Upload folder using huggingface_hub

Dockerfile

@@ -1,35 +1,34 @@
 FROM node:18
 
-# Cài đặt đầy đủ công cụ build
+# Cài đặt tools
 RUN apt-get update && apt-get install -y \
     python3 \
-    python3-pip \
     make \
     g++ \
     cmake \
-    curl \
     build-essential \
     && rm -rf /var/lib/apt/lists/*
 
+# Tạo user để chạy non-root
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH
+
 WORKDIR /app
 
-# Copy package files trước
-COPY package*.json ./
+# Copy files với quyền user
+COPY --chown=user package*.json ./
 
-# Cài đặt dependencies với cờ --unsafe-perm để tránh lỗi quyền root
-# Bỏ qua script postinstall lúc đầu để tránh build lỗi ngay
 RUN npm install --ignore-scripts
 
-# Copy toàn bộ code
-COPY . .
+COPY --chown=user . .
 
-# Tạo thư mục models
+# Tạo thư mục models và tải model
 RUN mkdir -p models
-
-# Tải model (dùng npx từ node_modules local)
 RUN npx --no node-llama-cpp pull --dir ./models hf:Qwen/Qwen3-1.7B-GGUF:Q8_0 --filename Qwen3-1.7B-Q8_0.gguf
 
-# Rebuild node-llama-cpp một cách thủ công để đảm bảo nó nhận diện đúng môi trường Linux
+# Rebuild
 RUN npm rebuild node-llama-cpp
 
 EXPOSE 7860

server.js

@@ -1,11 +1,10 @@
 import express from 'express';
-import { SystemMessage, HumanMessage, Runnable, LlamaCppLLM } from '../src/index.js';
+import { SystemMessage, HumanMessage, Runnable, LlamaCppLLM } from './src/index.js'; // Sửa lại đường dẫn import cho đúng cấu trúc flat nếu cần, nhưng ở đây src vẫn nằm trong root
 import bodyParser from 'body-parser';
+import path from 'path';
+import fs from 'fs';
 
-// ============================================================================ 
-// COPIED LOGIC FROM SOLUTION (Simplification for API)
-// ============================================================================ 
-
+// ... (Giữ nguyên phần Class Logic) ...
 class EmailClassifierRunnable extends Runnable {
     constructor(llm) {
         super();
@@ -22,7 +21,9 @@ class EmailClassifierRunnable extends Runnable {
         return [
             new SystemMessage(`You are an email classification assistant. Classify into: Spam, Invoice, Meeting Request, Urgent, Personal, Other.
             Respond in strict JSON format: {"category": "Name", "confidence": 0.95, "reason": "Explanation"}`),
-            new HumanMessage(`Classify this email:\nSubject: ${email.subject}\nBody: ${email.body}`)
+            new HumanMessage(`Classify this email:
+Subject: ${email.subject}
+Body: ${email.body}`)
         ];
     }
 
@@ -37,32 +38,44 @@ class EmailClassifierRunnable extends Runnable {
     }
 }
 
-// ============================================================================ 
-// SERVER SETUP
-// ============================================================================ 
-
 const app = express();
-const PORT = 7860; // Standard port for Hugging Face Spaces
+const PORT = 7860;
 
 app.use(bodyParser.json());
 
-// Initialize LLM Global
-console.log("Loading model...");
-const llm = new LlamaCppLLM({
-    modelPath: './models/Qwen3-1.7B-Q8_0.gguf',
-    temperature: 0.1,
-    maxTokens: 200
-});
-const classifier = new EmailClassifierRunnable(llm);
-console.log("Model loaded!");
+// Log directory content for debugging
+console.log("Current directory:", process.cwd());
+console.log("Files in models:", fs.readdirSync('./models'));
+
+let classifier = null;
+
+async function initModel() {
+    try {
+        console.log("Loading model...");
+        const modelPath = path.resolve('./models/Qwen3-1.7B-Q8_0.gguf');
+        console.log("Model absolute path:", modelPath);
+
+        const llm = new LlamaCppLLM({
+            modelPath: modelPath,
+            temperature: 0.1,
+            maxTokens: 200
+        });
+        
+        // Test invoke to make sure it works
+        await llm.invoke("Hello"); 
+        
+        classifier = new EmailClassifierRunnable(llm);
+        console.log("Model loaded and ready!");
+    } catch (err) {
+        console.error("FATAL ERROR loading model:", err);
+        process.exit(1); // Exit to let Docker restart (or show error)
+    }
+}
 
 app.post('/classify', async (req, res) => {
+    if (!classifier) return res.status(503).json({ error: "Model is loading..." });
     try {
         const { subject, body, from } = req.body;
-        if (!subject || !body) {
-            return res.status(400).json({ error: 'Missing subject or body' });
-        }
-
         const result = await classifier.invoke({ subject, body, from: from || 'unknown' });
         res.json(result);
     } catch (error) {
@@ -75,6 +88,8 @@ app.get('/', (req, res) => {
     res.send('AI Agent Email Classifier is Running. POST to /classify to use.');
 });
 
-app.listen(PORT, () => {
+// Start server immediately, load model in background
+app.listen(PORT, '0.0.0.0', () => {
     console.log(`Server running on port ${PORT}`);
+    initModel();
 });