You've already forked DataMate
- 实现三层对齐策略:规则层 + 向量相似度层 + LLM 仲裁层 - 规则层:名称规范化(NFKC、小写、去标点/空格)+ 规则评分 - 向量层:OpenAI Embeddings + cosine 相似度计算 - LLM 层:仅对边界样本调用,严格 JSON schema 校验 - 使用 Union-Find 实现传递合并 - 支持批内对齐(库内对齐待 KG 服务 API 支持) 核心组件: - EntityAligner 类:align() (async)、align_rules_only() (sync) - 配置项:kg_alignment_enabled(默认 false)、embedding_model、阈值 - 失败策略:fail-open(对齐失败不中断请求) 集成: - 已集成到抽取主链路(extract → align → return) - extract() 调用 async align() - extract_sync() 调用 sync align_rules_only() 修复: - P1-1:使用 (name, type) 作为 key,避免同名跨类型误合并 - P1-2:LLM 计数在 finally 块中增加,异常也计数 - P1-3:添加库内对齐说明(待后续实现) 新增 41 个测试用例,全部通过 测试结果:41 tests pass
128 lines
4.7 KiB
Python
128 lines
4.7 KiB
Python
from pydantic_settings import BaseSettings
|
|
from pydantic import SecretStr, model_validator
|
|
from typing import Optional
|
|
import logging
|
|
import os
|
|
|
|
_logger = logging.getLogger(__name__)
|
|
|
|
# 已知的弱默认凭据,生产环境禁止使用
|
|
_BLOCKED_DEFAULT_PASSWORDS = {"password", "123456", "admin", "root", "datamate123"}
|
|
_BLOCKED_DEFAULT_TOKENS = {"abc123abc123", "EMPTY"}
|
|
|
|
|
|
class Settings(BaseSettings):
|
|
"""应用程序配置"""
|
|
|
|
class Config:
|
|
env_file = ".env"
|
|
case_sensitive = False
|
|
extra = 'ignore'
|
|
|
|
# Service
|
|
app_name: str = "DataMate Python Backend"
|
|
app_version: str = "1.0.0"
|
|
app_description: str = "Adapter for integrating Data Management System with Label Studio"
|
|
|
|
host: str = "0.0.0.0"
|
|
port: int = 18000
|
|
|
|
# CORS
|
|
# allowed_origins: List[str] = ["*"]
|
|
# allowed_methods: List[str] = ["*"]
|
|
# allowed_headers: List[str] = ["*"]
|
|
|
|
# Log
|
|
log_level: str = "INFO"
|
|
debug: bool = True
|
|
log_file_dir: str = "/var/log/datamate/backend-python"
|
|
|
|
# Database
|
|
mysql_host: str = "datamate-database"
|
|
mysql_port: int = 3306
|
|
mysql_user: str = "root"
|
|
mysql_password: str = "password"
|
|
mysql_database: str = "datamate"
|
|
|
|
database_url: str = "" # Will be overridden by build_database_url() if not provided
|
|
|
|
@model_validator(mode='after')
|
|
def build_database_url(self):
|
|
"""如果没有提供 database_url,则根据 MySQL 配置构建"""
|
|
if not self.database_url:
|
|
if self.mysql_password and self.mysql_user:
|
|
self.database_url = f"mysql+aiomysql://{self.mysql_user}:{self.mysql_password}@{self.mysql_host}:{self.mysql_port}/{self.mysql_database}"
|
|
else:
|
|
self.database_url = f"mysql+aiomysql://{self.mysql_host}:{self.mysql_port}/{self.mysql_database}"
|
|
return self
|
|
|
|
|
|
# Label Studio
|
|
label_studio_base_url: str = "http://label-studio:8000"
|
|
label_studio_username: Optional[str] = "admin@demo.com"
|
|
label_studio_password: Optional[str] = "demoadmin"
|
|
label_studio_user_token: Optional[str] = "abc123abc123" # Legacy Token
|
|
|
|
label_studio_local_document_root: str = "/label-studio/local" # Label Studio local file storage path
|
|
label_studio_file_path_prefix: str = "/data/local-files/?d=" # Label Studio local file serving URL prefix
|
|
|
|
ls_task_page_size: int = 1000
|
|
|
|
# DataMate
|
|
dm_file_path_prefix: str = "/dataset" # DM存储文件夹前缀
|
|
|
|
# DataMate Backend (Java) - 用于通过"下载/预览接口"读取文件内容
|
|
datamate_backend_base_url: str = "http://datamate-backend:8080/api"
|
|
|
|
# Knowledge Graph - LLM 三元组抽取配置
|
|
kg_llm_api_key: SecretStr = SecretStr("EMPTY")
|
|
kg_llm_base_url: Optional[str] = None
|
|
kg_llm_model: str = "gpt-4o-mini"
|
|
kg_llm_temperature: float = 0.0
|
|
kg_llm_timeout_seconds: int = 60
|
|
kg_llm_max_retries: int = 2
|
|
|
|
# Knowledge Graph - 实体对齐配置
|
|
kg_alignment_enabled: bool = False
|
|
kg_alignment_embedding_model: str = "text-embedding-3-small"
|
|
kg_alignment_vector_threshold: float = 0.92
|
|
kg_alignment_llm_threshold: float = 0.78
|
|
|
|
# 标注编辑器(Label Studio Editor)相关
|
|
editor_max_text_bytes: int = 0 # <=0 表示不限制,正数为最大字节数
|
|
|
|
@model_validator(mode='after')
|
|
def check_default_credentials(self):
|
|
"""生产环境下检测弱默认凭据,拒绝启动。
|
|
|
|
通过环境变量 DATAMATE_ENV 判断环境:
|
|
- dev/test/local: 仅发出警告
|
|
- 其他(prod/staging 等): 抛出异常阻止启动
|
|
"""
|
|
env = os.environ.get("DATAMATE_ENV", "dev").lower()
|
|
is_dev = env in ("dev", "test", "local", "development")
|
|
issues: list[str] = []
|
|
|
|
if self.mysql_password in _BLOCKED_DEFAULT_PASSWORDS:
|
|
issues.append(f"mysql_password is set to a weak default ('{self.mysql_password}')")
|
|
|
|
if self.label_studio_password and self.label_studio_password in _BLOCKED_DEFAULT_PASSWORDS:
|
|
issues.append("label_studio_password is set to a weak default")
|
|
|
|
if self.label_studio_user_token and self.label_studio_user_token in _BLOCKED_DEFAULT_TOKENS:
|
|
issues.append("label_studio_user_token is set to a weak default")
|
|
|
|
if issues:
|
|
msg = "SECURITY: Weak default credentials detected: " + "; ".join(issues)
|
|
if is_dev:
|
|
_logger.warning(msg + " (acceptable in dev/test, MUST change for production)")
|
|
else:
|
|
raise ValueError(
|
|
msg + ". Set proper credentials via environment variables "
|
|
"before deploying to production."
|
|
)
|
|
return self
|
|
|
|
# 全局设置实例
|
|
settings = Settings()
|