Yes. The repository is MIT licensed, which permits commercial use, modification, and redistribution. Keep the license notice, and that is the whole obligation.
To DATA_DIR/uploads, which defaults to ./data/uploads. The directory is git-ignored. Nothing is uploaded anywhere else, there is no outbound telemetry in the codebase, and no account or email is required to run it.
Two practical reasons. Polars parses CSV and Parquet with the same API, so the upload endpoint does not need a format branch for every file type. And the lazy query plan means column selection and schema inference happen before data materializes, which keeps the code short on the ingestion side. If a project needs pandas specifically, the JSON contract stays identical while the engine swaps.
For profiling and descriptive statistics, the current collect-once design materializes one file in RAM at a time. Size the machine for the largest file you expect and it goes comfortably past a million rows. After that point, background jobs and schema contracts start to matter, and that is what the complete edition covers.
Authentication split the right way (JWT for humans, static API keys for machines), Pandera schema contracts, Celery/Redis job handling tuned for clean SIGTERM shutdown, a billing module pre-integrated with Polar.sh, and Docker multi-stage builds with healthchecks.
No. The complete edition is a one-time purchase of a codebase, self-hosted, with no seat limits and no per-server fee. It stays yours. If you want it, the checkout runs through Polar.sh.
Fauzi Matakupan, a statistics and data science student in Semarang, Indonesia. Matakupan Labs is the umbrella for the libraries built and published there. Questions reach me at fauzimatakupan@gmail.com.