All of Basedash, inside your boundary

Anywhere you can run Docker or Kubernetes: AWS, Azure, Google Cloud, DigitalOcean, an on-premises data center, or a hybrid setup, in any region. Smaller teams run Docker Compose on a single server, with or without the Distr agent, which Basedash recommends up to about 500 users. Larger rollouts use the Helm chart on Kubernetes 1.28 or later, with an ingress controller and a managed PostgreSQL database, so the app can scale across replicas.

Up to 100 users, one server with 2 vCPUs, 8 GB of memory, and a 50 GB SSD runs Basedash on Docker Compose; up to 500 users, 4 vCPUs, 16 GB, and 100 GB. Beyond that, plan on Kubernetes with about one app replica per 500 users at 2 vCPUs and 4 GiB each, plus managed PostgreSQL 16 or later. PostgreSQL is the only required dependency. Avoid burstable instance types, and talk to Basedash about anything above 10,000 users.

Nothing updates automatically. Releases use calendar versions, such as 2026.09.30, and each is published for Docker Compose and Helm with release notes in the Distr portal. With the Distr agent, you pick the new version in the portal; with plain Compose, you pull and restart; with Helm, you change the image tag and run helm upgrade. Migrations run on startup, so back up first, check the notes for migrations that can't be rolled back, and test in staging if you need strict change control.

OpenAI, Anthropic, Azure OpenAI, Amazon Bedrock, xAI, and OpenRouter, plus any OpenAI-compatible endpoint such as an internal LLM gateway. You set your own keys and choose the main and fast models with two environment variables, and Bedrock can use an IAM role instead of access keys. The app starts without an AI key, but chat, chart generation, and insights need one. Basedash tests against its default models, so others may produce lower-quality SQL and charts.

Not completely on its own. A self-hosted deployment validates its license with an HTTPS call to charts.basedash.com at startup and every six hours, and it won't start if that first check fails. Container images also come from registry.distr.sh. Everything else can stay internal: AI can run through Bedrock or your own gateway, and email, object storage, and data sources can all be internal services. If your network allows no outbound traffic at all, contact Basedash before you deploy.

Only what licensing needs. Every six hours the server sends a heartbeat with the app version and its organization and user counts, which Basedash uses for billing. Your data, queries, dashboards, and chats stay in your deployment and its PostgreSQL database, and prompts go only to the AI provider you configure. If a heartbeat fails, existing users keep working while the server retries, and a banner appears until it reconnects.

Self-hosting is part of the Enterprise plan and currently available to select customers. It starts with an email to [email protected], followed by an infrastructure review and sizing. You then get a license key and access to the Distr customer portal, which holds the container images, Compose files, Helm chart, and release notes for every version. Basedash supports the deployment after launch, and sandbox licenses for QA and CI environments aren't billed. It includes the same access controls as the cloud, from SSO and SCIM to audit logs. See Enterprise for the rest of the plan.