KaretThe self-hosted analytics stack

ETL pipelines you draw as a graph and dashboards you describe in YAML, on hardware you own, up in one Docker Compose file.

Get started

View on GitHub

Pipeline visualization

The graph is the editor. Sources, dimensions, mappings and tables are nodes; drag between them to say what feeds what, and the layout is saved with the pipeline so it looks the same to everyone.

Config-driven dashboards

A dashboard is a YAML document and every panel is a DuckDB query, edited in-app with autocomplete and inline validation. Click a chart to filter the rest.

Pipeline versioning

Every save is a numbered version with an author and a diff against what is live, and a restore writes forward rather than rewinding. Runs are pinned to the version that produced them, and a table can be read as of an earlier version or rolled back to one.

Docker Compose and go

One compose file brings up the web app, the worker, Postgres, Valkey and an S3-compatible store, from prebuilt images. Generate four secrets and an admin password hash, create three buckets, and that is the install.