An artificial intelligence Nautobot ReAct Agent
Welcome to the Nautobot AI Agent project! This application provides a natural language interface for interacting with Nautobot APIs, enabling CRUD (Create, Read, Update, Delete) operations through an intuitive chat-like interface powered by AI.
This project simplifies network management by combining AI-driven agents with the Nautobot API to streamline and automate common network tasks.
This has been forked from https://github.com/automateyournetwork/Nautobot_react_agent. Please follow John Capobianco at the following to see what he is doing with AI. Great stuff on his channels
- https://bsky.app/profile/automateyournetwork.ca
- https://www.youtube.com/@johncapobianco2527
- https://x.com/John_Capobianco
- https://www.linkedin.com/in/john-capobianco-644a1515/
Powered by ChatGPT (gpt-4o)
Requires OpenAI API Key
Offers high accuracy and performance for handling natural language queries.
Recommended for production use.
API costs apply.
- Importing secrets from .env file so that you don't have to keep readding them all the time as you make changes.
- A tool that fetches all of the Nautobot API endpoints so you don't have to manually fill that out for your deployment.
- Logic to handle large endpoint responses that are too large for gpt4o to handle.
- played around with prompt engineering trying to get the LLM to not hallucinate when figuring out how to handle the responses. (It still does dumb stuff when I asked it how many arista devices I have.)
So far I am not sure this has made much of an improvement, but it does answer the question when you ask it "how many devices do I have?" Asking it about specific devices doesn't really work though.
Powered by Local LLM using Ollama
Completely free and private: All computations happen locally.
No external API calls required.
Performance: Works well for basic tasks but is less sophisticated compared to the ChatGPT-based version.
Recommended for personal or offline use cases.
I've struggled with this llama and I can't get it to work at all. My head hurts from banging it against the wall.
Natural Language Interface: Interact with Nautobot APIs using plain English commands.
CRUD Operations: Perform Create, Read, Update, and Delete tasks on your Nautobot data.
API Validation: Ensures commands align with supported Nautobot API endpoints.
Dynamic Tools: Auto-detects and leverages the appropriate tools for each task.
Local or Cloud Options: Choose between the main branch for high performance or the Ollama branch for privacy and offline capabilities.
Docker and Docker Compose installed.
OpenAI API Key (for the main branch).
Optional: Ollama installed for the local branch.
git clone https://github.com/<your-repo-name>/nautobot_react_agent.git
cd nautobot_react_agentUpdate the docker-compose.yml with your values.
docker-compose builddocker-compose upOpen your browser and go to http://localhost:8501.
Configure API Keys
Provide your Nautobot API URL, Nautobot Token, and OpenAI API Key in the configuration page.
Provide only your Nautobot API URL and Nautobot Token.
Use natural language to manage your Nautobot data. Example commands:
"Fetch all devices in the DC1 site."
"Create a new VLAN in site DC2."
NautobotController: Manages interactions with the Nautobot API.
LangChain ReAct Agent: Dynamically selects tools to process natural language queries.
Streamlit Interface: Provides an intuitive chat-like web UI.
Q: Which branch should I use?
Use the main branch for production-grade performance and OpenAI's latest capabilities.
Use the Ollama branch for offline and private operations, but expect reduced performance.
Q: How do I switch between branches?
To use the Ollama branch, run:
git checkout ollamaThen re-run the Docker setup.
Docker Issues: Ensure Docker is running and your system meets the necessary prerequisites.
OpenAI Key Errors: Check that your API key is valid and added correctly.
Nautobot API Errors: Verify your Nautobot instance is accessible, and the API token has the required permissions.