Interface FtsSearchHit

0-based chunk position within the source.

v2.8.0 — content-source kind. Defaults to "md" for backward compat.

1-based ending line in the source (inclusive).

1-based starting line in the source.

Vault-relative path of the source note / PDF.

Flipped BM25 score — higher = better (the underlying FTS5 score is negative; we negate so callers can sort descending).

v7 — exactly 0 is a SENTINEL, not a measurement: it marks a hit found only through the identifier-parts pass ("found, not ranked"). Those hits are appended after every ranked hit, so ordering by score keeps them last; do not read a 0 as "BM25 scored this at zero", and do not filter on score > 0 unless dropping parts-only recall is intended.

Excerpt with matched tokens wrapped in «» and … truncation markers.

v7 caveat — the markers are present only when the query term occurs in the excerpted text. A hit reached through the identifier-parts pass is excerpted from that chunk's content COPY while the term that matched lives in the parts column, so such a snippet is a plain excerpt with no «». Verified against SQLite: snippet() marks nothing when the match is in an unexcerpted column.