JobHunter AI

AI‑powered resume ingestion, job analysis, candidate/job matching, and outreach drafting built on Next.js, SQLite, and NVIDIA Kimi.

Stack

• Next.js (App Router, React 19)
• better-sqlite3 (local persistence)
• LLM calls via NVIDIA Kimi (HTTP client in lib/kimi.js)
• SerpApi for job search imports

Quick Start

git clone <your-repo-url>
cd agent-development
cp .env.example .env.local
npm install
npm run dev

App runs at http://127.0.0.1:3000.

Environment Variables (.env.local)

• NVIDIA_API_KEY – required for resume parsing, job analysis, matching, and email generation.
• SERPAPI_KEY – required for job search import.
• DB_PATH – optional; path to the SQLite file. Defaults to ./jobhunter.db. The app will create parent folders if needed.

Scripts

• npm run dev – start dev server (Turbopack).
• npm run build – production build.
• npm run start – serve the production build.

Data Flow Overview

1. Resume upload: /resume → POST /api/resume → PDF parsed → LLM (parseResume) → saved to resumes.
2. Job analysis: /jobs → POST /api/jobs → LLM (analyzeJob) → saved to jobs.
3. Matching: /match → POST /api/match with resume_id + job_id → LLM (matchResumeToJob) → saved to matches.
4. Outreach: /email → POST /api/email with match_id + tone → LLM (generateEmail) → saved to emails.
5. Search import: /search → POST /api/serpapi (Google Jobs via SerpApi) → optional import to /api/jobs.

Notes / Operational Considerations

• The database file is local; use DB_PATH to point to a writable location in your deployment environment.
• Resume uploads are limited by PDF parseability; ensure PDFs contain text (not pure scans).
• API keys are required for end‑to‑end functionality; without them, LLM-backed routes will fail.

Troubleshooting

• If you see pdf-parse type errors, ensure NVIDIA_API_KEY is set and the uploaded file is a real PDF with text.
• Port conflicts on 3000: stop other Next.js instances or run npm run dev -- --port 3001.