8000 GitHub - iemafzalhassan/OutagePilot: OutagePilot uses a multi-agent system to autonomously detect, diagnose, and remediate infrastructure incidents, helping you achieve a self-healing Kubernetes environment. · GitHub
[go: up one dir, main page]

Skip to content

About

OutagePilot uses a multi-agent system to autonomously detect, diagnose, and remediate infrastructure incidents, helping you achieve a self-healing Kubernetes environment.

Topics

Resources

Code of conduct

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

OutagePilot

An autonomous AI SRE platform for Kubernetes.

OutagePilot uses a multi-agent system to autonomously detect, diagnose, and remediate infrastructure incidents, helping you achieve a self-healing Kubernetes environment.

License Issues Pull Requests Contributors Build Status Coverage Dependencies


📺 Watch the Demo

OutagePilot Demo

Table of Contents
  1. Features
  2. Architecture
  3. Tech Stack
  4. Getting Started
  5. Contributing
  6. License
  7. Contact

✨ Features

  • Autonomous Kubernetes Monitoring: Lightweight Go agents run directly in your cluster to detect common issues like CrashLoopBackOff, ImagePullBackOff, and OOMKilled.
  • AI-Powered Root Cause Analysis: A central "Brain" orchestrator powered by Google Gemini uses a ReAct loop to perform deep investigations into incident signals.
  • Dynamic Tool Use: The AI Brain dynamically uses a rich set of read-only Kubernetes tools (via MCP) to gather context, just like a human SRE would.
  • One-Click Remediation: The AI proposes a remediation plan (e.g., "Set image to nginx:latest" or "Rollback deployment"), which you can approve directly from the dashboard.
  • Real-Time Dashboard: A React-based frontend provides a live stream of incidents, agent status, and pending approvals.
  • Safe & Secure: Privileged operations are only executed after explicit user approval. All actions are audited.

🏗️ Architecture

OutagePilot uses a distributed, multi-agent architecture designed for safety and scalability.

graph TD
    subgraph User's K8s Cluster
        A[Go Watchdog Agent] -- gRPC Signal --> B{Backend Server}
        C(k8s-mcp-server) -- Tool Execution --> K8s(Kubernetes API)
    end

    subgraph OutagePilot Control Plane
        B -- 1. Creates Incident --> DB[(Postgres DB)]
        B -- 2. Starts Investigation --> AI((Gemini AI Brain))
        AI -- 3. Gathers Context --> MCP_Client[MCP Client]
        MCP_Client -- 4. Calls Read-Only Tools --> C
        AI -- 5. Proposes Fix --> B
        B -- 6. Awaits Approval --> UI[React Frontend]
    end

    subgraph User Interface
        UI -- Real-time Updates --> SSE
        User -- Approves Fix --> B
    end

    B -- 7. Executes Command --> MCP_Client
    MCP_Client -- 8. Calls Privileged Tool --> C

    style A fill:#3E6BFF,stroke:#fff,stroke-width:2px,color:#fff
    style C fill:#3E6BFF,stroke:#fff,stroke-width:2px,color:#fff
    style B fill:#1a1a1a,stroke:#fff,stroke-width:2px,color:#fff
    style UI fill:#1a1a1a,stroke:#fff,stroke-width:2px,color:#fff
    style AI fill:#f2b01e,stroke:#fff,stroke-width:2px,color:#000
Loading
  1. Go Watchdog Agents: These lightweight binaries run in your cluster, monitor for specific failure modes, and send a Signal to the backend when an issue is detected.
  2. Python Backend: The central "Brain" of the system. It receives signals, creates incidents, and orchestrates the AI investigation.
  3. Gemini AI Brain: A sophisticated prompt and ReAct (Reason+Act) loop that uses Gemini to investigate an incident. It behaves like a human SRE, using tools to gather information step-by-step.
  4. MCP Server: A Node.js server that acts as a secure gateway to the Kubernetes API. It exposes a set of "tools" (e.g., k8s_describe_pod) that the AI can call. This prevents the AI from having direct, unrestricted API access.
  5. React Frontend: The mission control center. It provides a real-time view of the system and is the single point for approving any mutating actions proposed by the AI.

🛠️ Tech Stack

Component Technologies
Backend Python, FastAPI, SQLAlchemy, Alembic
Frontend TypeScript, React, Vite, Zustand, Tailwind CSS
Agents Go
AI Google Gemini
Tooling MCP (Model Context Protocol)
Database PostgreSQL
Real-time Server-Sent Events (SSE)

🚀 Getting Started

To get OutagePilot running for local development, please follow the detailed instructions in our CONTRIBUTING.md guide.

The basic steps are:

  1. Prerequisites: Ensure you have Go, Python, Node.js/Bun, and Docker installed.
  2. Configuration: Set up your .env file with a GEMINI_API_KEY.
  3. Run Backend: Install Python dependencies and run the FastAPI server.
  4. Run Agents: Run the Go watchdog agents.
  5. Run Frontend: Install Node dependencies and run the Vite dev server.

🤝 Contributing

Contributions are what make the open-source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

Please read our CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

📜 License

This project is under a custom source-available license. Please see the LICENSE.md file for details.

  • Personal & Non-Commercial Use: Free to use.
  • Commercial & Enterprise Use: Requires a license. Please connect Afzal Hassan for details.

📧 Contact

Afzal Hassan - iemafzalhassan@gmail.com

Project Link: https://github.com/iemafzalhassan/OutagePilot

Live Demo: outage-pilot.smapatticare.com

About

OutagePilot uses a multi-agent system to autonomously detect, diagnose, and remediate infrastructure incidents, helping you achieve a self-healing Kubernetes environment.

Topics

Resources

Code of conduct

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages

0