Deployment Overview

HEBBS is distributed as a single statically-linked binary with embedded RocksDB. There are no external database dependencies — all storage, indexing, and embedding happen inside the process.

Deployment Modes

Standalone Binary

Run HEBBS directly on a host machine. Best for development, testing, and small-scale deployments.

hebbs-server --config config.toml

The server listens on gRPC (default port 50051) and exposes a Prometheus metrics endpoint (default port 9090).

Docker

Run HEBBS in a container with persistent volume mounts for data durability. Suitable for single-host production deployments.

docker run -d \
  -p 50051:50051 \
  -p 9090:9090 \
  -v hebbs-data:/data \
  ghcr.io/hebbs-ai/hebbs:latest

See Docker for full configuration.

Kubernetes (Helm)

Deploy on Kubernetes with the official Helm chart. Handles persistence, health checks, and scaling configuration.

helm install hebbs oci://ghcr.io/hebbs-ai/charts/hebbs

See Kubernetes for the full values reference.

Choosing the Right Mode

Mode	When to use
Standalone	Development, CI, single-tenant demos
Docker	Single-host production, docker-compose stacks
Kubernetes	Multi-tenant production, auto-scaling, high availability

Configuration

HEBBS is configured via a TOML file. Key sections:

Section	Purpose
[server]	Listen address, TLS, authentication
[storage]	Data directory, RocksDB tuning
[embedding]	Model path, batch size, hardware acceleration
[index]	HNSW parameters, temporal index settings
[reflect]	LLM provider, reflection pipeline configuration
[metrics]	Prometheus endpoint configuration

See the full configuration reference for all options.

Next Steps

• Docker — container deployment guide
• Kubernetes — Helm chart deployment
• Terraform (AWS) — infrastructure-as-code for AWS
• Monitoring — Prometheus and Grafana setup
• Production Checklist — go-live checklist