Matakupan Labs builds developer libraries that do one job and do it well. The first one is FastAPI-Data, an open-source data microservice starter you can read end to end in an afternoon.
Browse the code Read the docsA small FastAPI service that turns CSV and Parquet files into HTTP endpoints: upload a file once, then profile columns or compute descriptive statistics with plain GET requests. Polars runs lazy, so memory stays flat on files several times larger than RAM-to-disk limits would suggest.
curl -F "file=@transactions.csv" localhost:8000/api/v1/data/upload
# {"dataset_id": "..."}
curl localhost:8000/api/v1/data/{dataset_id}/profile
# dtypes, null counts, cardinality, per column
curl localhost:8000/api/v1/data/{dataset_id}/stats
# mean, std, quartiles — linear interpolation, not "nearest"That is the whole product surface, and it is tested. The repository includes a pytest suite that runs the upload-profile-stats flow end to end.
The open-source edition covers profiling and statistics. When a project grows into something that needs to live behind your own authentication, or hand off background jobs, the complete edition carries that part.
| FastAPI-Data (open source) | Complete edition | |
|---|---|---|
| Upload, profile, stats over HTTP | ✓ | ✓ |
| JWT login for humans | — | ✓ |
| API keys for services | — | ✓ |
| Pandera schema contracts | — | ✓ |
| Celery/Redis background jobs | — | ✓ |
| Docker Compose with healthchecks | — | ✓ |
| Integration-ready billing module | — | ✓ |