Deployment

HatiData runs anywhere — from your laptop to a production Kubernetes cluster.

Local (Development)

The fastest way to start. Runs DuckDB in-process, no containers needed.

pip install hatidata-agent
hati init
hati status

Connect with any Postgres client on port 5439:

psql -h localhost -p 5439 -U admin -c "SELECT 1"

Docker

Run HatiData as a container with persistent storage:

docker run -d \
  --name hatidata \
  -p 5439:5439 \
  -v hatidata-data:/data \
  -e HATIDATA_DEV_MODE=true \
  ghcr.io/hatios-ai/hatidata-proxy:latest

Docker Compose (Full Stack)

Run the proxy, control plane, and dashboard together:

git clone https://github.com/HatiOS-AI/HatiData.git
cd HatiData/dev
make up

Services:

• Proxy — localhost:5439 (accepts psql connections)
• Control Plane — localhost:8080 (REST API)
• Dashboard — localhost:5173 (admin UI)

Cloud (Managed)

For production workloads, HatiData Cloud runs in your VPC with managed infrastructure:

1. Sign up at hatidata.com/signup
2. Choose your cloud provider (AWS, GCP, or Azure)
3. HatiData provisions a dedicated instance in your region
4. Connect using the provided endpoint and API key

Cloud includes: managed upgrades, backups, monitoring, and hybrid SQL with premium embeddings.

Kubernetes (Self-Hosted)

Deploy on your own Kubernetes cluster using our Helm chart:

helm repo add hatidata https://charts.hatidata.com
helm install hatidata hatidata/hatidata \
  --set proxy.replicas=2 \
  --set controlPlane.replicas=2 \
  --set storage.class=gp3

See the Kubernetes guide for detailed configuration.

Environment Variables

Variable	Default	Description
HATIDATA_DEV_MODE	true	Enable dev mode (relaxed auth)
HATIDATA_LISTEN_ADDR	0.0.0.0:5439	Proxy listen address
HATIDATA_DUCKDB_MEMORY_LIMIT_MB	24000	DuckDB memory limit
HATIDATA_MAX_CONCURRENT_QUERIES	100	Max concurrent queries
HATIDATA_TLS_ENABLED	false	Enable TLS for wire protocol

For the full configuration reference, see Configuration.