed to download model: ${error}`); throw error; } finally { this.isDownloading = false; } } private async ensureModel(): Promise<void> { if (!existsSync(this.modelPath)) { await this.downloadModel(); } } private async loadModel(): Promise<LlamaModel> { if (this.model) { return this.model; } await this.ensureModel(); console.log('🤖 Loading model into memory...'); if (!this.llama) { this.llama = await getLlama(); } this.model = await this.llama.loadModel({ modelPath: this.modelPath, gpuLayers: 'auto', }); return this.model; } async summarize(content: string): Promise<string | null> { try { const model = await this.loadModel(); const context = await model.createContext(); const sequence = context.getSequence(); const session = new LlamaChatSession({ contextSequence: sequence, }); const prompt = `Summarize the following content concisely in 150-200 words, preserving the key facts and main insights:\n\n${content.substring(0, 8000)}\n\nSummary:`; const response = await session.prompt(prompt, { maxTokens: this.maxTokens, temperature: 0.3, }); await context.dispose(); return response.trim(); } catch (error) { console.error('LlamaEngine summarization error:', error); return null; } } async isAvailable(): Promise<boolean> { try { if (existsSync(this.modelPath)) { return true; } return true; } catch { return false; } } async cleanup(): Promise<void> { if (this.model) { await this.model.dispose(); this.model = null; } if (this.llama) { await this.llama.dispose(); this.llama = null; } } }