Where this leaves you

The corpus is fetched from the source, hashed, cut verbatim, dated, and held to its sources by a test. The index and the table each have a manifest, and a change to the corpus that was not pushed to them is red. Where a model chooses what to store, code decides whether it may be stored, by the one test a model cannot argue with: is this string in that string.

Everything the pipeline reads now exists and can be trusted. The next chapter puts the stages together.