feat(ebook): add phrase matching display and update search result structure

python/ebook_search/search.py

@@ -23,7 +23,7 @@ from python.ebook_search.embeddings import MODEL_DIMENSIONS, embed_query, get_em
 from python.ebook_search.protected_phrases.lib import (
     HydratedPhraseMatch,
     detect_protected_phrases_for_query,
-    phrase_hit_counts_for_chunks,
+    phrase_hits_for_chunks,
 )
 from python.ebook_search.rerank import rerank_chunks
 from python.ebook_search.timing import RuntimeStep, timed_result
@@ -58,6 +58,7 @@ class SearchResult:
     fused_score: float | None = None
     rerank_score: float | None = None
     phrase_hit_count: int = 0
+    matched_phrases: tuple[str, ...] = ()
     source_author: str | None = None
     chapter_title: str | None = None
     page_label: str | None = None
@@ -201,20 +202,24 @@ def apply_phrase_mention_boosts(
     chunk_ids = [candidate.chunk_id for candidate in candidates]
     try:
         with Session(engine) as session:
-            hit_counts = phrase_hit_counts_for_chunks(session, chunk_ids=chunk_ids, phrase_ids=phrase_ids)
+            phrase_hits = phrase_hits_for_chunks(session, chunk_ids=chunk_ids, phrase_ids=phrase_ids)
     except SQLAlchemyError as error:
         logger.warning("ebook_phrase_boost_unavailable error=%s", error)
         return candidates
 
-    if not hit_counts:
+    if not phrase_hits:
         return candidates
 
+    hit_counts = {
+        chunk_id: sum(hit.mention_count for hit in chunk_hits) for chunk_id, chunk_hits in phrase_hits.items()
+    }
     boosted = [
         replace(
             candidate,
             score=candidate.score + (hit_counts.get(candidate.chunk_id, 0) * phrase_hit_boost),
             fused_score=boosted_fused_score(candidate, hit_counts.get(candidate.chunk_id, 0), phrase_hit_boost),
             phrase_hit_count=hit_counts.get(candidate.chunk_id, 0),
+            matched_phrases=tuple(hit.phrase_text for hit in phrase_hits.get(candidate.chunk_id, ())),
             rank_source=phrase_rank_source(candidate.rank_source, hit_counts.get(candidate.chunk_id, 0)),
         )
         for candidate in candidates