Queries your database and keeps none of the data
Apache Superset is a business intelligence web application from the Apache Software Foundation, with a Python (Flask) server and a React front end. Superset holds none of the data it displays: each chart sends a SQL query to the database that does (Trino, PostgreSQL, or any engine with a Python DB-API driver and a SQLAlchemy dialect). A chart is built without code on a dataset, meaning a table or view whose metrics have been defined, or written by hand in SQL Lab, the built-in editor.
The README argues for this dataset-centric approach and says Superset can "replace or augment" proprietary BI tools. Metabase is aimed at people who want answers without knowing SQL, while Redash starts from a query written by someone who does.
The comparison of open source BI tools places Superset among the other ways to build dashboards without Power BI.
Forty-plus chart types on a lightweight semantic layer
- No-code charts: more than 40 preinstalled visualization types, maps included.
- SQL Lab: schema browser, unrestricted joins, Jinja templating. Queries that outlast a web request's timeout can be handed to Celery workers.
- Lightweight semantic layer: metrics and dimensions are defined once on the dataset, and a "virtual" dataset can be backed by a SQL query.
- Fine-grained access: roles, row-level security filters attached to datasets, and sign-in through OAuth or LDAP.
- Embedding: a REST API, plus a JavaScript SDK that puts a dashboard inside another application with a guest token.
- Alerts and reports: sent by email or Slack, on a SQL condition or a schedule, as PDF, PNG or CSV.
Four moving parts, with Kubernetes for production
Four components to run. The architecture docs list the application, a metadata database (PostgreSQL or MySQL; the default SQLite is not recommended for production), a cache, usually Redis, and Celery workers with their beat scheduler. The last two are optional, but alerts and async queries depend on them. Superset refuses to start without a SECRET_KEY, and the default lean Docker image has no database driver, not even for the metadata store, so the docs recommend building your own image. According to the FAQ, community admins report 8 GB of RAM and 2 vCPUs as enough for a moderately sized instance.
Kubernetes for production. Docker Compose is for trying Superset out: the project does not support it in production because it cannot provide high availability, and the Postgres it starts is not backed up. The recommended route is Kubernetes, which in the docs' words takes "someone who knows Kubernetes". The Helm chart has been deprecated since September 9, 2026 in favor of an operator, at version 0.3.0 since September 30, whose v1alpha1 API may still change between releases.
Built for teams that already have a warehouse. The FAQ describes Superset as a thin layer: when it is slow, the database is what needs tuning. The exploration UI works on a single table or view, so combining tables means preparing one upstream, defining a view, or writing SQL.
No paid edition. All the code is under Apache-2.0. The hosted offering comes from Preset, founded by Superset's creator. The company says its team contributes over 75% of the project's commits, and its free tier stops at five users.
A release every five to seven months. 4.1.0 shipped in November 2024, 5.0.0 in June 2025, 6.0.0 in December 2025 after four release candidates, and 6.1.0 on May 13, 2026. The CVE page lists four vulnerabilities fixed in 6.0.0, including a bypass of SQL Lab's read-only mode on PostgreSQL.