An AI-powered resume builder that generates and compiles professional LaTeX resumes from structured user input entirely in the browser. Both the compiled resume in pdf and latex can be downloaded
Making a professional resume for your profile is always complicated task, not because of the latex code itself but also in finding the right template, so for solving that purpose I created a Resume.AI. An AI powered website which generates latex code using groq llm, compiles the latex code using a safe e2b sandbox environment and displays the pdf right in your browser. A mix of simple form builder and overleaf.
On the first day of release it got 10+ active users.
Homepage
Dashboard — browse & create projects
Resume form editor
| Tool | Purpose |
|---|---|
| Next.js 15 (App Router) | Frontend + API routes |
| Clerk | Authentication |
| Inngest | Background job orchestration |
| Groq + LangChain | LLM-based LaTeX generation |
| E2B Code Interpreter | Remote sandboxed PDF compilation |
| Neon Postgres + Prisma | Database and ORM |
| React Query | Client-side data fetching and caching |
- Node.js 18+
- A Neon Postgres database
- A Clerk application
- A Groq API key
- An E2B API key
- An Inngest account (or local dev server)
git clone https://github.com/distroaryan/resume.ai.git
cd resume.ai
npm installCreate a .env.local file in the project root and fill these variables with your own API Keys
# Database
DATABASE_URL=
# Clerk
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=
CLERK_SECRET_KEY=
# Groq
GROQ_API_KEY=
# E2B
E2B_API_KEY=
# Inngest (required for local configuration of Inngest environment)
INNGEST_DEV=1npx prisma generate
npx prisma db push# Start the Next.js dev server
npm run dev
# In a separate terminal, start the Inngest dev server
npx inngest-cli@latest devThe app will be available at http://localhost:3000.
The core workflow is fully asynchronous:
User submits resume form
│
▼
generateProjectAndCompileAction (Server Action)
│
├── Inngest Step 1: Groq rewrites LaTeX → saved to DB
└── Inngest Step 2: E2B compiles LaTeX → PDF saved to DB
│
▼
Client polls checkProjectStatus → PDF rendered in editor
- New projects are seeded from pre-built templates on disk (
/templates/{name}/main.tex+main.pdf). - Authentication is handled entirely by Clerk — no user data is stored in the application database.
- Project state is tracked via a
statusfield (pending → Processing → Completed / Failed). - Backend logic lives in Next.js Server Actions (
src/action.ts) — no traditional REST API routes exceptGET /api/get-project-pdffor binary PDF serving.
For full architecture documentation and data flow diagrams, see Architecture.
For open issues and contribution opportunities, see Contribution.
The project uses Vitest for unit testing. All external dependencies (Clerk, Prisma, Inngest, fs) are fully mocked so tests run in isolation — no real database or API keys required.
| File | What it tests |
|---|---|
src/action.test.ts |
All five server actions: auth guards, validation, happy paths |
src/app/api/get-project-pdf/route.test.ts |
PDF route: missing/invalid UUID (400), not found (404), success (200) |
npm run testTests run in Node environment and complete in under 1 second.
A GitHub Actions workflow (.github/workflows/ci.yml) runs automatically on every push and pull request to main.
Push / PR to main
│
▼
┌─────────────────────────────┐
│ ubuntu-latest / Node 20 │
│ 1. npm ci │
│ 2. npm run lint │
│ 3. npm run test │
└─────────────────────────────┘
npm ci(notnpm install) is used to ensure a reproducible, lock-file-exact install on every run.- Lint runs before tests so style/type issues are caught first.
- No environment secrets are needed — all dependencies are mocked in tests.
Contributions are welcome. Please open an issue before submitting a pull request for significant changes. See Contribution for a list of known improvements and contribution guidelines.
This project is licensed under the MIT License.