// the find
databuddy-analytics/Databuddy
Open-source product analytics for startups: track visitors, events, funnels, and goals without cookies, and ask Databunny, the built-in AI analyst. Uptime, feature flags, and short links in the same dashboard.
Databuddy is an open-source, cookieless product analytics tool that puts events, funnels, goals, feature flags, short links, and uptime monitoring in one dashboard. It is aimed at startups that want analytics without a consent banner and are willing to run the stack themselves or pay for hosting.
- The storage split fits the workload. Postgres holds app state, ClickHouse takes event volume, and a separate Basket collector handles ingestion, so event traffic can scale apart from the dashboard and API.
- The DQL setup asks for a restricted dql_user with its own CLICKHOUSE_DQL_URL and warns against reusing admin credentials. That is the right default for a feature that runs queries against stored data.
- The AI analyst shows the query behind each answer, so its output can be checked against the data instead of taken on trust.
- Self-hosting turns off hosted billing and the project's own telemetry, which suits a privacy-focused tool. The AGPL-3.0 license, with the scan and pulumi packages under MIT, is clearly stated.
- The self-hosting steps cannot be followed yet. They need a release with the databuddy-init image, and the README says none has been published, so anyone cloning main hits a dead end at the first step.
- The stack is heavy for a small team: Postgres, ClickHouse, Redis, and several app services. Upgrades run interactive db:push prompts and a separate ClickHouse setup script, so someone has to own the operational work.
- Several features need extra setup beyond a default install. The AI analyst needs AI_GATEWAY_API_KEY, website research needs CONTEXT_DEV_API_KEY, DQL needs a separate database user, and email and social login are optional extras.
- With CLICKHOUSE_CLUSTER set, the upgrade script does not add columns or indexes, so you apply them by hand. That is the setup most likely to matter once event volume grows.