// the find
apache/superset
Apache Superset is a Data Visualization and Data Exploration Platform
Superset is a self-hosted BI platform for building dashboards and running SQL queries against virtually any SQL-speaking database. It's a mature, Apache-governed project targeting data teams that want to own their analytics stack instead of paying Tableau or Looker. At 73k stars and active daily commits, it's the most serious open-source BI option available.
The database connector breadth is genuinely impressive — 60+ adapters covering everything from DuckDB to SAP HANA, all pluggable via SQLAlchemy dialects. The SQL Lab editor is the best part of the product: async queries, query history, results caching, and Jinja templating that lets you parameterize queries without building a full semantic layer. The semantic layer (datasets with virtual columns and metrics) is a real productivity win — define a metric once, reuse it across every chart instead of copy-pasting SQL. The Helm chart and Docker Compose setup are well-maintained, which matters a lot for a project with this many moving parts (Flask, Celery, Redis, Postgres).
The feature flag system is a mess — there are ~80 flags, many of which have been 'experimental' for years, and the docs lag behind which ones are actually safe to enable in production. Upgrades are painful: the migration history is long and occasionally has conflicts, and the UPDATING.md changelog is required reading before every version bump. The permission model (FAB-based RBAC) is notoriously confusing — getting row-level security right requires understanding a non-obvious combination of roles, datasets, and RLS filters that isn't well documented. Self-hosting the full stack (Flask app + async workers + websocket server + caching layer) is a significant ops burden; if your team doesn't already run Celery and Redis, plan for that overhead before committing.