Filter exact music identities before lexical and semantic candidate limits

crates/searchi-service/README.md

@@ -41,3 +41,9 @@ Audio timestamps represent segment starts, not exact word locations. Smaller
 windows alone have not demonstrated accurate spoken navigation. No-answer
 calibration, music discrimination, exact spoken positions, process isolation,
 unchanged-vector reuse and production platform checks remain release gates.
+
+Exact artist/title constraints use the core's normalized structured probe keys
+before ANN limits and in the eligibility cache key. Metadata-only writes advance
+the corpus revision and dirty path so warmed readers cannot retain old eligibility.
+Empty metadata-only hybrid requests use core results without model inference.
+These explicit constraints do not establish abstention for free-form topic queries.