Newsee 实体查找插件

graph TD
    A[用户输入文本] --> B[实体类型选择]
    B --> C{精确匹配?}
    C -->|是| D[返回精确匹配结果]
    C -->|否| E{启用模糊匹配?}
    E -->|是| F[基于编辑距离的模糊匹配]
    F --> G[过滤低分结果]
    G --> H{找到匹配?}
    E -->|否| H
    H -->|是| I[返回模糊匹配结果]
    H -->|否| J[ChromaDB 向量检索]
    J --> K[计算语义相似度]
    K --> L[过滤低分结果]
    L --> M[返回向量检索结果]

graph TD
    A[main.py - 插件入口] --> B[FindNewseeStoreProvider]
    B --> C[EntityFinderMySQL]
    C --> D[DatabaseConnection]
    C --> E[ChromaDB]
    B --> F1[find-project 工具]
    B --> F2[find-org 工具]
    B --> F3[find-target 工具]
    E --> G1[projects 集合]
    E --> G2[orgs 集合]
    E --> G3[targets 集合]

# 安装依赖
python -m venv venv
source venv/bin/activate

pip install -r requirements.txt

CREATE DATABASE IF NOT EXISTS `newsee-view`
    CHARACTER SET utf8mb4
    COLLATE utf8mb4_unicode_ci;

import asyncio
from provider.entity_finder_mysql import EntityFinderMySQL

async def main():
    # 配置数据库连接
    db_config = {
        "host": "localhost",
        "port": 3306,
        "user": "your_username",
        "password": "your_password",
        "database": "newsee-view"
    }

    # 初始化查找器
    finder = EntityFinderMySQL()
    await finder.initialize(db_config)

    # 查找实体
    results = finder.search("星河湾", entity_type="project", top_k=3)
    print(results)

    # 查找组织
    results = finder.search("龙游", entity_type="org", top_k=3)
    print(results)

    # 查找指标
    results = finder.search("旧欠", entity_type="target", top_k=3)
    print(results)

if __name__ == "__main__":
    asyncio.run(main())

DEFAULT_CONFIG = {
    "fuzzy_match_threshold": 0.6,  # 模糊匹配阈值 - 降低以捕获更多结果
    "vector_search_threshold": 0.3,  # 向量搜索阈值 - 降低以捕获更多结果
    "top_k": 3,  # 默认返回结果数量
    "enable_fuzzy": True,  # 是否启用模糊匹配
    "enable_vector_search": True,  # 是否启用向量搜索
    "default_top_k": 3,  # 默认返回结果数量
    "batch_size": 25,  # 批量大小
}

{
    "id": "P001",  # 项目ID
    "name": "星河湾",  # 项目名称
}

{
    "id": "O001",  # 组织ID
    "name": "星河湾1号楼",  # 组织名称
}

{
    "id": "T001",  # 指标ID
    "name": "旧欠",  # 指标名称
    "unit": "元",  # 指标单位
}

# 运行测试脚本
python -m tests.test_plugin

查询示例: "星河湾"

响应:
{
  "success": true,
  "query": "星河湾",
  "entities": [
    {
      "id": "P001",
      "name": "星河湾",
      "type": "project",
      "confidence": 0.95,
      "match_type": "vector"
    }
  ],
  "message": "找到 1 个匹配实体"
}

查询示例: "龙游"

响应:
{
  "success": true,
  "query": "龙游",
  "entities": [
    {
      "id": "O001",
      "name": "龙游公司",
      "type": "org",
      "confidence": 0.92,
      "match_type": "vector"
    }
  ],
  "message": "找到 1 个匹配实体"
}

查询示例: "旧欠"

响应:
{
  "success": true,
  "query": "旧欠",
  "entities": [
    {
      "id": "T001",
      "name": "旧欠",
      "type": "target",
      "confidence": 1.0,
      "match_type": "exact"
    }
  ],
  "message": "找到 1 个匹配实体"
}