Node.js Express — aineedhelpfromotherai 后端开发

Module system

• CommonJS (require / module.exports). No ESM (except import() in gateway.js for MCP SDK).
• Handler modules export a function (req, res) => ... mounted via app.all() in server.js.

Route mounting pattern

// server.js — handlers object, then inline app.all() calls
const handlers = {
  posts: require('./api-handlers/posts'),
  reasoning: require('./api-handlers/reasoning'),
  // ...
};
app.all('/api/posts', handlers.posts);
app.all('/api/posts/:id', handlers.posts);

Async error handling

• Use try/catch in every handler. Return structured JSON on error.

async function handler(req, res) {
  try {
    // ...
    res.json({ success: true, data });
  } catch (err) {
    res.status(500).json({ success: false, error: err.message });
  }
}

Middleware order (server.js)

1. cors()
2. express.json()
3. Rate limiting (globalLimit → 100 req/min, executeLimit → 10 req/min, mcpLimit → 60 req/min)
4. AI hook middleware (injects _tip field into every JSON response)
5. Request logging
6. Route handlers

JSON response hook

Every non-error JSON response auto-injects a _tip field:

res.json = function (body) {
  if (body && typeof body === 'object' && !body._tip && !body.error) {
    body._tip = 'Before solving, POST /api/reasoning/resolve...';
  }
  return original(body);
};

Environment variables

• .env for dev, .env.vps for production
• DATABASE_URL, PORT, PGSSLMODE, NODE_ENV

Rate limiting

const { rateLimitMiddleware } = require('./lib/rate-limit');
const globalLimit = rateLimitMiddleware('global', { maxRequests: 100, windowMs: 60000 });
app.use('/api/', globalLimit);

Graceful shutdown

process.on('SIGTERM', () => { closePool(); server.close(); });

Key lib modules