converting to pydantic-settings

python/ebook_search/rerank.py

@@ -13,8 +13,6 @@ if TYPE_CHECKING:
     from python.ebook_search.search import SearchResult
 
 logger = logging.getLogger(__name__)
-RERANK_SCORE_WEIGHT = 0.7
-HYBRID_SCORE_WEIGHT = 0.3
 
 
 @dataclass(frozen=True)
@@ -41,7 +39,7 @@ def rerank_chunks(query: str, candidates: list[SearchResult], config: RerankConf
         (
             replace(
                 result,
-                score=final_rerank_score(result, scores[result.chunk_id].score, candidates),
+                score=final_rerank_score(result, scores[result.chunk_id].score, candidates, config),
                 rerank_score=scores[result.chunk_id].score,
             )
             for result in candidates
@@ -110,9 +108,14 @@ def clamp_score(score: float) -> float:
     return min(max(score, 0.0), 1.0)
 
 
-def final_rerank_score(result: SearchResult, rerank_score: float, candidates: list[SearchResult]) -> float:
+def final_rerank_score(
+    result: SearchResult,
+    rerank_score: float,
+    candidates: list[SearchResult],
+    config: RerankConfig,
+) -> float:
     """Combine rerank relevance with normalized hybrid retrieval evidence."""
-    return (RERANK_SCORE_WEIGHT * rerank_score) + (HYBRID_SCORE_WEIGHT * normalized_hybrid_score(result, candidates))
+    return (config.score_weight * rerank_score) + (config.hybrid_weight * normalized_hybrid_score(result, candidates))
 
 
 def normalized_hybrid_score(result: SearchResult, candidates: list[SearchResult]) -> float: