- Conversational control — trigger DAGs, view runs, pause/resume schedules, inspect logs.
- LLM‑powered intelligence — plug in OpenAI · Gemini · Claude · Groq.
- RBAC aware — honours Airflow permissions when calling its REST API.
- Persistent memory — summaries & checkpoints stored in PostgreSQL (LangGraph).
- Deployment parity — same image on Docker Compose, Kubernetes, ECS.
Full guide →
Getting Started
# Clone & copy env template
git clone https://github.com/NikhilSuthar/airflow-copilot.git
cd airflow-copilot
cp .env.example .env # fill in keys / secrets
# Launch Copilot + ngrok + bot‑updater
docker compose -f docker-compose.local.yml up -d
- Ensure Airflow (≥ 2.5) is reachable at
${AIRFLOW_BASE_URL}
- Ngrok publishes an HTTPS tunnel
- bot‑updater patches the Azure Bot endpoint automatically
- Open Microsoft Teams → chat with Airflow Copilot 🚀
🚀 See Airflow Copilot in action inside Microsoft Teams — from updating credentials to triggering DAGs and viewing logs.
🧾 Or explore the same examples → Chat Examples
Read the deep dive → Architecture
Mode | Guide | Best For |
---|---|---|
Docker Compose (local) | Local Deployment |
Dev / PoC |
Docker Compose (prod) | Docker based Deployment |
Small teams |
Kubernetes | Kubernetes based Deployment |
Cloud & scale |
Each mode uses the same .env
. Bring your own DB & TLS.
All runtime settings are environment variables.
See the reference → Environment Variables
Licensed under the MIT License © 2025 Nikhil Suthar.