ymq 789960a1cf refactor(schema): rag 模块全部数据表加 rag_ 前缀(模块跨应用复用防冲突)
- models/ 重建为正确表定义格式(summary+fields+indexes),覆盖全部 11 个表
  修正原 models/ 误放 CRUD 格式副本(缺 summary)导致 json2ddl 100% 失败被静默忽略
- json/ 6 个 CRUD 定义 tblname 加前缀 + 文件改名
- 129 处 SQL/sor 表名加 rag_ 前缀(URL 路径 knowledge_bases_list 保持不变)
- 新增 init/migrate_rag_prefix.sql 幂等 RENAME TABLE 迁移
2026-08-25 14:24:37 +08:00

106 lines
4.1 KiB
Plaintext

import os
ns = params_kw.copy()
kb_id = ns.get('kb_id', '')
if not kb_id:
return {"widgettype": "urlwidget",
"options": {"url": entire_url('/rag/knowledge_bases_list/index.ui')}}
env = request._run_ns
userorgid = await env.get_userorgid()
deleted = {"documents": 0, "chunks": 0, "tags": 0, "media_tags": 0,
"files": 0, "vdb": 0, "graph": 0, "entities": 0}
async with get_sor_context(env, 'rag') as sor:
kb = await sor.sqlExe(
"SELECT id, name FROM rag_knowledge_bases WHERE id=${id}$ AND (org_id=${org_id}$ OR org_id IS NULL)",
{"id": kb_id, "org_id": userorgid})
if not kb:
return {"widgettype": "urlwidget",
"options": {"url": entire_url('/rag/knowledge_bases_list/index.ui')}}
docs = await sor.sqlExe(
"SELECT id, file_path FROM rag_documents WHERE kb_id=${kb_id}$",
{"kb_id": kb_id})
doc_ids = [d.id for d in docs]
# 1. VDB cleanup
if doc_ids:
chunks = await sor.sqlExe(
"SELECT vector_id FROM rag_document_chunks WHERE doc_id IN (${ids}$) AND vector_id IS NOT NULL AND vector_id != ''",
{"ids": doc_ids})
vector_ids = [c.vector_id for c in chunks]
if vector_ids:
try:
client = StreamHttpClient()
await client.request('POST',
'https://vectordb.opencomputing.net:10443/v1/delete',
json={"colname": kb_id, "ids": vector_ids})
deleted["vdb"] = len(vector_ids)
except Exception as e:
info(f'delete_kb vdb error: {e}')
# 2. Graph DB cleanup
try:
client2 = StreamHttpClient()
await client2.request('POST',
'http://localhost:9092/api/graph/delete',
json={"graph": kb_id})
deleted["graph"] = 1
except Exception as e:
info(f'delete_kb graph error: {e}')
# 3. Entities & relations
er = await sor.sqlExe(
"SELECT COUNT(*) AS cnt FROM rag_entity_relations WHERE kb_id=${kb_id}$",
{"kb_id": kb_id})
await sor.sqlExe("DELETE FROM rag_entity_relations WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
deleted["entities"] += er[0].cnt if er else 0
ent = await sor.sqlExe(
"SELECT COUNT(*) AS cnt FROM rag_entities WHERE kb_id=${kb_id}$",
{"kb_id": kb_id})
await sor.sqlExe("DELETE FROM rag_entities WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
deleted["entities"] += ent[0].cnt if ent else 0
# 4. Chunks
ch = await sor.sqlExe(
"SELECT COUNT(*) AS cnt FROM rag_document_chunks WHERE doc_id IN (${ids}$)",
{"ids": doc_ids})
await sor.sqlExe("DELETE FROM rag_document_chunks WHERE doc_id IN (${ids}$)", {"ids": doc_ids})
deleted["chunks"] = ch[0].cnt if ch else 0
# Media tags
mt = await sor.sqlExe("SELECT COUNT(*) AS cnt FROM rag_media_tags WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
await sor.sqlExe("DELETE FROM rag_media_tags WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
deleted["media_tags"] = mt[0].cnt if mt else 0
# Tags
t = await sor.sqlExe("SELECT COUNT(*) AS cnt FROM rag_tags WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
await sor.sqlExe("DELETE FROM rag_tags WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
deleted["tags"] = t[0].cnt if t else 0
# Documents
await sor.sqlExe("DELETE FROM rag_documents WHERE kb_id=${kb_id}$", {"kb_id": kb_id})
deleted["documents"] = len(doc_ids)
# KB itself
await sor.sqlExe("DELETE FROM rag_knowledge_bases WHERE id=${id}$", {"id": kb_id})
# Physical files — paths are like /44/126/174/61/file.mp4
for d in docs:
fp = d.file_path or ''
if fp and fp.startswith("/"):
try:
real_path = env.realpath(fp)
if os.path.exists(real_path):
os.remove(real_path)
deleted["files"] += 1
except Exception:
pass
info(f'delete_kb {kb_id}: {deleted}')
return {"widgettype": "urlwidget",
"options": {"url": entire_url('/rag/knowledge_bases_list/index.ui')}}