// the find
running-elephant/datart
Datart is a next generation Data Visualization Open Platform
Datart is a self-hosted BI/data visualization platform built primarily in Java (Spring Boot) with a React/TypeScript frontend. It covers the full spectrum from individual charts to dashboards, storyboards, and scheduled reports. Aimed at teams that want a Metabase/Superset alternative they can embed or extend without paying for a SaaS license.
1. Plugin architecture is real, not just claimed — Source, Chart, and Visualization layers each expose a pluggable interface, so you can add a custom data connector or chart type without forking. 2. Row-level permission model via RelSubjectColumns and SQL variables is more granular than most open-source BI tools, which either skip it or bolt it on awkwardly. 3. Uses Apache Calcite for SQL generation, which means dialect translation and function normalization are handled properly rather than being a pile of regex substitutions. 4. Gossip-protocol cluster support (GossiperManager) suggests someone actually thought about horizontal scaling, not just single-node deployment.
1. Last meaningful commit was February 2025 and activity has been sparse for over a year — the issues tracker has open bugs with no responses, which is a bad sign for a platform you'd put production dashboards on. 2. Documentation is almost entirely in Chinese; the English README is thin and the deployment guide assumes you already know what you're doing. Non-Chinese-speaking teams will hit walls fast. 3. The Java backend is a MyBatis + Spring Boot monolith with generated mapper files — the kind of codebase where adding a new entity requires touching six files by hand, and the generated SQL provider pattern doesn't age well as query complexity grows. 4. No mention of incremental query caching or query federation — for large datasets you're hitting the source directly every time, which will surface as a performance problem at scale.