File size: 2,567 Bytes
2f6097e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107

"""
ZKCoding Knowledge + Embedding 集成实验

使用原项目的 KnowledgeCatalog 和 SQLite Repository,
然后调用自己编写的 KnowledgeReranker。

运行:
python -m experiments.knowledge_integration_test
"""

import tempfile
from pathlib import Path

from src.adapters.indexes.knowledge_sqlite import (
    SqliteKnowledgeRepository,
)
from src.modules.knowledge.application.catalog import (
    KnowledgeCatalog,
)
from src.modules.skill.domain.models import Actor

from experiments.knowledge_reranker import KnowledgeReranker


def main():
    # 1. 创建测试用户
    actor = Actor(
        tenant_id="test-tenant",
        user_id="developer",
        roles=[],
    )

    # 2. 创建临时数据库
    with tempfile.TemporaryDirectory() as temp_dir:
        folder = Path(temp_dir)

        repo = SqliteKnowledgeRepository(
            folder / "index.sqlite"
        )

        # 3. 复用组长已有的测试数据生成函数
        from src.modules.knowledge.tests.test_knowledge import (
            write_fixture,
        )

        corpus_dir = folder / "corpus"
        catalog_path = write_fixture(corpus_dir)

        repo.import_corpus(
            corpus_dir,
            actor.tenant_id,
            catalog_path,
        )

        # 4. 创建真正的 Knowledge Service
        catalog = KnowledgeCatalog(repo)

        # 5. 调用原有 SQLite FTS 检索
        query = "inventory"

        records = catalog.search(
            actor,
            query,
            ["fixture-java"],
        )

        print("\n【SQLite 检索结果】")
        print("找到文档数量:", len(records))

        for record in records:
            print(
                record["citation_id"],
                record["title"],
                record["source_url"],
            )

        # 6. 调用我们新增的 Embedding 重排序
        reranker = KnowledgeReranker()

        ranked = reranker.rerank(
            query,
            records,
        )

        print("\n【Embedding 重排序结果】")

        for record in ranked:
            print(
                record["citation_id"],
                "相似度:",
                round(record["semantic_score"], 4),
                "来源:",
                record["source_url"],
            )

        # 7. 验证引用信息没有丢失
        assert len(ranked) == len(records)
        assert all(
            r["source_url"] for r in ranked
        )
        print("\n集成测试通过:搜索与重排序已连接!")


if __name__ == "__main__":
    main()