Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/mauroperez055/infoJobs/llms.txt

Use this file to discover all available pages before exploring further.

The InfoJobs DevBoard backend is an Express 5 REST API written with ES Modules. It follows a layered architecture where each concern lives in its own directory: routes declare URL patterns, controllers hold business logic, models handle data access, and Zod schemas validate every incoming payload. Two middleware layers — CORS and a per-route rate limiter — sit in front of the AI endpoint to control access and prevent abuse.

​
Directory structure

backend/
├── config.js           # Global constants and environment config
├── index.js            # Server entry point
├── jobs.json           # Flat JSON data store
├── routes/
│   ├── jobs.js         # CRUD endpoints for job listings
│   └── ai.js           # AI summary endpoint (Ollama)
├── controllers/
│   └── jobs.js         # Request/response logic for jobs
├── models/
│   └── job.js          # Data-access layer (reads/writes jobs.json)
├── schemas/
│   └── jobs.js         # Zod schemas: validateJob, validatePartialJob
├── middlewares/
│   └── cors.js         # Origin-allowlist CORS middleware factory
└── package.json

​
Layered architecture

Every HTTP request flows through the same pipeline before a response is returned: 
Incoming request
       │
       ▼
 Middleware layer
  (CORS, JSON body parser, rate limiter on /ai)
       │
       ▼
   Router layer
  (routes/jobs.js  ·  routes/ai.js)
       │
       ├── Zod validation middleware (POST / PATCH only)
       │
       ▼
 Controller layer
  (controllers/jobs.js)
       │
       ▼
   Model layer
  (models/job.js  →  jobs.json)
       │
       ▼
    Response

​
Configuration

All server-wide constants and environment-driven settings are exported from config.js: 
export const DEFAULTS = {
  LIMIT_PAGINATION: 10,
  LIMIT_OFFSET: 0,
  PORT: 1234
}

export const CONFIG = {
  MODEL_AI: process.env.MODEL_AI ?? 'mistral/devstral-small-2'
}
DEFAULTS.PORT is used by index.js as the fallback when process.env.PORT is not set. CONFIG.MODEL_AI is exported for future use but is currently not wired into the AI route — the /ai/summary/:id handler hard-codes model: 'qwen2.5:3b' directly in the ollama.chat() call.

​
Router modules

​
/jobs — job listing CRUD

Defined in routes/jobs.js, mounted at /jobs in index.js. Every write operation (POST and PATCH) passes through an inline Zod validation middleware before reaching the controller; invalid payloads receive a 400 response with structured error details before the controller is ever called.
MethodPathValidationController action
GET/jobs—JobController.getAll
GET/jobs/:id—JobController.getId
POST/jobsvalidateJob (full schema)JobController.create
PUT/jobs/:id—JobController.update
PATCH/jobs/:idvalidatePartialJob (partial schema)JobController.parcialUpdate
DELETE/jobs/:id—JobController.delete

​
/ai — AI job summaries

Defined in routes/ai.js, mounted at /ai in index.js. A express-rate-limit instance is applied as router-level middleware, limiting each IP to 5 requests per minute with standard RateLimit response headers (draft-8 spec). Requests that exceed the limit receive a 429 with a JSON error message. The single endpoint GET /ai/summary/:id works as follows:
1

Look up the job

JobModel.getById(id) reads the matching record from jobs.json. Returns 404 if not found.
2

Build the prompt

A multi-line Spanish-language prompt is assembled from the job’s titulo, empresa, ubicacion, and descripcion fields, instructing the model to produce a 4–6 sentence Spanish-language summary.
3

Stream from Ollama

ollama.chat() is called with model: 'qwen2.5:3b' and stream: true. Response headers are set to Content-Type: text/plain; charset=utf-8 and Transfer-Encoding: chunked.
4

Forward chunks to the client

Each chunk from the async iterator is written directly to the response with res.write(content). The response is finalised with res.end() once the stream is exhausted.
5

Handle errors gracefully

If an error occurs before any headers are sent, the handler sends a 500 JSON response. If headers have already been flushed (mid-stream), it calls res.end() to close the connection cleanly.

​
CORS middleware

middlewares/cors.js exports a corsMiddleware factory that wraps the cors npm package with an explicit origin allowlist: 
import cors from 'cors';

const ACCEPTED_ORIGINS = [
  'http://localhost:5173'
]

export const corsMiddleware = ({ acceptedOrigins = ACCEPTED_ORIGINS } = {}) => {
  return cors({
    origin: (origin, callback) => {
      if (!origin) {
        return callback(null, true);
      }

      if (acceptedOrigins.includes(origin)) {
        return callback(null, true);
      }
      console.log(origin)
      return callback(new Error('Origen no permitido'));
    }
  })
}
Requests with no Origin header (e.g. server-to-server or same-origin calls) are allowed through unconditionally. Browser requests from any origin not in ACCEPTED_ORIGINS are rejected with an error. In the current index.js the factory is commented out in favour of cors() with open defaults — use corsMiddleware() to lock down the allowed origins in production.

​
Running the server

# Development — restarts automatically on file changes (nodemon)
npm run dev

# Production — runs directly with Node.js
npm start
The server logs Servidor levantado en http://localhost:1234 on startup. The PORT environment variable overrides the default. The server also sets app.set('trust proxy', 1) so that the rate limiter reads the real client IP when the app is running behind a reverse proxy such as Nginx or Cloudflare.
Job data is loaded from jobs.json into memory when the server starts. Any jobs created or modified via the API are written back to the file, but the in-process rate-limit counters (for /ai) are not persisted — they reset on every restart. There is no database: if jobs.json is deleted or corrupted, all job data is lost. Back up the file before deploying changes to a shared environment.

Build docs developers (and LLMs) love

Get started for freeTalk to us