CVConnect

CVConnect is a resume optimisation workspace: upload a PDF/DOCX resume, analyse a job description, inspect the match score and skill gaps, then review AI-proposed edits before exporting.

Quick start

1. Copy .env.example to .env and set the two JWT secrets. Add an OpenAI API key to enable AI rewriting (the app has a safe local fallback for development).
2. Start the stack: docker compose up --build.
3. In a second terminal run docker compose exec backend npx prisma migrate deploy.
4. Visit http://localhost:5173; the API health check is at http://localhost:5000/health.

For local development, run npm install in backend and frontend, then npm run dev in both folders. In ml-service, create a virtual environment and run pip install -r requirements.txt && python app.py.

Architecture

• frontend/ — React, Redux Toolkit, Tailwind, Recharts, and accessible editor controls.
• backend/ — Express API, Prisma/Postgres persistence, Redis caching, JWT auth, Socket.IO progress events and S3-ready upload abstraction.
• ml-service/ — FastAPI service for skill extraction and embedding similarity; it degrades to a lightweight lexical model when large ML models are unavailable.

API

All API routes below require Authorization: Bearer <access token> except POST /api/auth/register, POST /api/auth/login, and GET /health.

Method	Route	Purpose
POST	/api/resumes/upload	Upload a PDF/DOCX (10 MB maximum) and parse it
POST	/api/jobs/analyze	Extract job skills and requirements
POST	/api/resumes/:id/match	Calculate weighted matching score
POST	/api/resumes/:id/rewrite	Generate tracked, honest optimisation proposals
GET	/api/resumes	Paginated resume history
GET/PUT	/api/users/me	Read/update profile

Production notes

Use managed Postgres, Redis, object storage, a secret manager and an HTTPS reverse proxy in production. Set a restrictive CLIENT_ORIGIN, rotate JWT secrets, apply migrations in CI/CD before rollout, and configure Sentry/Prometheus endpoints through your platform. Setting the AWS_S3_* variables stores source documents with server-side encryption; leaving them empty keeps local development storage-free. The included GitHub Actions workflow validates backend and frontend on each push.