Search the public-record substrate

What this searches. Every record LexDOGE has ingested — council legislation, contracts, budget line items, FOIA-released documents, and the per-page chunks behind every PDF. The query runs against Postgres tsvector indexestsvectorA Postgres data type that stores a document's text in a form optimized for full-text search. LexDOGE generates one tsvector per row across every searchable table (legislation, contracts, budgets, documents, chunks) and indexes them with GIN, so the /search page returns matches in milliseconds even at millions of rows., so quoted phrases, boolean OR/AND/NOT, and negation (prefix with -) all work.

Examples: "sole source", pension OR opeb, contract -amendment, vendor concentration.

How ranking works. Postgres' cover-density rankingts_rank_cdPostgres' cover-density ranking function. Given a search query and a tsvector, it returns a relevance score that weights matches by how close the query terms appear in the source text. LexDOGE sorts search results by ts_rank_cd descending, so the most relevant matches surface first. scores each match by how close the query terms appear in the source text. The snippet you see is generated by ts_headlinets_headlineThe Postgres function that extracts a short snippet from a document with the matching query terms highlighted in <b>…</b> tags. LexDOGE uses ts_headline on every search result so you can see which words matched in context, not just that the document matched., which extracts the most relevant fragment and highlights matching words with <b> tags.

Coverage. The index covers every record in the substrate listed in §1 Foundation. When a new source comes online (KPPA, EMMA, KORA, BoardDocs), it's included automatically — no schema migration is required to add new content to an existing table.