ADR 0006 — Ingest the corpus from the sibling rather than re-crawling it

ga · floor-software-engineer

Decision

Ingest the corpus from the sibling rather than re-crawling it

Why

The bytes are already on disk, fetched by the process that owns fetching. Re-crawling to obtain them again spends hours and a rate limit to reach identical content, and puts a second crawl policy against hosts the sibling is already polite to.

Measured

7,783 documents, 115.8 MB, 41 seconds. 2,384 files carried no manifest row -- 1,350 being the API reference -- recovered by a path derivation verified against a live 200 text/markdown.

Inputs and outputs

This decision declares none. Every name in this section, when present, resolves against the capability vocabulary or the floor registry -- a typo fails the generator rather than rendering.

Where it is