An in-memory analysis pipeline

From records to a result

A small analysis pipeline has distinct stages even when all its data is already in memory. Using the vocabulary from the start of this part:

1. Validate: decide which input records satisfy the data contract.
2. Transform and summarize: place values into a useful result shape and combine repeated observations.
3. Present: turn the structured summary into consistently ordered rows.

Keeping these stages separate makes intermediate values easy to inspect. It also gives an agent a clearer specification than a single request to “analyse this data”.

You can check your current points from the blue blob in the bottom-right corner of the page.