Hybrid RAG: BM25+Dense (sqlite-vec/BGE-M3) + cross-encoder reranker (bge-reranker-v2-m3)

src/kpaa/guides/index.py

@@ -159,6 +159,40 @@ def search(
     ]
 
 
+def get_chunks(
+    chunk_ids: Iterable[str], *, db_path: Path | None = None
+) -> dict[str, GuideChunk]:
+    """chunk_id 리스트 → {chunk_id: GuideChunk} dict (누락된 id 는 dict 에 없음).
+
+    Hybrid retrieval 의 dense 검색 결과(chunk_id 만 반환)를 GuideChunk 로 복원할 때 사용.
+    """
+    ids = [str(i) for i in chunk_ids]
+    if not ids:
+        return {}
+    path = db_path or default_db_path()
+    if not path.exists():
+        return {}
+    placeholders = ",".join("?" * len(ids))
+    conn = _connect(path)
+    try:
+        rows = conn.execute(
+            f"SELECT * FROM guides WHERE chunk_id IN ({placeholders})", ids
+        ).fetchall()
+    finally:
+        conn.close()
+    return {
+        r["chunk_id"]: GuideChunk(
+            chunk_id=r["chunk_id"], doc_id=r["doc_id"],
+            doc_title=r["doc_title"], doc_date=r["doc_date"],
+            section=r["section"], chunk_no=r["chunk_no"],
+            body=r["body"], pages=r["pages"] or "",
+            source_pdf=r["source_pdf"],
+            chunk_context=(r["chunk_context"] if "chunk_context" in r.keys() else "") or "",
+        )
+        for r in rows
+    }
+
+
 def count(db_path: Path | None = None) -> int:
     path = db_path or default_db_path()
     if not path.exists():