Hybrid keyword and semantic search within one filing or transcript. When too few excerpts match every query word, the search automatically broadens to match any of the words — precise matches keep their rank and broader hits only fill the remainder.

Search-result envelope (additive over the plain list, #7056): top-k retrieval has no meaningful "total matches", so instead of a fake total the response echoes the applied limit — count == limit means the result set was capped and a larger limit may surface more excerpts. count < limit means no further excerpts cleared relevance at this limit — EXCEPT on endpoints that take maxResultsPerCompany, where a non-zero per-company cap can drop excerpts below the overall limit, so a short page there is not proof of exhaustion.