fix(backend): 落实 PR #9 评审意见
- 检索调优参数(rrf_k/rerank/rerank_candidates/score_threshold)透传到引擎实际执行 - Recall 去重,避免同一 Note 多 Block 重复导致 Recall 超 1 - RAG 运行改为后台异步执行:创建即 queued + 202,支持取消与 SSE 实时事件 - 数据集元数据校验,坏文件隔离跳过;citation_required 语义修正 - modes 空/重复校验;配置快照记录模型版本与索引元信息 Co-Authored-By: Claude Code <noreply@anthropic.com>
This commit is contained in:
@@ -24,6 +24,7 @@ class RerankerProvider(Protocol):
|
||||
"""统一 Reranker 接口:输入候选块,输出按相关性重排后的候选块。"""
|
||||
|
||||
model_id: str
|
||||
version: str
|
||||
|
||||
async def rerank(self, query: str, candidates: list[RankedCandidate]) -> list[RankedCandidate]: ...
|
||||
|
||||
@@ -32,6 +33,7 @@ class LexicalReranker:
|
||||
"""轻量精排:query 与块正文的词面重叠度,与归一化后的原始分数加权求和。"""
|
||||
|
||||
model_id = "lexical-v1"
|
||||
version = "1"
|
||||
|
||||
def __init__(self, lexical_weight: float = 0.5) -> None:
|
||||
self.lexical_weight = lexical_weight
|
||||
|
||||
Reference in New Issue
Block a user