wb-jobs is a production-grade, AI-powered CLI application designed to automate search, harvest detailed metadata, evaluate fit, and orchestrate job applications on LinkedIn and Indeed. By combining Playwright browser automation, SQLite persistence, and state-of-the-art LLMs (Gemini, Groq), wb-jobs removes the tedious aspects of job hunting while keeping you firmly in control.

⚠️ Personal / educational project. Automating LinkedIn or Indeed may violate their Terms of Service — use responsibly, human-in-the-loop, and at your own risk. Not affiliated with, endorsed by, or connected to any platform. See DISCLAIMER.md.

❓ Why Use wb-jobs?

• Anti-Bot Evasion: Operates using a real Chrome instance with randomized human-like delays, mouse curves, and scroll patterns.
• Deep Metadata Extraction: Automatically harvests applicant counts, profile matching status, and hiring team details from job pages.
• Dynamic AI Scoring: Employs configurable system prompts to evaluate job descriptions against your experience, calculating matching scores and selecting resume variants.
• Human-in-the-Loop Safeguards: Handles form filling and uploads automatically, but stops for manual verification before submission, preventing unintended or messy applications.
• Clean Local History: Persists every visited, skipped, and applied position in a local SQLite database, supporting instant data exports to JSON, CSV, and Markdown.

📐 Architecture & Workflow

Here is the high-level workflow of how wb-jobs processes and manages your job applications:

📂 Project Directory Structure

.
├── docs/
│   └── images/
│       └── wb_jobs_logo.png     # Application logo
├── bin/
│   └── cli.js                  # CLI bin entry point
├── profile/
│   └── profile.json.example    # Profile credentials and matching weights template
├── prompts/                    # LLM Prompts folder
│   ├── match_prompt.txt
│   ├── cover_letter.txt
│   └── screening_questions.txt
├── src/                        # Source TypeScript code
│   ├── apply.ts                # LinkedIn scraping & application
│   ├── applyIndeed.ts          # Indeed scraping & application
│   ├── openJob.ts              # LinkedIn detailed page parsing
│   ├── openIndeedJob.ts        # Indeed detailed page parsing
│   ├── tracker.ts              # SQLite persistence & database migrations
│   └── ...
├── package.json
└── tsconfig.json

🚀 Installation & Setup

1. Prerequisites

Ensure you have Node.js (v18+) and Google Chrome installed on your system.

2. Local Development Setup

Clone the repository, install dependencies, and build the TypeScript source:

git clone https://github.com/wissemb11/wb-jobs.git
cd wb-jobs
npm install
npm run build

3. Run Globally

To make the CLI command wb-job available globally anywhere in your terminal, link it:

npm link
# or install globally from the local path:
npm install -g .

Now you can invoke the CLI globally via wb-job:

wb-job --help

[!TIP] Native Addon Troubleshooting (better-sqlite3 error)
If you encounter an error stating Could not locate the bindings file when executing wb-job, it means npm's safety mechanism blocked better-sqlite3's native C++ compilation script. Fix it by running:

# Approve the install script for better-sqlite3
npm install-scripts approve better-sqlite3

# Re-install globally with execution permissions enabled for both
npm install -g --allow-scripts=wb-jobs,better-sqlite3 .

⚙️ Configuration & Environment

Step 1: LLM Credentials

Copy the environment template and fill in your API keys in the generated .env file:

cp .env.example .env

Supported providers include:

• Gemini: GEMINI_API_KEY (Recommended default)
• Groq: GROQ_API_KEY
• OpenRouter: OPENROUTER_API_KEY

Step 2: Experience Profile & Resumes

1. Copy the profile template:

   cp profile/profile.json.example profile/profile.json

2. Open profile/profile.json and enter your name, target roles, work summary, and core skills.
3. Place your resume PDF files in the profile/ folder (e.g. profile/resume_fullstack.pdf, profile/resume_ai.pdf).

🛠️ Usage Manual

Step 1: Persist Login Sessions

To scrape and apply as a logged-in user, run the login utility to capture cookies:

wb-job login:linkedin
wb-job login:indeed

This opens a browser window. Once you log in, session details are persisted locally in linkedin.json and indeed.json (automatically ignored by Git).

Step 2: Running Application Pipelines

Run search, metadata harvesting, and human-in-the-loop applications:

# Search & evaluate LinkedIn roles globally
wb-job apply:linkedin --role="Senior Full-Stack Developer" --place="Germany|Remote" --easy --date="week"

# Search and harvest Indeed listings
wb-job apply:indeed --role="AI Engineer" --place="Remote" --date="day"

CLI Parameters & Flags:

• --role="Role1|Role2": Pipe-separated keywords for roles.
• --place="Loc1|Loc2": Pipe-separated target locations.
• --easy: Filter and apply only to Easy Apply / Quick Apply jobs.
• --date="day" | "week" | "month": Filter jobs by post date.
• --submit=SCORE: Automatically submit if the match score is above this threshold (e.g., 85). If omitted, the CLI always pauses for review.
• --save: Save all evaluated jobs and outcome states to the local database.

Step 3: Exporting Saved Data

Extract job statistics and logs from the SQLite database to various formats:

# Export to CSV format
wb-job export-db --as csv

# Export to Markdown, JSON, or Text
wb-job export-db --as md
wb-job export-db --as json

Step 4: Resetting the Database

To clear logs and start with a clean database:

wb-job purge-db

📚 Documentation

For deeper documentation, check the wiki/ directory:

• Architecture & Engine Details
• Comprehensive Command Parameters
• Scoring System & Prompt Tuning
• API Configuration Guide
• LinkedIn Auto-Apply Tutorial
• CLI Commands & Terminal Visualizations
• Global Relocation & Application Strategy