AI Job Agent 🤖

An intelligent automation system that scrapes job postings, tailors resumes using AI, and automates job application submissions.

📋 Overview

AI Job Agent is a FastAPI-based application that leverages LangChain, OpenAI, and Playwright to streamline the job application process. It consists of three core services that work together to automate job hunting:

1. Job Scraper: Extracts job details from various platforms (LinkedIn, Indeed, etc.)
2. Resume Tailoring: Uses LLM to customize resumes for specific job descriptions
3. Application Submitter: Automates form filling and submission (LinkedIn Easy Apply)

✨ Features

• 🔍 Intelligent Job Scraping: Playwright-based scrapers for dynamic content
• 🤖 AI-Powered Resume Customization: LangChain + OpenAI GPT-3.5 integration
• 📝 Automatic Application Submission: Browser automation with Playwright
• 💾 Database Persistence: SQLAlchemy with PostgreSQL/SQLite support
• 🚀 RESTful API: FastAPI with automatic interactive documentation
• 🛡️ Robust Error Handling: Graceful degradation when services unavailable
• 🧪 Comprehensive Testing: Unit and integration tests with pytest

🛠️ Tech Stack

Backend

• FastAPI: Modern async web framework
• LangChain: LLM orchestration framework
• OpenAI API: GPT-3.5-turbo for resume tailoring
• Playwright: Browser automation for scraping/submission
• SQLAlchemy: ORM for database operations

Database

• PostgreSQL: Production database (recommended)
• SQLite: Development/testing database

Libraries

• BeautifulSoup4: HTML parsing
• Pydantic: Data validation
• Uvicorn: ASGI server

📦 Installation

Prerequisites

• Python 3.10+
• PostgreSQL (optional, SQLite works for development)
• OpenAI API key

Setup

1. Clone the repository

git clone <repository-url>
cd ai-job-agent

2. Create virtual environment

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Install Playwright browsers

playwright install chromium

5. Configure environment variables

cp .env.example .env
# Edit .env with your credentials

Required environment variables:

DATABASE_URL=postgresql://user:password@localhost/dbname
OPENAI_API_KEY=sk-your-openai-api-key

6. Initialize database

# Database tables are created automatically on first run
# Or manually with:
python -c "from app.database import engine, Base; from app.models import *; Base.metadata.create_all(bind=engine)"

🚀 Usage

Start the server

Development:

uvicorn app.main:app --reload

Production:

uvicorn app.main:app --host 0.0.0.0 --port 8000

API Documentation

Once running, visit:

• Interactive Docs: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

API Endpoints

1. Scrape Job Posting

POST /jobs/scrape
Content-Type: application/json

{
  "url": "https://www.linkedin.com/jobs/view/123456"
}

Response:

{
  "id": 1,
  "title": "Senior Software Engineer",
  "company": "Tech Corp",
  "description": "Job description...",
  "url": "https://...",
  "source": "linkedin",
  "created_at": "2025-11-24T12:00:00"
}

2. Tailor Resume

POST /resumes/tailor
Content-Type: application/json

{
  "base_resume": "Your base resume content...",
  "job_description": "Target job description..."
}

Response:

{
  "id": 1,
  "content": "Tailored resume content optimized for the job...",
  "base_resume": false,
  "created_at": "2025-11-24T12:00:00"
}

3. Submit Application

POST /applications/submit
Content-Type: application/json

{
  "job_id": 1,
  "resume_id": 1
}

Response:

{
  "id": 1,
  "job_id": 1,
  "resume_id": 1,
  "status": "submitted",
  "created_at": "2025-11-24T12:00:00"
}

📁 Project Structure

ai-job-agent/
├── app/
│   ├── __init__.py
│   ├── main.py                 # FastAPI application & endpoints
│   ├── database.py             # Database configuration
│   ├── models.py               # SQLAlchemy models
│   ├── agent.py                # LangChain agent setup
│   └── services/
│       ├── __init__.py
│       ├── scraper.py          # Base scraper & factory
│       ├── resume.py           # LLM resume builder
│       ├── submitter.py        # Application submitter
│       └── scrapers/
│           ├── __init__.py
│           └── linkedin.py     # LinkedIn-specific scraper
├── tests/
│   ├── test_basic.py           # Service unit tests
│   ├── test_scraper.py         # Scraper tests
│   ├── test_resume_llm.py      # LLM integration tests
│   └── test_api.py             # API endpoint tests
├── requirements.txt
├── .env.example
└── README.md

🔄 Implementation Workflow

1. Job Scraping Flow

User provides URL
    ↓
ScraperFactory selects appropriate scraper
    ↓
Playwright launches browser → Navigates to URL
    ↓
BeautifulSoup parses HTML → Extracts data
    ↓
Job saved to database → Returns Job object

2. Resume Tailoring Flow

User provides base resume + job description
    ↓
ResumeBuilder initializes ChatOpenAI
    ↓
Prompt template formats input
    ↓
LLM generates tailored resume
    ↓
Resume saved to database → Returns Resume object

3. Application Submission Flow

User provides job_id + resume_id
    ↓
Application record created (status: PENDING)
    ↓
ApplicationSubmitter routes to correct submitter
    ↓
LinkedIn: Launch browser → Login → Navigate to job → Click Easy Apply
    ↓
Fill form fields → Submit → Update status

🧪 Testing

Run all tests:

DATABASE_URL=sqlite:///./test.db pytest tests/

Run specific test file:

DATABASE_URL=sqlite:///./test.db pytest tests/test_api.py -v

With coverage:

DATABASE_URL=sqlite:///./test.db pytest --cov=app tests/

⚙️ Configuration

Database Options

PostgreSQL (Production):

DATABASE_URL=postgresql://user:password@localhost:5432/jobagent

SQLite (Development):

DATABASE_URL=sqlite:///./dev.db

Scraper Configuration

The ScraperFactory automatically selects scrapers based on URL:

• Contains linkedin.com → LinkedInScraper
• Default → MockScraper

LLM Configuration

Edit app/services/resume.py to customize:

self.llm = ChatOpenAI(
    temperature=0.7,        # Creativity (0-1)
    model="gpt-3.5-turbo"  # Model selection
)

Available models:

• gpt-3.5-turbo (faster, cheaper)
• gpt-4 (higher quality, slower)

🔒 Security Notes

• Never commit .env file or API keys to version control
• Use environment variables for all sensitive data
• Implement rate limiting for production deployments
• Consider using background tasks (Celery/Arq) for long-running operations
• Add authentication/authorization for production API

🚧 Current Limitations

1. LinkedIn Scraper: Only handles public job pages (no authentication)
2. Application Submitter: Skeleton implementation (requires manual login setup)
3. Error Recovery: Limited retry logic for network failures
4. Rate Limiting: No built-in request throttling

🎯 Future Enhancements

• Complete LinkedIn Easy Apply automation with cookie-based auth
• Add Indeed, Glassdoor scrapers
• Implement background task queue (Celery)
• Add job matching/filtering based on criteria
• Email notifications for application status
• Web UI dashboard for monitoring
• Export applications to CSV/PDF
• Multi-tenant support with user authentication

🤝 Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open Pull Request

📝 License

This project is licensed under the MIT License - see LICENSE file for details.

🙏 Acknowledgments

• FastAPI - Web framework
• LangChain - LLM orchestration
• Playwright - Browser automation
• OpenAI - GPT models

📧 Support

For issues and questions:

• Open an issue on GitHub
• Contact: [your-email@example.com]

---

Note: This tool is for educational purposes. Always respect websites' Terms of Service and robots.txt when scraping. Use responsibly and ethically.