AI Server Management Agent

A self-hosted AI agent for server management and troubleshooting that combines RAG (Retrieval Augmented Generation) with secure file access.

Key Features

• Secure file reading with Redis caching
• RAG-based document retrieval with vector search
• Asynchronous operations
• SSH-based file access with security controls
• Web interface for easy interaction
• Comprehensive file search and analysis

Architecture

The agent uses a layered architecture that separates concerns:

1. File Access Layer

   • Secure SSH connections
   • Redis caching
   • Path filtering and security controls

2. Knowledge Layer

   • FAISS vector store
   • Document processing
   • Similarity search

3. Intelligence Layer

   • Claude API integration
   • Context building
   • Response generation

4. Interface Layer

   • Gradio web UI
   • Async request handling
   • Session management

Components

/src
├── core/           # Core agent logic and query handling
├── tools/          # File reading and caching services
├── knowledge_base/ # Vector store and document processing
├── rag/           # Retrieval and context generation
├── server_management/ # SSH and file operations
└── ui/            # Web interface

Installation

1. System Requirements:

   • Python 3.10+
   • Redis server
   • SSH access to target servers

2. Setup:

   # Clone repository
   git clone [repository-url]
   
   # Create virtual environment
   python -m venv venv
   source venv/bin/activate
   
   # Install dependencies
   pip install -r requirements.txt
   
   # Configure environment
   cp .env.example .env
   # Edit .env with your settings

Configuration

Required environment variables:

• ANTHROPIC_API_KEY: Your Claude API key
• Server configurations (in .env):

  SERVERS='[
    {
      "name": "server1",
      "ip": "xxx.xxx.xxx.xxx",
      "search_paths": ["/path1", "/path2"]
    }
  ]'

Usage

1. Start Redis server:

   redis-server

2. Configure SSH keys:

   ssh-keygen -t ed25519
   ssh-copy-id user@server

3. Start the agent:

   python src/ui/app.py

4. Access web interface at http://localhost:7860

Security Features

• No direct LLM access to servers
• File content cached in Redis
• SSH key-based authentication
• Configurable excluded paths
• Path filtering and validation
• Access logging and monitoring

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT