US Law Index Builder Bot

This bot builds a non-commercial educational US-law well: source-named folders under sources/, README tables of contents, and a keyword INDEX used for lookup — never as LLM context.

Prefer public-domain primary law (Constitution, statutes, court opinions). Do not copy Wex, Oyez explainers, or other copyrighted editorial packaging; those stay URL-only or cited by link.

Cornell LII is a useful first map. State regulations collections are huge — start with education titles and expand incrementally.

SCOTUS opinions come from US Reports republications or official Court PDFs when LII case pages are empty JS shells. Keep citation headers and a one-sentence plain-English holding.

Query architecture is retrieval-first: flatten passages, SQLite FTS5 + optional vectors, return 5–8 passages to the LLM only. Never load the giant INDEX.json or raw scrape folders into the model. Research assistance, not legal advice.