Context-Aware Learning and Intelligent Command Orchestrator

AI-powered browser automation with GPT reasoning, visual intelligence, and distributed architecture.

Tags: ai-automation browser-automation gpt-4 playwright web-scraping ocr intelligent-agents python celery redis

# Install dependencies
pip install -r requirements/base.requirements.txt

# Set API key
export OPENAI_API_KEY=sk-your-key

# Launch interactive shell
./bin/calico

• AI-Powered: GPT-4 integration for intelligent task planning and execution
• Visual Intelligence: OCR support for CAPTCHAs, images, and complex layouts using Tesseract and Google Cloud Vision
• Interactive CLI: Active shell with tab completion and real-time monitoring
• Anti-Detection: Enhanced browser fingerprinting evasion using Patchright
• Session Management: Organized storage for screenshots, logs, and training data
• Distributed Architecture: Scalable design with Celery workers and Redis for session isolation
• MCP Integration: Browser control via Model Context Protocol (MCP) WebSocket service

Calico supports two browser control methods:

1. Python Server (default): Direct Playwright integration with built-in stealth features
2. MCP Server (alternative): Model Context Protocol WebSocket service for remote browser control

Key environment variables:

# AI Services
OPENAI_API_KEY=sk-...                    # Required: GPT-4 API key
GOOGLE_APPLICATION_CREDENTIALS=./keys/   # Optional: Google Vision OCR

# Browser Automation (Python Server - default)
PLAYWRIGHT_HEADLESS=true                # Browser display mode

# Browser Automation (MCP Server - alternative)
MCP_WS_URL=ws://localhost:7001          # MCP WebSocket service URL

# Session Storage
SESSION_STORAGE_DIR=./sessions          # Base directory for session data

Session storage structure:

./sessions/{session-uuid}/
├── metadata.json
├── photos/
├── logs/
└── data/

• Calico Core: Main automation engine with GPT-4 reasoning
• Browser Layer: Playwright/Patchright for stealth automation (default: Python server, alternative: MCP WebSocket server)
• Vision Layer: Multi-provider OCR (Tesseract, Google Cloud Vision)
• Task Queue: Celery with Redis for distributed processing
• Session Storage: UUID-based isolation for screenshots, logs, and data

MIT License - see LICENSE file for details.