Add viw_kg_form_neighbors: 1-hop subgraph per form (GraphRAG expand step)

Materialize each form's L2 neighbors (upstream/downstream documents) and L3 neighbors (referenced entity tables) so the retriever can expand a vector-recalled form-card into a subgraph context via a single JOIN. 702 forms have downstream, 476 upstream, 1467 reference entity tables.

Add viw_kg_form_neighbors: 1-hop subgraph per form (GraphRAG expand step)
Materialize each form's L2 neighbors (upstream/downstream documents) and L3 neighbors (referenced entity tables) so the retriever can expand a vector-recalled form-card into a subgraph context via a single JOIN. 702 forms have downstream, 476 upstream, 1467 reference entity tables.
zichun
1 parent 11f33a4d
Showing 2 changed files with 50 additions and 4 deletions
sql/README.md
sql/viw_kg_form_neighbors.sql
@@ -33,14 +33,17 @@ mysql --no-defaults -h127.0.0.1 -P33307 -uroot -plocal xlyweberp_saas
    1597 边；目标多为 `ele*`/`sis*` 主数据字典。`bTgtMaster` 标记引用主数据。
 5. **`viw_kg_form_card`** — 每表单一段可向量化的结构化卡片文本 `sCard`（合并节点+上下游+字段中文名）。
 6. **`viw_kg_field_dict`** — 中文名→字段→表 术语字典（38342 条），带外键 join 提示，供 NL2SQL 接地。
-7. **`export_form_cards.sh`** — 导出 1748 张卡片为 JSONL（供灌向量库）。生成文件 gitignore，按需重跑。
+7. **`viw_kg_form_neighbors`** — 每表单的 1-hop 子图（上/下游单据 + 引用实体表），GraphRAG 的「图扩展」步物化。
+8. **`export_form_cards.sh`** — 导出 1748 张卡片为 JSONL（供灌向量库）。生成文件 gitignore，按需重跑。
 ## 一键重建
 ```sh
-DB="mysql --no-defaults -h127.0.0.1 -P33307 -uroot -plocal xlyweberp_saas"
-for v in viw_ai_useful_forms viw_kg_edge_flow viw_kg_form viw_kg_edge_ref viw_kg_form_card viw_kg_field_dict; do
-  $DB < "$v.sql"
+# 注意：zsh 不对未加引号的变量做分词，用数组
+DB=(mysql --no-defaults -h127.0.0.1 -P33307 -uroot -plocal xlyweberp_saas)
+for v in viw_ai_useful_forms viw_kg_edge_flow viw_kg_form viw_kg_edge_ref \
+         viw_kg_form_card viw_kg_field_dict viw_kg_form_neighbors; do
+  "${DB[@]}" < "$v.sql"
 done
 ./export_form_cards.sh form_cards.jsonl   # → 1748 行 JSONL
 ```
+-- ============================================================
+-- 视图：viw_kg_form_neighbors —— ERP 知识图谱·表单一跳子图 (GraphRAG 扩展步)
+-- 每个表单一行，物化它的 1-hop 邻域：下游单据 / 上游单据 / 引用的实体表。
+-- 这是 GraphRAG 检索的「图扩展」步骤：向量召回一张表单卡片后，直接 JOIN 本视图
+-- 取出它的 L2(单据流转) + L3(实体引用) 邻居，拼成子图上下文喂给 LLM。
+--
+--   L2 下游 sDownIds/sDownForms  = 本单据可生成的目标单据(viw_kg_edge_flow, src侧, 非自环)
+--   L2 上游 sUpForms            = 可生成本单据的上游单据(edge_flow, tgt侧)
+--   L3 引用 sRefTables          = 本表数据源引用到的实体/主数据表(viw_kg_edge_ref, 去重)
+--
+-- 节点粒度 = 表单 sFormId (1748)。L2 按菜单粒度，L3 按数据源表粒度。
+-- 表内 GROUP_CONCAT 邻居列表都很短(下游≤5, 引用≤18)，默认 max_len 足够。
+-- ============================================================
+CREATE OR REPLACE VIEW viw_kg_form_neighbors AS
+SELECT
+    fm.sFormId,
+    fm.sModuleId,
+    fm.sFormTitle,
+    fm.sDomain,
+    fm.sDataSource,
+    fm.iUpstream,
+    fm.iDownstream,
+    up.names   AS sUpForms,      -- 上游单据名
+    dn.names   AS sDownForms,    -- 下游单据名
+    dn.ids     AS sDownIds,      -- 下游单据 moduleId(便于程序继续扩展)
+    rf.tables  AS sRefTables     -- 引用的实体/主数据表
+FROM viw_kg_form fm
+LEFT JOIN (
+    SELECT sSrcModuleId AS mid,
+           GROUP_CONCAT(DISTINCT sTgtName ORDER BY sTgtName SEPARATOR '，') AS names,
+           GROUP_CONCAT(DISTINCT sTgtModuleId SEPARATOR ',')                AS ids
+    FROM viw_kg_edge_flow WHERE bSelfLoop = 0 GROUP BY sSrcModuleId
+) dn ON dn.mid = fm.sModuleId
+LEFT JOIN (
+    SELECT sTgtModuleId AS mid,
+           GROUP_CONCAT(DISTINCT sSrcName ORDER BY sSrcName SEPARATOR '，') AS names
+    FROM viw_kg_edge_flow WHERE bSelfLoop = 0 GROUP BY sTgtModuleId
+) up ON up.mid = fm.sModuleId
+LEFT JOIN (
+    SELECT sSrcTable,
+           GROUP_CONCAT(DISTINCT sTgtTable ORDER BY sTgtTable SEPARATOR ',') AS tables
+    FROM viw_kg_edge_ref GROUP BY sSrcTable
+) rf ON rf.sSrcTable = LOWER(fm.sDataSource);