Resume Refresh

Paste your resume, describe what you're targeting, and get a polished, ATS-ready draft — section by section.

Live demo: resume-refresh-ten.vercel.app

---

What it does

• Accepts a resume as plain text, PDF, TXT, or MD
• Optionally paste your LinkedIn headline/about/experience for added context
• Returns bullet-by-bullet improvement suggestions plus a full rewrite
• Exports to PDF or DOCX with right-aligned job dates and clean section formatting
• Works without an API key (rule-based suggestions); plug in OpenAI or a local Ollama model to unlock AI rewrite

Run locally

npm install
cp .env.example .env   # fill in at minimum APP_SECRET
npm start

Open http://localhost:3210. Node 20+ required.

Environment variables

Variable	Required	Description
APP_SECRET	Yes	Random secret for session signing. Generate: node -e "console.log(require('crypto').randomBytes(32).toString('hex'))"
OPENAI_API_KEY	No	Enables AI-powered rewrite via OpenAI (gpt-4.1-mini)
INFERENCE_PROVIDER	No	Set to ollama to use a local model instead of OpenAI
OLLAMA_URL	No	Ollama base URL (default: http://127.0.0.1:11434)
OLLAMA_MODEL	No	Model name (default: qwen2.5:7b)
LINKEDIN_CLIENT_ID	No	Only needed if you want LinkedIn sign-in
LINKEDIN_CLIENT_SECRET	No	LinkedIn OAuth
PUBLIC_BASE_URL	No	Full origin URL used for OAuth redirects (e.g. https://your-domain.com)

See .env.example for the full list with comments.

Using a local model (Ollama)

brew install ollama
ollama pull qwen2.5:7b
ollama serve

Then in .env:

INFERENCE_PROVIDER=ollama

Works well for bullet strengthening and wording cleanup. For full resume rewrites, a cloud model gives better results.

Deploy to Vercel

vercel

Set the environment variables above in the Vercel dashboard. The vercel.json at the repo root handles routing and security headers.

Deploy elsewhere

Any Node 20+ host works. Run node src/server.js and set the environment variables.

Tested on Render, Railway, Fly.io, and plain Linux VPS.

Tests

npm test          # unit tests (Node built-in test runner)
npm run test:e2e  # Playwright end-to-end tests

Project structure

api/        Vercel serverless function handlers
src/        Core logic (app router, inference, resume analysis, export)
public/     Static frontend (HTML, CSS, JS)

Security

• Sessions signed with APP_SECRET
• Rate-limited per IP: 20 analyze / 8 rewrite / 12 export requests per minute
• Security headers on all responses (CSP, X-Frame-Options, Referrer-Policy, Permissions-Policy, X-Content-Type-Options)
• Request body and file upload size limits enforced
• No resume data stored server-side

LinkedIn sign-in (optional, not surfaced in the default UI)

The server includes a LinkedIn OAuth flow, but the default UI uses paste-only for LinkedIn content. To enable sign-in:

1. Create a LinkedIn app and enable Sign In with LinkedIn using OpenID Connect
2. Fill in LINKEDIN_CLIENT_ID, LINKEDIN_CLIENT_SECRET, and PUBLIC_BASE_URL in .env
3. Add your redirect URL to the LinkedIn app:
   • Production: https://your-domain.com/api/auth/linkedin/callback
   • Local: http://127.0.0.1:3210/api/auth/linkedin/callback

Note: LinkedIn's standard OAuth only provides basic identity (name, email, photo). Full profile data like experience and skills history is not available via API — paste that content directly into the LinkedIn field instead.

License

MIT