JDExtractor

A local-first tool that fetches job postings, parses them, and uses AI (DeepSeek or Kimi) to tailor your resume and cover letter to each role. Includes contact management and AI-powered networking follow-ups. Ships as a single self-contained binary with a modern Svelte web UI.

How it works

1. Paste a job posting URL (or raw text) into the web UI
2. JDExtractor fetches and parses the job description via jina.ai
3. Your selected AI backend rewrites your resume and cover letter templates to match the role
4. Track applications, manage networking contacts, and generate AI follow-up messages — all from one place

Prerequisites

• A DeepSeek API key (pay-as-you-go, very cheap), or
• A Kimi API key (Moonshot AI, experimental support)

Installation

Download the latest binary for your platform from the Releases page.

# Linux / macOS
tar -xzf jdextractor_linux_amd64.tar.gz
./jdextractor

# Windows — extract the zip, then run:
.\jdextractor.exe

Quick start

# First run: create the config and template directories
./jdextractor setup

# Add your resume and cover letter templates to:
#   config/templates/resume.txt
#   config/templates/cover.txt (optional)

# Start the web UI (default: http://localhost:8080)
./jdextractor serve          # or: serve --open to auto-launch browser

Then open http://localhost:8080 and configure your AI backend in Settings.

Web UI

The Svelte UI has five main sections:

• Dashboard — Application stats, activity charts, match score distribution, and an overdue follow-up queue
• Jobs — Filterable table of all applications with inline editing, file viewers, and token usage tracking
• Process — Submit job URLs (single or batch) or paste raw text; streaming output shows real-time progress
• Contacts — Manage networking contacts, log conversations, generate AI follow-up messages, and link contacts to jobs
• Settings — Backend selection (DeepSeek / Kimi), API key, model, templates, and networking prompt configuration

A unified search bar in the header searches across both jobs and contacts.

CLI

# Process a URL
./jdextractor generate https://jobs.example.com/some-role

# Batch-process multiple URLs concurrently
./jdextractor generate --batch urls.txt

# Process a local file
./jdextractor generate --local path/to/job.txt

# List tracked applications
./jdextractor list

# Update application status (draft, applied, interviewing, offer, rejected)
./jdextractor status <dir-prefix> applied

# Contacts
./jdextractor contacts add --name "Jane Doe" --company Acme --role "Eng Manager"
./jdextractor contacts list
./jdextractor contacts log <id> --channel email --summary "Discussed the role"
./jdextractor contacts followup <id>          # AI-generated follow-up message
./jdextractor contacts overdue                # List contacts past their follow-up date
./jdextractor contacts link <contact-id> <job-dir>
./jdextractor contacts status <id> connected
./jdextractor contacts delete <id>

Build from source

Requires Go 1.25+ and Node.js (for the UI).

git clone https://github.com/Magd74NA/jdextractor.git
cd jdextractor
make build        # builds UI assets, then the Go binary
./out/jdextractor

Other targets: make clean, make run, make snapshot, make fmt, make vet.

Features

• Modern Svelte UI — Dashboard with activity charts, job statistics, and inline editing
• Dual backend support — DeepSeek (deepseek-chat, deepseek-reasoner) or Kimi K2.5 (experimental)
• Contact management — Track networking contacts with relationship status, tags, and conversation threads
• AI follow-ups — Generate context-aware follow-up messages with suggested timing and channel
• PII sanitization — Emails and phone numbers are redacted before sending data to the LLM
• Unified search — Search across jobs and contacts from a single search bar
• Streaming processing — Real-time progress updates as documents are generated
• Batch processing — Process multiple job URLs concurrently
• Local-first — All data stored relative to the binary; no cloud uploads or hidden config directories
• Customizable prompts — Edit system prompts and task lists for both resume generation and networking

Notes

• URL fetching depends on jina.ai to convert web pages to clean text
• DeepSeek: deepseek-chat recommended for most cases; deepseek-reasoner for complex roles
• Kimi: K2.5 model is experimental and still being tested

License

MIT