✨ See it live

⚙️ How it works + pricing

📱 Mobile view

Introduction

Vitex is an AI-powered resume generation platform that transforms a job description and your professional background into a polished, ATS-optimized resume PDF and cover letter. An 8-step AI pipeline handles everything from JD parsing and skill matching to server-side PDF compilation, with Typst compiling PDFs locally in under 100ms. It is agent-ready: every capability is reachable over an authenticated HTTP API (the web UI and the public v1 API share one pipeline core), and billing is outcome-based — you are charged only when a resume is successfully produced. The UI follows the Phantom design system: a soft, flat aubergine-and-lavender aesthetic with pill geometry, whisper-weight typography, and generous whitespace.

Vitex is built for a person plus their AI assistant on a "one core, N thin adapters" design — "The API is the UI". Everything the web app does, an agent can do over the public v1 HTTP API, the published vitex-cli, or the hosted MCP connector. The guiding framing is Career as Code: your career facts are the source, each tailored PDF is a reproducible build artifact, the refinement chain is a series of commits, outcome billing means you pay per successful build, and the exported .typ source means zero lock-in. See ADR 0003.

Key Features

• AI Resume Generation -- Paste a job description + describe your background, get a tailored resume PDF
• 8-Step AI Pipeline -- JD parsing, background parsing, match analysis, tailoring, ATS scoring, cover letter, document generation, server-side PDF compilation
• Agent-Ready API -- Operate the whole product over HTTP with an API key; no browser, 2FA, or CAPTCHA (see docs/api/v1.md, the OpenAPI spec at public/openapi.yaml, and the runnable curl playbook at public/skill.md)
• CLI & MCP server -- vitex-cli (bin vitex) is a thin client over the hosted API: a token-cheap CLI for terminals and coding agents, plus vitex mcp (a stdio MCP server via npx -y vitex-cli mcp) for Claude Desktop, Claude Code, and Cursor
• Hosted MCP connector -- a browser-OAuth remote MCP server at https://www.vitex.org.nz/api/mcp lets non-technical users connect Vitex inside ChatGPT or Claude with a sign-in (no API key to paste): ChatGPT guide, Claude guide. For agent-driven setup from docs alone, see llms-install.md
• Outcome-Based Billing -- Credits are charged only when a resume is successfully produced (failures are free), idempotently
• 7 Professional Templates -- Auto-selected by AI based on industry and role (two-column, modern-cv, executive, creative, compact, banking, academic)
• ATS Optimization -- Real-time ATS compatibility scoring with actionable feedback
• Cover Letter Generation -- Automatically generated alongside the resume
• Typst-Powered PDF -- Local compilation in <100ms, no external APIs required
• Natural Language Refinement -- Describe changes in plain English; AI applies a targeted, free refinement (resume and/or cover letter) — not a full regeneration
• Voice Profile -- Save a writing sample on a candidate profile so generated cover letters match your voice
• Public Career Endpoint -- Publish a profile to a stable, agent-readable page at /p/<slug> (HTML, /json, and /md); contact PII and raw text are never exposed
• Credit System -- Stripe-powered subscription plans for usage management
• Cloud Storage -- Persistent resume management with user accounts (My Resumes history, search, and re-open)

How It Works

1. Input -- Paste a job description and describe your professional background on the homepage
2. Generate -- AI analyzes, matches, tailors, and compiles your resume through an 8-step pipeline streamed via SSE
3. Refine & Export -- Review the PDF preview, ATS score, and cover letter; refine with natural language; download or share

Tech Stack

Getting Started

Prerequisites

• Node.js 18+
• Docker (for production deployment)
• Typst binary (for local PDF compilation)

Installation

git clone https://github.com/ChanMeng666/easy-resume.git
cd easy-resume
npm install

Environment Variables

Create a .env.local file:

# AI
OPENAI_API_KEY=sk-...

# Database
DATABASE_URL=postgresql://user:password@host:5432/dbname

# Neon Auth (Stack Auth)
NEXT_PUBLIC_STACK_PROJECT_ID=...
NEXT_PUBLIC_STACK_PUBLISHABLE_CLIENT_KEY=...
STACK_SECRET_SERVER_KEY=...

# Payments (Stripe)
STRIPE_SECRET_KEY=...
STRIPE_WEBHOOK_SECRET=...
STRIPE_PRICE_CREDITS_5=...
STRIPE_PRICE_PRO_MONTHLY=...
STRIPE_PRICE_UNLIMITED_MONTHLY=...

# Canonical app origin (drives the OAuth issuer + advertised OAuth/MCP/public URLs)
NEXT_PUBLIC_APP_URL=http://localhost:3000   # prod: https://www.vitex.org.nz

# Optional: model tiering + observability
AI_MODEL_EXTRACT=gpt-5.4-mini-2026-03-17   # JD parse
AI_MODEL_REASON=gpt-5.5-2026-04-23         # generation / quality-critical steps
AI_MODEL_CHAT=gpt-5.5-2026-04-23           # edit-agent tool loop (same as reason)
AI_TELEMETRY_ENABLED=false

See .env.example for the full annotated list. Rate limiting and webhook idempotency use the existing Neon Postgres — no Redis required.

Development

npm run dev          # Start dev server on http://localhost:3000
npm run build        # Production build
npm run start        # Start production server
npm run lint         # Run ESLint

Deployment

Vitex runs as a Docker container on a VPS behind Traefik reverse proxy, with GitHub Actions for CI/CD.

# Build Docker image
docker build -t vitex .

# Run container
docker run -p 3000:3000 --env-file .env vitex

The production deployment uses:

• DigitalOcean VPS as the host
• Docker for containerization
• Traefik for reverse proxy and TLS
• GitHub Actions for automated build and deploy on push

Architecture

Monolith Next.js App (Docker Container on VPS)
|
|-- Pages
|   |-- /                    Landing page: JD + background input
|   |-- /editor              Result review: PDF preview, ATS score, cover letter, refinement
|   |-- /resumes             My Resumes history (search, open, download, delete)
|   |-- /profiles            Candidate profiles (background + voice sample) + publish
|   |-- /applications        Application tracker
|   |-- /dashboard           Credits, billing + Connections & API Keys
|   |-- /pricing             Subscription plans
|   |-- /p/[slug]            Public career endpoint (HTML, +/json, +/md)
|
|-- Transports (thin adapters over one shared core -- "the API is the UI")
|   |-- POST /api/generate     SSE stream for the web UI
|   |-- /api/v1/resumes        Public agent REST API (API key, job-based) + /api/v1/me
|   |-- /api/mcp               Hosted remote MCP (Streamable HTTP, OAuth-protected)
|   |-- /api/oauth/*           OAuth 2.1 Authorization Server (facade minting API keys)
|   |-- vitex-cli              Published npm package: CLI + stdio MCP server
|
|-- Backend Core (src/server/, transport-agnostic)
|   |-- 8-Step Pipeline: JD parse || background parse -> match -> tailor
|   |   -> ATS score || cover letter -> render (Typst) -> compile (PDF)
|   |-- Outcome billing: charge once, only on a compiled PDF, idempotent
|   |-- Auth: getCaller() resolves API key or Neon Auth cookie session
|   |-- Errors: machine-readable envelope; structured JSON logs
|
|-- Services
|   |-- Neon PostgreSQL (Drizzle ORM) -- data, credits, rate limits, job queue
|   |-- Neon Auth / Stack Auth (authentication)
|   |-- Stripe (payments)
|   |-- OpenAI (3 tiers: gpt-5.4-mini extract / gpt-5.5 reason / gpt-5.5 chat)

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

Chan Meng

• LinkedIn: chanmeng666
• GitHub: ChanMeng666
• Email: chanmeng.dev@gmail.com
• Website: chanmeng.org

---

---

---