DataEase: a self-hosted BI platform you install with one curl command

The gap DataEase fills between raw SQL and a finished dashboard

Most teams that need a dashboard face the same fork. Write the charts yourself against a charting library, and you own every axis label, filter and permission check forever. Buy a hosted BI product, and your query results leave your network. DataEase targets the middle: a self-hosted platform where the connection to the database, the chart construction and the sharing are all part of one application.

The README frames the audience as people who want to "快速分析数据并洞察业务趋势" without writing the visualization layer. In practice that means an analyst or operations person who can point at a table, drag a dimension onto an axis, and publish the result. The README also states the deployment options explicitly: a quick-start script for evaluation, and an offline installation package that the project recommends for production.

It is a Java application. The repository's top level holds a core directory for the backend, a drivers directory, an installer directory and a sdk directory, which tells you the project is organized as a deployable server rather than a library you embed. If you were hoping to import DataEase into your own application as a rendering component, the layout does not suggest that is the intended use.

How DataEase connects sources and turns them into charts

The README lists the data sources in three groups. OLTP databases include MySQL, Oracle, SQL Server, PostgreSQL, MariaDB, Db2, TiDB and MongoDB-BI. OLAP databases include ClickHouse, Apache Doris, Apache Impala and StarRocks. Data warehouses and lakes are represented by Amazon RedShift, and files by Excel and CSV. There is also an API data source category.

Underneath, the README names the processing stack: Apache Calcite and Apache SeaTunnel. Calcite is a SQL parsing and planning framework, and SeaTunnel is a data integration engine. Reading those two together with the source list, the plausible architecture is that DataEase translates the drag-and-drop query you build in the UI into SQL that Calcite can plan and push down to the source, while SeaTunnel handles the movement of data for sources that cannot be queried directly or for extracted datasets. The README does not spell out the boundary between the two, so treat that as an inference from the named components rather than documented behaviour.

The frontend is Vue.js with Element, and the charts come from AntV. The application database is MySQL. The Dockerfile confirms the runtime shape: an Alpine image with OpenJDK 21, a jar at /opt/apps/app.jar, drivers mounted from the drivers directory, map files copied from mapFiles, and RUNNING_PORT set to 8100. A HEALTHCHECK probes that port with nc every 15 seconds. That is a conventional single-container Spring Boot deployment, and it also tells you the map rendering data ships inside the image, which matters if your dashboards use geographic charts.

Installing DataEase and building a first chart

The README's quick start is a single command, run as root on a Linux server with at least 2 cores and 4 GB of memory. It streams a shell script from an Aliyun OSS bucket and executes it.

curl -sSL https://dataease.oss-cn-hangzhou.aliyuncs.com/quick_start_v2.sh | bash

When it finishes, the README gives the default credentials: username admin, password DataEase@123456. Log in and change that password before the instance is reachable from anywhere you do not control. Piping a remote script into bash also means you are trusting whatever that URL returns at the moment you run it, which is a reasonable thing to do on a throwaway VM and a poor thing to do on a server holding customer data.

For production the README points at a different path, the offline installation and upgrade documentation under dataease.io. That package is the one to use when the host has no outbound internet access, which is common for the databases DataEase is meant to query.

The repository's own Dockerfile shows the container contract if you build the image yourself. It expects the backend jar at core/core-backend/target/CoreApplication.jar, mounts drivers into /opt/dataease2.0/drivers/, and sets the port through an environment variable.

ENV JAVA_APP_JAR=/opt/apps/app.jar
ENV RUNNING_PORT=8100
ENV JAVA_OPTIONS="-Dfile.encoding=utf-8 -Dloader.path=/opt/apps -Dspring.config.additional-location=/opt/apps/config/"

After login, the first real task is creating a data source: pick the database type, supply host, port, database name and credentials, and test the connection. Then create a dataset from a table or a custom SQL query, and drag fields into a chart. The README does not walk through these screens in text, so the online documentation at dataease.cn/docs/v2/ is where the field-by-field detail lives.

What the README does not tell you about running DataEase

The README is a landing page, not an operations manual, and the gaps are worth naming before you commit.

Backup and restore are not covered. There is a MySQL instance behind the application, and the repository layout suggests configuration and metadata live there, but the README says nothing about how to dump it or how to restore a dashboard after a failed upgrade. The offline installation page is described as covering installation and upgrade; rollback is not mentioned in the README at all.

Resource sizing is a single line: 2 cores and 4 GB. That is a floor for the application container, not a statement about what happens when a ClickHouse query returns ten million rows into an AntV chart. The README gives no guidance on query timeouts, result limits or caching.

The licence is GPL-3.0. If you plan to embed DataEase inside a product you distribute, or to modify it and ship the result under different terms, that licence governs what you must do with your changes. The README states the licence and links to the text; it does not offer an alternative commercial licence, and nothing in the repository suggests one exists. For internal dashboards this is a non-issue. For a commercial product built on top of the code, it is the first thing to resolve.

Finally, the primary documentation language is Chinese. The README links translations into English, Traditional Chinese, Japanese, Portuguese, Arabic, German, Spanish, French, Korean, Indonesian and Turkish, but the online docs link points at dataease.cn/docs/v2/, and the depth of the translated material is not something the README establishes.

DataEase compared with Metabase and Superset

The two comparisons people search for are Metabase and Apache Superset, and the difference is mostly about where the query logic lives.

Metabase is a JVM application too, but its model is question-first: you ask a question in a visual builder, save it, and assemble saved questions into a dashboard. It leans toward letting non-technical users explore a single connected database without a modelling layer. DataEase leans toward dataset-first: you define a dataset, then build charts from its fields. That extra step is friction for a quick ad-hoc question and an advantage when the same metric is reused across many dashboards.

Superset is Python, and its centre of gravity is SQL. Analysts write queries in SQL Lab, and the visualization layer sits on top of those queries. Its chart catalogue is large and its access to the underlying SQL is direct. DataEase's drag-and-drop path means less SQL, and the README's mention of Calcite and SeaTunnel suggests the platform is doing query translation on your behalf. That is convenient until you need a query the translation layer cannot express, at which point the escape hatch matters and the README does not describe one.

A practical split: Superset if your analysts write SQL and want the full surface of it. Metabase if you want the shortest path from a connected database to a shared question. DataEase if you want a self-hosted platform with a defined dataset model, a broad connector list including OLAP engines and API sources, and a documented offline installation path for air-gapped hosts.

Maintenance, releases and the SQLBot integration

The most recent release listed is v2.10.26, dated 2026-07-30, and the last push to the repository was on the same day. Two earlier releases, v2.10.25 and v2.10.24, landed in June 2026. The README describes the release rhythm as monthly iteration, and the three release dates are consistent with that.

The repository is not archived, and the default branch is dev-v2, which means the v2 line is where development happens. Upgrade cost is the thing to plan for. Monthly releases on a self-hosted BI server mean you either track them or you fall behind and face a large jump later. The README recommends the offline installation package for production and links documentation covering installation and upgrade, but it does not describe a rollback procedure, so an upgrade plan that assumes you can reverse a bad release is not supported by anything in this repository's README.

The SQLBot integration is worth a separate note. The README lists it as an advantage, describing integration with SQLBot for "智能问数", natural-language querying. SQLBot is a separate repository under the same GitHub organization, not a component inside this one. Enabling it means deploying and operating a second service, and the README does not describe the configuration that connects the two.

On licence: GPL-3.0 is a copyleft licence. Reading it as an engineer rather than a lawyer, the practical question is whether you are distributing modified DataEase code, and if so, what the licence requires you to publish. Internal deployment behind a company firewall does not raise that question in the same way. Distribution does.

Editorial conclusion

Adopt DataEase if you want a self-hosted dashboard layer over databases you already run, and if your team is comfortable with a Chinese-first documentation set and a GPL-3.0 licence. Do not adopt it if you need a permissive licence for a closed-source product, or if you expect the README to tell you how to upgrade or roll back a deployment; it does not. Verify first that the data source you care about appears in the supported list, and that you can run the offline installation package rather than the quick-start script before you point it at production data.

Frequently asked questions

Official sources

1. Official documentation
2. Official README
3. Project repository
4. Release notes