ChaosPilot

ChaosPilot is an AI-powered log analysis platform designed to automatically ingest production logs, detect patterns and anomalies, classify incidents, and generate actionable remediation plans. It leverages AI agents to analyze error, warning, and critical logs, providing auto-classification by severity and impact, smart response planning, and safe automated fixes with rollback capability. The platform also includes intelligent alerting and a live dashboard to visualize metrics, workflows, and insights in real-time. Users interact with ChaosPilot through the MCP-enabled ADK (Agent Development Kit) tooling and the web frontend, enabling you to run the ADK API server, manage agents, and observe automated analysis and remediation cycles.

To use ChaosPilot, start the MCP environment as described in the setup guide. Launch the ADK API server for agent management and analysis, then run the MCP Toolbox to orchestrate tools and agents. The web frontend provides UI access to the log analysis and fix recommendations workflow. You can select the Log Analyzer agent to submit logs, the Fix Recommender agent to obtain concrete remediation steps, and the Automated Incident Response flow to generate end-to-end response plans. The UI also supports debugging via the ADK UI, allowing you to interact with individual agents and inspect confidence scores, suggested fixes, and rollback options.

Prerequisites:\n- Python 3.9+ (or the Python environment you rely on in your setup)\n- uv (the lightweight Python tool runner) installed and accessible in PATH\n- Node.js 16+ if you plan to use the frontend (optional for MCP tooling)\n- Access to Google Cloud (BigQuery) and/or Azure OpenAI if you intend to use those integrations.\n\nStep-by-step installation:\n1) Clone the ChaosPilot repository:\nbash\ngit clone https://github.com/pmutua/ChaosPilot\ncd ChaosPilot\n\n2) Create and activate a Python virtual environment using uv:\nbash\n# Create environment (uv manages venv under the hood in this setup)\nuv venv\n\n# Activate virtual environment (Linux/macOS)\nsource .venv/bin/activate\n# Activate virtual environment (Windows)\n.venv\Scripts\activate\n\n3) Install dependencies via uv (or install normally via pip if preferred):\nbash\nuv sync\n\n4) Rename the environment template and configure environment variables:\nbash\n# Copy and edit environment file as needed\ncp .env.template .env # then populate required values\n\n5) Start the ChaosPilot services per the setup guide (start MCP Toolbox, then ADK API server, then the frontend):\nbash\n# Start MCP Toolbox (from repo root)\ncd mcp-toolbox\ntoolbox --tools-file="tools.yaml"\n\n# Start ADK API Server (agent_manager) with CORS enabled (from repo root)\nadk api_server agent_manager --allow_origins="*"\n\n# Start Frontend (optional if you want the UI)\ncd ../web\nnpm install\nnpm start\n\n6) Verify that you can access the application endpoints and the UI (as documented in the Quick Start and Setup guides).

Environment variables and integration notes:\n- APP_NAME, VERSION, and MODEL in the .env file should reflect your deployment.\n- GOOGLE_API_KEY and AZURE_API_KEY are required if you enable Google Gemini or Azure OpenAI integrations. Populate these with valid credentials or placeholders for secure handling.\n- GOOGLE_GENAI_USE_VERTEXAI determines whether Vertex AI is used for GenAI inference.\n- The ADK tooling (adk) is used to run API servers and debug agents; ensure you run from the correct repository subdirectory (agent_manager, web, mcp-toolbox) as shown in the Quick Start.\n- For production deployments, follow the Setup & Deployment guide to configure GCP/Cloud Run, IAM roles, and billing. Typical issues include CORS misconfigurations, missing API keys, or incorrect environment variable values; consult the Troubleshooting section of the docs if you encounter them.\n- If you’re using uv, you can leverage the unified venv and dependencies management to simplify repeated deployments across environments.