Hara for Data Scientists¶
Hara is useful when an analysis needs to become a live, inspectable part of a web application—not just a notebook result or a batch job.
Use a kernel session to keep the values, transformations, and evidence for an experiment together. A browser workspace, a command-line session, and a tool can inspect the same running system.
(def observations [12 19 23 18 28])
(def mean
(/ (reduce + observations) (count observations)))
{:mean mean
:count (count observations)}
The important distinction is that the values are not only printed: they remain available for inspection, controlled changes, visual output, and a durable project record.
A practical workflow¶
- Load or create a bounded dataset.
- Evaluate one transformation at a time.
- Inspect the value and any visual result.
- Keep the verified step in source.
Hara does not replace a specialised Python or R ecosystem. It provides a portable kernel boundary for applications that need their data work, runtime state, and visible output to remain connected.