refactor(server): split oversized backend services

server/src/app/services/document_intelligence.py

@@ -2,178 +2,20 @@ from __future__ import annotations
 
 import json
 import re
-from dataclasses import dataclass
 from decimal import Decimal, InvalidOperation
 from typing import Any
 
-from pydantic import BaseModel, Field, ValidationError
+from pydantic import ValidationError
 from sqlalchemy.orm import Session
 
-
-@dataclass(frozen=True, slots=True)
-class DocumentField:
-    key: str
-    label: str
-    value: str
-
-
-@dataclass(frozen=True, slots=True)
-class DocumentInsight:
-    document_type: str
-    document_type_label: str
-    scene_code: str
-    scene_label: str
-    expense_type: str
-    fields: tuple[DocumentField, ...] = ()
-    classification_source: str = "rule"
-    classification_confidence: float = 0.0
-    evidence: tuple[str, ...] = ()
-    warnings: tuple[str, ...] = ()
-
-
-@dataclass(frozen=True, slots=True)
-class DocumentRule:
-    document_type: str
-    document_type_label: str
-    scene_code: str
-    scene_label: str
-    expense_type: str
-    keywords: tuple[str, ...]
-    score_bias: float = 0.0
-
-
-@dataclass(frozen=True, slots=True)
-class RuleMatch:
-    rule: DocumentRule | None
-    confidence: float
-    evidence: tuple[str, ...]
-    score: float
-
-
-class LlmDocumentClassification(BaseModel):
-    document_type: str = Field(default="other")
-    scene_code: str = Field(default="other")
-    scene_label: str = Field(default="其他票据")
-    expense_type: str = Field(default="other")
-    confidence: float = Field(default=0.0, ge=0.0, le=1.0)
-    evidence: list[str] = Field(default_factory=list)
-    fields: list[DocumentField] = Field(default_factory=list)
-
-
-DEFAULT_RULE = DocumentRule(
-    document_type="other",
-    document_type_label="其他单据",
-    scene_code="other",
-    scene_label="其他票据",
-    expense_type="other",
-    keywords=(),
-    score_bias=0.0,
+from app.services.document_intelligence_rules import DEFAULT_RULE, DOCUMENT_RULES, DOCUMENT_TYPE_RULE_MAP, SUPPORTED_DOCUMENT_TYPES
+from app.services.document_intelligence_types import (
+    DocumentField,
+    DocumentInsight,
+    LlmDocumentClassification,
+    RuleMatch,
 )
 
-DOCUMENT_RULES: tuple[DocumentRule, ...] = (
-    DocumentRule(
-        document_type="flight_itinerary",
-        document_type_label="机票/航班行程单",
-        scene_code="travel",
-        scene_label="差旅票据",
-        expense_type="travel",
-        keywords=("电子行程单", "航班号", "航班", "机票", "登机", "航空", "客票"),
-        score_bias=0.34,
-    ),
-    DocumentRule(
-        document_type="train_ticket",
-        document_type_label="火车/高铁票",
-        scene_code="travel",
-        scene_label="差旅票据",
-        expense_type="travel",
-        keywords=("铁路电子客票", "电子客票", "高铁", "火车", "动车", "铁路", "车次", "检票", "二等座", "一等座", "票价"),
-        score_bias=0.32,
-    ),
-    DocumentRule(
-        document_type="hotel_invoice",
-        document_type_label="酒店住宿票据",
-        scene_code="hotel",
-        scene_label="住宿票据",
-        expense_type="hotel",
-        keywords=("住宿", "房费", "客房", "入住", "离店", "酒店", "宾馆", "间夜"),
-        score_bias=0.16,
-    ),
-    DocumentRule(
-        document_type="taxi_receipt",
-        document_type_label="出租车/网约车票据",
-        scene_code="transport",
-        scene_label="交通票据",
-        expense_type="transport",
-        keywords=("滴滴出行", "滴滴", "网约车", "出租车", "打车", "乘车", "用车", "叫车", "车费", "车资", "的士", "快车", "专车", "订单号", "上车", "下车", "起点", "终点", "里程", "司机"),
-        score_bias=0.38,
-    ),
-    DocumentRule(
-        document_type="parking_toll_receipt",
-        document_type_label="停车/通行费票据",
-        scene_code="transport",
-        scene_label="交通票据",
-        expense_type="transport",
-        keywords=("停车费", "通行费", "过路费", "收费站", "停车场", "停车"),
-        score_bias=0.28,
-    ),
-    DocumentRule(
-        document_type="meal_receipt",
-        document_type_label="餐饮票据",
-        scene_code="meal",
-        scene_label="餐饮票据",
-        expense_type="meal",
-        keywords=("餐饮", "餐费", "用餐", "饭店", "酒楼", "餐厅", "食品", "外卖", "咖啡"),
-        score_bias=0.14,
-    ),
-    DocumentRule(
-        document_type="office_invoice",
-        document_type_label="办公用品票据",
-        scene_code="office",
-        scene_label="办公用品票据",
-        expense_type="office",
-        keywords=("办公用品", "文具", "耗材", "打印纸", "墨盒", "硒鼓", "键盘", "鼠标"),
-        score_bias=0.14,
-    ),
-    DocumentRule(
-        document_type="meeting_invoice",
-        document_type_label="会议/会务票据",
-        scene_code="meeting",
-        scene_label="会务票据",
-        expense_type="meeting",
-        keywords=("会议", "会务", "会展", "论坛", "会议室", "会场"),
-        score_bias=0.12,
-    ),
-    DocumentRule(
-        document_type="training_invoice",
-        document_type_label="培训票据",
-        scene_code="training",
-        scene_label="培训票据",
-        expense_type="training",
-        keywords=("培训", "课程", "讲师", "教材", "学费", "认证"),
-        score_bias=0.12,
-    ),
-    DocumentRule(
-        document_type="vat_invoice",
-        document_type_label="增值税发票",
-        scene_code="other",
-        scene_label="通用发票",
-        expense_type="other",
-        keywords=("发票代码", "发票号码", "价税合计", "增值税", "电子发票"),
-        score_bias=-0.08,
-    ),
-    DocumentRule(
-        document_type="receipt",
-        document_type_label="一般收据/凭证",
-        scene_code="other",
-        scene_label="其他票据",
-        expense_type="other",
-        keywords=("收据", "凭证", "票据"),
-        score_bias=-0.18,
-    ),
-)
-
-DOCUMENT_TYPE_RULE_MAP = {rule.document_type: rule for rule in DOCUMENT_RULES}
-SUPPORTED_DOCUMENT_TYPES = tuple(DOCUMENT_TYPE_RULE_MAP.keys()) + ("other",)
 
 AMOUNT_PATTERNS = (
     re.compile(