AI-powered incident response assistant built with LangGraph ReAct agents and Model Context Protocol (MCP) for DevPR 2025 conference.
📊 Presentation: https://docs.google.com/presentation/d/1caZhQFIXv2K4e1ljIR9I85Xv51vegV3-S64tanx4Q4M/edit?usp=sharing
# 1. Setup Kubernetes cluster and all services
make setup
# 2. Configure local DNS (requires sudo)
make setup-hosts
# 3. Copy environment variables
cp .env.example .env # Edit with your tokens
# 4. Install Python dependencies
uv install
# 5. Launch Discord bot
python -m src.vibedebugger_discord.chatbot- Docker (v20.10+) - Container runtime
- Kind (v0.20+) - Local Kubernetes clusters
- kubectl (v1.28+) - Kubernetes CLI
- k9s (optional) - Terminal UI for Kubernetes
- Python (3.10+) - Agent runtime
- uv - Fast Python package manager
- Node.js (v20) via nvm - For Backstage only
- Discord Bot Token - For chat interface
make agent PROMPT="Your question" # Run agent from terminal| Command | Description |
|---|---|
make setup |
Create Kind cluster with all services (ArgoCD, Prometheus, Grafana, Backstage) |
make teardown |
Destroy cluster and cleanup |
make setup-hosts |
Configure *.localdev.me domains in /etc/hosts |
make status |
Check health of all components |
make validate |
Run validation tests |
make port-forward |
Expose all services locally |
make get-passwords |
Retrieve admin credentials |
make deploy-sample |
Deploy nginx sample app for testing |
make logs-<service> |
View logs (argocd/prometheus/grafana/backstage/discord-bot) |
hello-langrafo/
├── src/ # Agent source code
│ ├── agent/ # LangGraph ReAct agent core
│ │ ├── graph.py # Main agent orchestration
│ │ └── constants.py # System prompts and configurations
│ ├── tools/ # MCP tool implementations
│ │ ├── k8s.py # Kubernetes operations
│ │ ├── prometheus.py # Metrics and alerts
│ │ ├── argocd.py # GitOps management
│ │ └── backstage.py # Service catalog
│ ├── mcps/ # MCP server declarations
│ └── vibedebugger_discord/ # Discord bot interface
│ └── chatbot.py # Bot implementation
├── cluster/ # Kubernetes infrastructure
│ ├── resources/ # K8s manifests
│ │ ├── argocd/ # GitOps deployment configs
│ │ ├── backstage/ # Developer portal
│ │ ├── grafana/ # Dashboards and visualizations
│ │ ├── prometheus/ # Monitoring and alerting
│ │ └── sample-apps/ # Test applications
│ ├── scripts/ # Automation scripts
│ └── Makefile # Cluster management commands
├── scripts/ # Helper utilities
│ └── agent_cli.py # Terminal interface for agent
├── assets/ # Documentation images
├── docker-compose.yml # Langfuse observability stack
├── langgraph.json # LangGraph configuration
├── pyproject.toml # Python dependencies
└── uv.lock # Locked dependencies
After make setup, services are available at:
| Service | URL | Default Credentials |
|---|---|---|
| ArgoCD | http://argocd.localdev.me | admin / (run make get-passwords) |
| Backstage | http://backstage.localdev.me | No auth required |
| Grafana | http://grafana.localdev.me | admin / admin |
| Prometheus | http://prometheus.localdev.me | No auth required |
Note: Run
make setup-hostsonce to configure local DNS
- Configure
.envwith your Discord bot token - Run
python -m src.vibedebugger_discord.chatbot - Send messages in Discord channel
# Direct query
make agent PROMPT="Help me investigate nginx-sample in apps namespace"
# Or using the CLI script
python scripts/agent_cli.py "Check pod status in all namespaces"Create .env file with:
# Required
OPENAI_API_KEY=sk-...
DISCORD_BOT_TOKEN=...
DISCORD_CHANNEL_ID=...
# Optional (uses defaults)
PROMETHEUS_URL=http://localhost:9090
ARGOCD_SERVER=localhost:8080
BACKSTAGE_URL=http://localhost:7007Click to view screenshots
- ReAct Agent: Reasoning + Acting loop with LangGraph
- MCP Tools: Direct integrations with K8s, Prometheus, ArgoCD, Backstage
- Local Kind Cluster: Full observability stack for testing
- Discord Interface: Real-time incident response chat
MIT - See LICENSE file for details
This project was created for DevPR Conference 2025. Feel free to fork and adapt for your needs!
Built with ❤️ by Gabriel Dantas