feat(rag): dense+BM25 hybrid on MilvusClientV2, drop SDK path

Replace legacy MilvusServiceClient knowledge search/write with a single
MilvusClientV2 hybrid store (BM25 function + dense ANN + RRFRanker).
Use collection biz_hybrid and require knowledge reindex.
This commit is contained in:
zhuyongxin
2026-07-27 18:49:20 +08:00
parent 376ad0c241
commit f035538531
22 changed files with 813 additions and 936 deletions
@@ -32,6 +32,26 @@
- 分块去重与 hybrid 同规划、分里程碑交付(先共用地基,再开 hybrid)
```
## 实现状态(2026-07-27)
| 里程碑 | 状态 | 说明 |
|---|---|---|
| 交付 1 chunk 身份/去重/SearchPort | **已完成并归档** | `2026-07-27-rag-chunk-evidence-identity-dedup` |
| 交付 2a 应用层 multi-path+RRF | **已完成并归档** | `2026-07-27-rag-hybrid-search-rrf`(已被 2b 取代为生产路径) |
| 交付 2b 真 BM25 hybrid + 废弃 SDK | **已完成并归档** | `2026-07-27-rag-bm25-hybrid-drop-sdk` |
**当前生产知识路径:**
```text
VectorIndexService / VectorSearchService
-> MilvusHybridKnowledgeStore (MilvusClientV2 only)
collection: milvus.collection (default biz_hybrid)
mode: retrieval.search.mode = dense | hybrid
hybrid: dense ANN + BM25 sparse ANN + RRFRanker
```
**运维必做:** 全量重灌知识库到 `biz_hybrid`;旧 `biz` collection 不再被知识路径使用。
---
## 1.1 交付拆分:分块去重 + Hybrid 同规划