bonk1t agent-os .cursorrules file for Python (stars: 33)

# Agent OS Platform Design

## System Overview
The Agent OS Platform is a FastAPI-based system for running AI agents with dynamic tool creation and secure execution capabilities.

## Core Architecture

### Components
1. **Master Container (FastAPI)**
   - Handles user interactions
   - Manages agent lifecycles
   - Coordinates tool execution
   - Runs on Python 3.13 (slim image)

2. **E2B Integration**
   - On-demand sandbox creation for tool execution
   - One sandbox per tool execution
   - Short-lived containers (destroyed after tool completion)
   - Maximum execution time: 10 minutes (OpenAI limit)
   - Firecracker microVM-based isolation
   - Supports any Linux-compatible framework
   - Handles package installation and I/O

3. **Redis Message Bus**
   - Inter-component communication
   - State management
   - Event distribution

4. **Tool Registry**
   - Firestore-based tool storage
   - Tools stored as standalone Python code
   - Each tool includes all dependencies
   - No caching or hot-reloading

### Development Environment
1. **Local Development**
   - Docker Compose based setup
   - E2B for sandbox isolation
   - Redis for message passing
   - Firestore emulator for storage

2. **Production Environment**
   - Similar to local but with real Firestore
   - E2B for production sandboxes
   - Potential future K8s migration

## Security Model

### Tool Execution
1. **E2B Sandbox**
   - Created on-demand for each tool execution
   - Destroyed immediately after tool completion
   - No session pooling or reuse
   - Clean environment for each execution
   - Resource limits enforcement
   - Network isolation
   - Secure file system access

2. **Tool Validation**
   - Static code analysis
   - Security checks
   - Dependency scanning
   - Automated testing

### Access Control
1. **Authentication**
   - Firebase Authentication
   - JWT validation
   - Role-based access

2. **Authorization**
   - Tool execution permissions
   - Admin capabilities
   - User isolation

## System Self-Improvement

### AI-Driven Updates
1. **Code Generation**
   - AI creates/modifies platform code
   - Generates pull requests
   - Must pass all tests
   - Manual review required

2. **Tool Creation**
   - AI generates new tools
   - Automated validation
   - Security scanning

### Quality Assurance
1. **Testing**
   - Very high coverage requirement
   - Integration tests
   - Security tests
   - Performance benchmarks

2. **Monitoring**
   - Execution metrics
   - Error tracking
   - Resource usage
   - User analytics

## Implementation Notes

### Current State
- Single container deployment
- File-based tool storage
- Basic security validation
- Manual tool approval

### Migration Path
1. **Phase 1: E2B Integration**
   - Add E2B sandbox support
   - Implement on-demand container creation
   - Add container cleanup after tool execution
   - Implement security checks
   - Add execution timeouts (10-minute limit)

2. **Phase 2: Message Bus**
   - Redis integration
   - State management
   - Event system

3. **Phase 3: Tool Registry**
   - Move to pure Firestore storage
   - Remove file-based storage
   - Implement standalone tool format

### Design Principles
1. **Simplicity**
   - Minimal abstractions
   - Clear responsibilities
   - Direct communication

2. **Security**
   - Isolated execution
   - Validated tools
   - Access control

3. **Maintainability**
   - Clear documentation
   - High test coverage
   - Automated improvements

## Maintenance Instructions

This file (.cursorrules) serves as the source of truth for system design.
When making significant changes:

1. Update this file first
2. Ensure consistency with implementation
3. Remove or update other docs if needed
4. Keep this file in sync with:
   - README.md (high-level overview)
   - docker-compose.yml (deployment)
   - backend/services/* (implementation)

---
Last Updated: [Current Date]
Update this file when making architectural changes.