Tailored Cover Letter Generator

Generate personalized cover letters from your resume and job postings using AI.

Features

• Parse PDF resumes; scrape job URLs (Playwright) or paste/upload job text
• Gemini generates structured letter sections; LaTeX → PDF (optional logo)
• Edit via chat (natural language); guardrail blocks off-topic requests

How It Works

Job input (URL/text/PDF) + resume PDF → scrape/parse → Gemini (sections) → inject into LaTeX template → pdflatex → PDF. User can edit in UI or ask for changes in chat; edits go through a guardrail, then Gemini returns revised sections; diff and recompile without another scrape.

Architecture

• Frontend: Next.js, TypeScript, Tailwind. frontend/. API routes proxy to backend.
• Backend: FastAPI, backend/. LangChain for orchestration (generate, edit, guardrail). Optional AWS RDS (Postgres + pgvector) for template storage. Bearer API_KEY; CORS via CORS_ORIGINS.

Backend modules: api.py — routes, LaTeX build, PDF compile. main.py — generate, edit, guardrail (Gemini). scraper.py — Playwright. resume_parser.py — PDF text. database.py — AWS RDS + pgvector (optional).

Routes: POST /generate (resume + job → PDF + sections), POST /compile (sections → PDF), POST /edit (instruction + sections → proposed sections or guardrail message), POST /latex (sections → LaTeX), GET /health.

Project Structure

tailored/
├── frontend/     # Next.js, app/api, components, lib
├── backend/      # api.py, main.py, scraper.py, resume_parser.py, database.py,
│                 # cover_letter_template.tex, Dockerfile, requirements.txt
└── README.md

Local Development

Prereqs: Node 18+, Python 3.10+, pdflatex, Gemini API key.

Backend: cd backend && python3 -m venv venv && source venv/bin/activate && pip install -r requirements.txt && playwright install chromium. Add .env with GEMINI_API_KEY=.... Run: uvicorn api:app --reload --port 8000.

Frontend: cd frontend && npm install && npm run dev. App at http://localhost:3000.

Environment Variables (Backend)

Variable	Description
GEMINI_API_KEY	Required for generate/edit.
API_KEY	Optional; Bearer auth on protected routes.
CORS_ORIGINS	Optional; comma-separated origins.
RDS_*	Optional; AWS RDS Postgres + pgvector (database.py).

Deployment

Backend: Docker + AWS ECS/Fargate (see Dockerfile).

Troubleshooting

• pdflatex missing: brew install --cask basictex then eval "$(/usr/libexec/path_helper)".
• Blank PDF / fonts: sudo tlmgr update --self && sudo tlmgr install lmodern.
• Scraper: Slow or missing content → raw page text still sent to AI.