AI TAC (Home Edition)

An AI-powered network troubleshooting assistant that combines GPT-4's analytical capabilities with automated network device interaction.

TUI Demo

CLI Demo

Overview

AI TAC is a Go-based CLI tool that simulates a network TAC engineer by automatically executing diagnostic commands and analyzing their output. It uses GPT-4 to determine appropriate troubleshooting steps while maintaining strict safety measures to prevent any harmful commands from being executed.

Features

• 🔒 Safe command execution with built-in protections
• 🤖 AI-powered troubleshooting flow
• 📝 Automated command execution on network devices
• 💾 Persistent troubleshooting history
• ✨ Formatted markdown output for results

Prerequisites

• Go 1.x or higher
• OpenAI API key
• Network devices running Cisco IOS-XR
• Required Go packages (see Dependencies section)

Installation

1. Clone the repository:

git clone https://www.github.com/metajar/ai-tac
cd ai-tac

2. Install dependencies:

go mod download

3. Set up your OpenAI API key:

export OPENAI_API_KEY='your-api-key-here'

Usage

Run the program with the following flags:

go run cmd/cli/main.go -question "What is your troubleshooting question?" -host "device-ip-or-hostname"

Example:

go run cmd/cli/main.go -question "Why is interface GigabitEthernet0/0/0/0 down?" -host "192.168.1.1"

Build The TUI

You can build the TUI by running:

go build -o ./ai-tac ./cmd/tui

You can then run the ai-tac binary and have fun.

How It Works

1. You provide a troubleshooting question and target device
2. GPT-4 suggests appropriate diagnostic commands
3. Commands are automatically executed on the target device
4. Results are analyzed by GPT-4
5. Process continues until the issue is identified or more information is needed
6. Final analysis is presented in formatted markdown

Dependencies

• github.com/charmbracelet/glamour - For markdown rendering
• github.com/openai/openai-go - OpenAI API client
• github.com/scrapli/scrapligo - Network device automation
• Standard Go libraries

Security Features

• Strict command filtering to prevent configuration changes
• No debug commands allowed
• Prevention of resource-intensive operations
• Authentication handling with proper credential management

Configuration

The CLI tool currently uses hardcoded credentials for device access:

• Username: clab
• Password: clab@123

For production use, you should modify these to use environment variables or a secure configuration file.

The TUI has a prompt at the beginning of the troubleshooting session to set these.

Limitations

• Currently supports only Cisco IOS-XR devices
• Requires direct network access to devices
• Depends on GPT-4 API availability
• Not intended for production troubleshooting

Contributing

Contributions are welcome! Some areas that could use improvement:

• Multi-vendor support
• Secure credentials management
• Command template library
• Result caching and optimization
• Test coverage

Future Enhancements

• Support for additional network vendors
• Integration with existing monitoring systems
• Custom command template support
• Result database for knowledge building
• Enhanced security features

Disclaimer

This tool is for educational and development purposes only. It should not be used as a replacement for professional network troubleshooting in production environments.

Support

Srsly? There is none