iMery Backend Development Patterns

Express.js API Structure

iMery uses a centralized index.js for routing.

// ✅ Route grouping in index.js
app.post('/auth/login', async (req, res) => { ... });
app.get('/posts', async (req, res) => { ... });
app.post('/posts', upload.single('image'), async (req, res) => { ... });

Database Patterns (MySQL/TiDB)

We use mysql2 with db.js for initialization and schema management.

Prepared Statements

ALWAYS use ? placeholders to prevent SQL injection.

// ✅ GOOD
const [results] = await db.execute("SELECT * FROM Posts WHERE user_id = ?", [
  userId,
]);

// ❌ BAD
const [results] = await db.execute(
  `SELECT * FROM Posts WHERE user_id = ${userId}`,
);

Schema Initialization (db.js)

The database is automatically initialized on server start.

const schema = `
CREATE TABLE IF NOT EXISTS Users ( ... );
CREATE TABLE IF NOT EXISTS Posts ( ... );
`;
await connection.query(schema);

Authentication (JWT & bcryptjs)

• •bcryptjs: Used for hashing passwords before saving to DB.
• •jsonwebtoken: Used for generating tokens upon login.

const token = jwt.sign(
  { id: user.id, email: user.email },
  process.env.JWT_SECRET,
  { expiresIn: "24h" },
);

AI Integration Workflow

RunPod & Gemini Integration

The /analyze/:id endpoint orchestrates multiple external APIs.

1. •Input: Image URL from S3.
2. •Step 1: Call RunPod for style identification.
3. •Step 2: Call Gemini for AI summary and audio prompt generation.
4. •Step 3: Update the Posts table with analysis results.

File Uploads (S3)

We use multer-s3 to upload directly to AWS.

const upload = multer({
  storage: multerS3({
    s3: s3Client,
    bucket: process.env.AWS_S3_BUCKET,
    key: (req, file, cb) => {
      cb(null, `${Date.now()}-${file.originalname}`);
    },
  }),
});