Sheikh-Ai

Sheikh-Ai is an advanced AI Agent System designed for complex workflows involving browser automation, code execution, and multi-agent coordination.

Project Overview

This repository contains the core infrastructure for a sandboxed execution environment where AI agents can safely interact with a full Ubuntu system, including a web browser and a shell.

Architecture

Sandbox Environment

The sandbox is a Dockerized Ubuntu 22.04 environment that includes:

• Graphical Support: Xvfb for virtual frame buffer, x11vnc for VNC access, and websockify/noVNC for web-based VNC viewing.
• Browser: Google Chrome and Playwright for browser automation.
• Process Management: Supervisor is used to manage all background processes (X server, VNC server, API server).

Sandbox API

A FastAPI-based control API provides an interface for the agent to:

• File Operations: Read, write, list, and delete files.
• Shell Execution: Run arbitrary shell commands with timeout control.
• Browser Tools: Navigate pages, take screenshots, and interact with web elements via Playwright.

Project Structure

.
├── backend/
│   └── sandbox_api/          # FastAPI server implementation
├── docker/
│   └── sandbox/              # Dockerfile and Supervisor configuration
├── frontend/                 # Frontend application (in progress)
├── .gitignore
└── README.md

Getting Started

Prerequisites

• Docker
• Python 3.10+ (for local development/testing)

Running the Sandbox

To build and run the sandbox environment:

docker build -t sheikh-ai-sandbox -f docker/sandbox/Dockerfile .
docker run -p 8000:8000 -p 6080:6080 sheikh-ai-sandbox

• API Server: Accessible at http://localhost:8000
• VNC (noVNC): Accessible at http://localhost:6080/vnc.html

API Endpoints

• GET /health: Health check.
• GET /files/list: List files in a directory.
• GET /files/read: Read file content.
• POST /files/write: Write content to a file.
• DELETE /files/delete: Delete a file or directory.
• POST /shell/execute: Execute a shell command.
• POST /browser/goto: Navigate to a URL.
• GET /browser/screenshot: Take a screenshot of the current page.
• POST /browser/click: Click an element.
• POST /browser/type: Fill an input field.

Testing

You can run the API tests locally (requires fastapi, uvicorn, httpx installed):

python backend/sandbox_api/test_api.py