docs(mvp): update RAG docs for py-rag extraction

- rewrite RAG architecture doc for py-rag service contract mapping, ingest and rebuild ops
- refresh observability/trace doc for post-L0 single-attempt semantics
- add py-rag chain exploration note; mark superseded Milvus/L0 notes with status banners
- close ISS-017 (superseded by L0 sinking); mark knowledge_domain table orphaned
- refresh architecture/mvp/engineering indexes
This commit is contained in:
zhuyongxin
2026-09-30 17:03:50 +08:00
parent 9cf162482d
commit 473bb5d004
16 changed files with 597 additions and 343 deletions
@@ -1,5 +1,12 @@
# 混合检索上线之后:为什么还要统一 qualityScore,以及上一代后处理错在哪
> **现状说明(2026-09-29)**:本文讨论的后处理排序/去重/判级仍在 Java 侧
> (`KnowledgeEvidencePostProcessor` / `RetrievalScoreNormalizer`);
> 但分数语义已变:py-rag 服务端返回 rerank 绝对分(scoreLabel=`RERANK`,[0,1] 越大越好),
> quality 直传,不再走本文所述 dense L2 / hybrid rank 归一化分支(分支保留作兼容)。
> 判级阈值 0.75/0.5 与 py-rag 契约一致。当前架构见
> [../../architecture/RAG知识检索架构.md](../../architecture/RAG知识检索架构.md)。
**日期**:2026-07-28
**范围**:hybrid 检索后的分数语义、后处理排序、相关度闸门、scoreLabel 约定
**读者**:已经(或准备)上 dense+BM25+RRF,却发现「召回变了、质量判断还拧着」的工程同学