Anti-Detection & Stealth

• External stealth JavaScript (stealth.js) injected into browser context
• Kernel-level click forcing (force=True) to bypass DOM traps
• Gaussian-distributed random delays (3–8 seconds)
• Human-like typing speed (30–120ms per character)
• navigator.webdriver flag hidden
• Fake browser plugins and language headers
• Real Chrome user-agent spoofing
• Cookie persistence (avoids repeated login detection)

Code Quality

• mypy --strict — 0 errors across 16 source files
• ruff — 0 warnings (9 rules: E, F, W, I, N, UP, B, A, SIM)
• Cyclomatic Complexity — ≤7 average (radon)
• 100% type hints — ElementHandle, AsyncSession, SecretStr
• Custom exception hierarchy — 15 specific error types
• StrEnum for state management — zero magic strings

Python 3.14 Features

• PEP 649 — Native deferred annotations (no from __future__ import annotations)
• datetime.UTC — Modern alias instead of timezone.utc
• AsyncGenerator[T] — Without redundant second None parameter
• zip(strict=True) — Length validation on paired iterations
• Unquoted forward refs — Clean type annotations (DatabaseManager | None)

🚀 Quick Start

1. Clone & Install

git clone https://github.com/LeandroPG19/linkedin-easyapply-antidetection-bot.git
cd linkedin-easyapply-antidetection-bot
python -m venv .venv && source .venv/bin/activate
pip install -e .
playwright install chromium

2. Configure

cp .env.example .env

Edit .env with your credentials:

DEEPSEEK_API_KEY=sk-your-key-here
LINKEDIN_EMAIL=your-email@example.com
LINKEDIN_PASSWORD=your-password
RESUME_PATH=./resume.pdf
DRY_RUN=true  # Start with dry run!

Copy and edit the resume template:

cp resume.yaml.example resume.yaml

Edit search_config.yaml with your search preferences.

3. Run

# Dry run (does not submit applications)
python -m linkedin_bot.main

# Live mode (actually submits)
# Set DRY_RUN=false in .env first
python -m linkedin_bot.main

# Or use the registered CLI
linkedin-apply

📁 Architecture

linkedin-easyapply-antidetection-bot/
├── linkedin_bot/
│   ├── __init__.py
│   ├── main.py            # Main orchestrator + CLI entry point
│   ├── config.py           # Pydantic Settings (.env) + SearchConfig (YAML)
│   ├── logger.py           # structlog (JSON structured logging)
│   ├── enums.py            # StrEnum: ApplicationStatus, CircuitState
│   ├── exceptions.py       # Hierarchy of 15 custom exceptions
│   ├── stealth.js          # Anti-detection JavaScript (injected into browser)
│   ├── ai_engine.py        # DeepSeek AI + Circuit Breaker + Retry
│   ├── browser.py          # Playwright session + stealth + human delays
│   ├── linkedin_auth.py    # Login + session verification
│   ├── job_search.py       # URL builder + listings extractor
│   ├── applicator.py       # Easy Apply form filler + multi-step navigation
│   ├── tracker.py          # Application tracking (dedup by job_id)
│   └── db/
│       ├── __init__.py
│       ├── models.py       # SQLAlchemy 2.0 models (ApplicationRecord)
│       ├── session.py      # DatabaseManager singleton (async SQLite)
│       └── repository.py   # Repository pattern (UPSERT + stats)
├── cookies/                # Saved sessions (gitignored)
├── logs/                   # bot_database.db + logs (gitignored)
├── resume.yaml.example     # Resume template (edit and rename to resume.yaml)
├── search_config.yaml      # Search filters
├── .env.example            # Credentials template
├── pyproject.toml          # Dependencies + ruff + mypy config
└── README.md

Layer Diagram

┌─────────────────────────────────────────┐
│            main.py (Orchestrator)       │
│  CLI → Config → Auth → Search → Apply  │
├─────────────────────────────────────────┤
│     ai_engine.py     │    tracker.py    │
│  Circuit Breaker     │  Deduplication   │
│  Retry + Backoff     │  Stats (by enum) │
├──────────────────────┼──────────────────┤
│     browser.py       │    db/           │
│  Playwright + Stealth│  SQLAlchemy 2.0  │
│  Human delays        │  Repository      │
│  Cookie persistence  │  DatabaseManager │
└──────────────────────┴──────────────────┘

⚙️ Configuration

.env — Credentials & Behavior

search_config.yaml — Job Search Filters

keywords:
  - "Software Engineer"
  - "Backend Developer"
remote_only: true
experience_levels: [3, 4]    # 2=Entry, 3=Associate, 4=Mid-Senior
date_posted: 2               # 1=24h, 2=week, 3=month
locations: [""]               # Empty = no filter
blacklist_titles:
  - "senior"
  - "staff"
blacklist_companies:
  - "Acme Corp"
min_match_score: 50           # 0-100 (AI threshold)

resume.yaml — Your Profile

See resume.yaml.example for a complete template with all supported fields.

🛡️ Resilience Patterns

🔧 Development

Code Verification

# Activate virtualenv
source .venv/bin/activate

# Linting (9 rules)
ruff check linkedin_bot/

# Type checking (strict mode)
mypy linkedin_bot/ --strict

# Cyclomatic complexity
radon cc linkedin_bot/ -a -nc

# All together
ruff check linkedin_bot/ && mypy linkedin_bot/ --strict && radon cc linkedin_bot/ -a -nc

Dev Dependencies

pip install mypy types-PyYAML radon ruff

Tech Stack

💰 Cost

DeepSeek API is extremely affordable:

• ~50 applications/day × ~3 API calls = 150 calls
• ~$0.03 USD/day

🔒 Security Notes

• No credentials are stored in the repository — all secrets via .env (gitignored)
• No personal data in the codebase — resume data via resume.yaml (gitignored)
• Template files provided — .env.example and resume.yaml.example with placeholder data
• Anti-traversal protection — all file paths resolved via pathlib.resolve()
• SecretStr — passwords never logged or serialized

📜 License

MIT — See LICENSE for details.