Self-host Raven on your own infrastructure.

One Docker Compose command. No telemetry, no callback to a vendor. Runs on anything from a 4-GB VPS to a Raspberry Pi 5.

System requirements

Raven runs comfortably on a modest VPS or a Raspberry Pi 5. The numbers below are guidance — the actual footprint depends on the embedding model and corpus size.

ResourceMinimumRecommended
CPU2 cores (x86-64 or ARM64)4+ cores
RAM4 GB8 GB+ (16 GB if running local LLMs)
Disk20 GB SSD100 GB+ for embeddings + objects
OSLinux with Docker 24+Ubuntu 24.04 LTS or Debian 13
NetworkOutbound to your model providerSame; or fully air-gapped with Ollama

Five-minute quick start

One command brings up Raven, PostgreSQL with pgvector, Valkey, and the AI worker. Out of the box it points at Ollama on the host; swap in OpenAI/Anthropic by editing one env file.

# 1. Grab the production Compose file (canonical, pinned to main)
curl -fsSL https://raw.githubusercontent.com/ravencloak-org/Raven/main/docker-compose.yml \
  -o docker-compose.yml

# 2. Generate secrets and start
docker compose up -d

# 3. Open the app
open http://localhost:8080

Full guide and edge / Raspberry Pi variants: github.com/ravencloak-org/Raven.

Upgrades and rollbacks

Every release is tagged with semver and ships its own migration plan. Upgrade in place with docker compose pull && docker compose up -d. Roll back by pinning the previous tag in your Compose file — migrations are forward-only but always backward-compatible within a minor version.

Subscribe to release notifications by watching the GitHub repo, or follow the release feed.

Community support

Self-hosted Raven is fully supported by the open community. Open an issue or a discussion on GitHub — the maintainers and other operators are usually responsive within a day.