AIRegulation/AIRegulation-Deployment
7
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
wangwei
2026-04-23 14:50:24 +08:00
parent 448e078d99
commit af0eb007a0
5 changed files with 327 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

70
scripts/05_init_db.sh
View File

@@ -38,42 +38,80 @@ info "Step 2/3：初始化 Milvus Collections..."
if docker compose ps milvus | grep -q "healthy"; then
docker compose run --rm --no-deps compliance-backend \
python3 -c "
import asyncio
from pymilvus import connections, Collection, CollectionSchema, FieldSchema, DataType, utility
connections.connect(host='milvus', port='19530')
print('Milvus 连接成功')
def create_collection(name, description):
def create_regulation_chunks():
'''
regulation_chunks：法规专用 collection，带 sparse 向量和条款号字段。
开发阶段每次运行均强制重建（保持 schema 最新）。
'''
name = 'regulation_chunks'
if utility.has_collection(name):
utility.drop_collection(name)
print(f' 已删除旧 collection {name}（schema 升级）')
fields = [
FieldSchema(name='pk', dtype=DataType.VARCHAR, is_primary=True, max_length=128),
FieldSchema(name='file_id', dtype=DataType.VARCHAR, max_length=128),
FieldSchema(name='workspace_id', dtype=DataType.VARCHAR, max_length=128),
FieldSchema(name='chunk_idx', dtype=DataType.INT64),
FieldSchema(name='content', dtype=DataType.VARCHAR, max_length=65535),
FieldSchema(name='dense_vec', dtype=DataType.FLOAT_VECTOR, dim=1024),
FieldSchema(name='sparse_vec', dtype=DataType.SPARSE_FLOAT_VECTOR),
FieldSchema(name='clause_no', dtype=DataType.VARCHAR, max_length=64, default_value=''),
FieldSchema(name='article_no', dtype=DataType.VARCHAR, max_length=128, default_value=''),
FieldSchema(name='regulation_name', dtype=DataType.VARCHAR, max_length=512, default_value=''),
FieldSchema(name='metadata', dtype=DataType.JSON),
]
schema = CollectionSchema(fields, description='法规条款向量库（含稀疏向量和条款号）')
col = Collection(name, schema)
col.create_index('dense_vec', {
'metric_type': 'COSINE',
'index_type': 'HNSW',
'params': {'M': 16, 'efConstruction': 200}
})
col.create_index('sparse_vec', {
'metric_type': 'IP',
'index_type': 'SPARSE_INVERTED_INDEX',
'params': {'drop_ratio_build': 0.2}
})
col.load()
print(f' Collection {name} 创建完成（dense+sparse 双索引）')
def create_simple_collection(name, description):
'''doc_chunks / case_library：通用 collection，仅 dense 向量。'''
if utility.has_collection(name):
print(f' Collection {name} 已存在，跳过')
return
fields = [
FieldSchema(name='id', dtype=DataType.VARCHAR, is_primary=True, max_length=128),
FieldSchema(name='file_id', dtype=DataType.VARCHAR, max_length=128),
FieldSchema(name='pk', dtype=DataType.VARCHAR, is_primary=True, max_length=128),
FieldSchema(name='file_id', dtype=DataType.VARCHAR, max_length=128),
FieldSchema(name='workspace_id', dtype=DataType.VARCHAR, max_length=128),
FieldSchema(name='chunk_idx', dtype=DataType.INT64),
FieldSchema(name='content', dtype=DataType.VARCHAR, max_length=65535),
FieldSchema(name='dense_vec', dtype=DataType.FLOAT_VECTOR, dim=1024), # BGE-M3 dense
FieldSchema(name='metadata', dtype=DataType.JSON),
FieldSchema(name='chunk_idx', dtype=DataType.INT64),
FieldSchema(name='content', dtype=DataType.VARCHAR, max_length=65535),
FieldSchema(name='dense_vec', dtype=DataType.FLOAT_VECTOR, dim=1024),
FieldSchema(name='metadata', dtype=DataType.JSON),
]
schema = CollectionSchema(fields, description=description)
col = Collection(name, schema)
# 创建向量索引（HNSW，适合调研阶段）
index_params = {
col.create_index('dense_vec', {
'metric_type': 'COSINE',
'index_type': 'HNSW',
'params': {'M': 16, 'efConstruction': 200}
}
col.create_index('dense_vec', index_params)
})
col.load()
print(f' Collection {name} 创建完成')
create_collection('regulation_chunks', '法规条款向量库')
create_collection('doc_chunks', '企业文档向量库')
create_collection('case_library', '行业案例库')
create_regulation_chunks()
create_simple_collection('doc_chunks', '企业文档向量库')
create_simple_collection('case_library', '行业案例库')
print('Milvus 初始化完成')
" 2>&1