Jobby

Personal job application tracker with AI-powered resume enhancement.

Uses Groq's free API tier (Llama 3.3 70B) for AI features — no subscription required, just a free API key.

Prerequisites

• Node.js 20+
• A free Groq API key — sign up at console.groq.com (no credit card required)

Quick Start

npm install
npm run dev

Open http://localhost:3000, navigate to the Config page, and paste your Groq API key.

How it works

1. Job Tracker — full CRUD for job applications stored in a local DuckDB file.
2. Resume Enhancer — upload your base resume PDF, select a job, paste the job description (or fetch it from a URL). Jobby then:
   • Extracts text from your PDF
   • Asks the LLM to generate an optimised version
   • Generates a tailored PDF
   • Lets you review and edit before downloading

The LLM returns structured JSON for the enhanced resume; a new PDF is generated from scratch with a clean Helvetica layout optimised for ATS compatibility.

Note: An earlier approach used PyMuPDF to apply in-place text replacements directly onto the original PDF (preserving the original design). It was abandoned due to persistent issues: replaced spans lost their original font styles, text overflowed bounding boxes, and overlapping characters made the output unreadable. Generating a fresh PDF proved far more reliable.

Configuration

All settings live in jobby.config.json at the project root (gitignored — never committed). They are managed via the Config page in the UI:

Setting	Description
groq_api_key	Your Groq API key (gsk_...)
llm_model	Model to use for AI tasks (default: llama-3.3-70b-versatile)
target_currency	Default currency for salary conversion (default: EUR)
duckdb_path	Path to the DuckDB file (managed automatically)

Architecture

npm run dev
  └── Next.js :3000  -- UI, API routes, LLM calls, PDF generation

Local data:
  ./data/            -- DuckDB + PDFs (persists between runs, git-ignored)

External:
  api.groq.com       -- Llama 3.3 70B inference (free tier)

Stack

Layer	Technology
Frontend + API	Next.js (React, Turbopack)
Database	DuckDB (embedded, file-based)
PDF text extraction	unpdf (Mozilla PDF.js)
AI	Groq SDK + Llama 3.3 70B (free tier)
PDF generation	@react-pdf/renderer

Project structure

scripts/          fix-missing-history.ts (DuckDB maintenance utility)
src/
  app/            Pages and API routes
  components/     Sidebar, forms, dialogs, editors
  lib/            DuckDB, Groq LLM wrapper, PDF tools, config
data/             DuckDB file + uploaded PDFs (created on first run, git-ignored)