政务服务 AI Agent 办件材料智能预审与要素交叉核验架构实战：破解材料缺漏、跨证照信息冲突与容缺受理判定难题

production-government-material-pre-audit-engine.ts（政务办件材料预审与要素核验引擎）

/**
 * 百信政务 AI Agent 办件材料预审与要素交叉核验引擎
 * 生产级实现：涵盖材料拓扑校验、跨证照实体对齐、容缺受理守卫与补正通知生成
 */

// 1. 办件材料类型与法定容缺属性定义
export enum MaterialCategory {
  CORE_MANDATORY = 'CORE_MANDATORY',       // 核心法定材料（不可容缺）
  IDENTITY_PROOF = 'IDENTITY_PROOF',       // 主体资格证明（不可容缺）
  CONDITIONAL_PREREQ = 'CONDITIONAL_PREREQ', // 条件性前置材料
  AUXILIARY_TOLERABLE = 'AUXILIARY_TOLERABLE' // 辅助证明（可容缺）
}

// 2. 事项库材料定义接口
export interface MatterMaterialDef {
  code: string;                          // 材料编码（五级四同标准编码）
  name: string;                          // 材料名称
  category: MaterialCategory;            // 材料分类
  isTolerable: boolean;                  // 是否支持容缺后补
  tolerableMaxDays?: number;             // 允许容缺补正的最长工作日
  requiredFields: string[];              // 必须抽取的要素字段清单
  dependencies?: string[];               // 前置依赖材料编码
}

// 3. 用户实际提交的材料抽取产物
export interface ExtractedMaterialData {
  materialCode: string;                  // 对应的材料编码
  fileName: string;                      // 文件名
  confidence: number;                    // 整体抽取置信度 (0.0 ~ 1.0)
  hasValidStamp: boolean;                // 是否加盖有效印章
  extractedFields: Record<string, {
    value: string;                       // 抽取值
    confidence: number;                  // 字段置信度
    bbox?: [number, number, number, number]; // 坐标包围盒
  }>;
}

// 4. 预审结果状态枚举
export enum PreAuditStatus {
  ACCEPTED = 'ACCEPTED',                         // 材料齐全，直接受理
  CONDITIONAL_ACCEPTED = 'CONDITIONAL_ACCEPTED', // 允许容缺受理（先办后补）
  SUPPLEMENT_REQUIRED = 'SUPPLEMENT_REQUIRED',   // 需补正材料（阻断受理）
  REJECTED = 'REJECTED'                          // 严重不合规/阻断驳回
}

// 5. 补正与瑕疵项明细接口
export interface AuditDefect {
  type: 'MISSING_MATERIAL' | 'FIELD_CONFLICT' | 'EXPIRED' | 'MISSING_STAMP' | 'LOW_CONFIDENCE';
  materialCode: string;
  materialName: string;
  field?: string;
  severity: 'FATAL' | 'TOLERABLE' | 'WARNING';
  description: string;
  suggestion: string;
}

// 6. 统一社会信用代码校验函数 (GB 32100-2015 标准算法)
export function validateUSCC(uscc: string): boolean {
  if (!uscc || uscc.length !== 18) return false;
  const chars = '0123456789ABCDEFGHJKLMNPQRTUWXY';
  const weights = [1, 3, 9, 27, 19, 26, 16, 17, 20, 29, 25, 13, 8, 24, 10, 30, 28];

  let sum = 0;
  for (let i = 0; i < 17; i++) {
    const charIndex = chars.indexOf(uscc[i]);
    if (charIndex === -1) return false;
    sum += charIndex * weights[i];
  }
  const checkRemainder = (31 - (sum % 31)) % 31;
  const expectedCheckChar = chars[checkRemainder];
  return uscc[17] === expectedCheckChar;
}

// 7. 地址模糊容差比对器（层次化相似度匹配）
export function compareAddressWithTolerance(addr1: string, addr2: string): { match: boolean; similarity: number } {
  const normalize = (s: string) => s.replace(/[\s\-_#（）\(\)]/g, '').replace(/中国|开发区|软件谷/g, '');
  const clean1 = normalize(addr1);
  const clean2 = normalize(addr2);

  if (clean1 === clean2) return { match: true, similarity: 1.0 };

  // 计算公共最长子串比例与编辑距离综合得分
  const maxLength = Math.max(clean1.length, clean2.length);
  if (maxLength === 0) return { match: true, similarity: 1.0 };

  // 简易编辑距离算法
  const matrix: number[][] = [];
  for (let i = 0; i <= clean1.length; i++) matrix[i] = [i];
  for (let j = 0; j <= clean2.length; j++) matrix[0][j] = j;

  for (let i = 1; i <= clean1.length; i++) {
    for (let j = 1; j <= clean2.length; j++) {
      const cost = clean1[i - 1] === clean2[j - 1] ? 0 : 1;
      matrix[i][j] = Math.min(matrix[i - 1][j] + 1, matrix[i][j - 1] + 1, matrix[i - 1][j - 1] + cost);
    }
  }

  const distance = matrix[clean1.length][clean2.length];
  const similarity = (maxLength - distance) / maxLength;

  // 设定阈值：政务地址规范度通常在 0.88 以上视为同义微调
  return { match: similarity >= 0.88, similarity };
}

// 8. 核心智能预审流水线调度器
export class GovernmentMaterialPreAuditEngine {
  private matterDef: MatterMaterialDef[];
  private referenceDate: Date;

  constructor(matterDef: MatterMaterialDef[], referenceDate: Date = new Date()) {
    this.matterDef = matterDef;
    this.referenceDate = referenceDate;
  }

  public auditSubmission(submissions: ExtractedMaterialData[]): {
    status: PreAuditStatus;
    defects: AuditDefect[];
    noticeMarkdown?: string;
  } {
    const defects: AuditDefect[] = [];
    const submissionMap = new Map<string, ExtractedMaterialData>();
    for (const sub of submissions) {
      submissionMap.set(sub.materialCode, sub);
    }

    // 第一阶段：材料齐全度与 DAG 拓扑依赖检查
    for (const matDef of this.matterDef) {
      const submitted = submissionMap.get(matDef.code);
      if (!submitted) {
        defects.push({
          type: 'MISSING_MATERIAL',
          materialCode: matDef.code,
          materialName: matDef.name,
          severity: matDef.isTolerable ? 'TOLERABLE' : 'FATAL',
          description: `法定必需材料【${matDef.name}】未上传。`,
          suggestion: `请提供盖章或签字的【${matDef.name}】扫描件或电子证照。`
        });
        continue;
      }

      // 检查盖章完整性（核心材料必须有印章）
      if (matDef.category === MaterialCategory.CORE_MANDATORY && !submitted.hasValidStamp) {
        defects.push({
          type: 'MISSING_STAMP',
          materialCode: matDef.code,
          materialName: matDef.name,
          severity: 'FATAL',
          description: `材料【${matDef.name}】未检测到有效行政公章或电子印章。`,
          suggestion: '请加盖企业公章或有权法定代表人名章后重新扫描上传。'
        });
      }

      // 检查关键字段抽取置信度
      for (const field of matDef.requiredFields) {
        const fieldData = submitted.extractedFields[field];
        if (!fieldData || fieldData.confidence < 0.80) {
          defects.push({
            type: 'LOW_CONFIDENCE',
            materialCode: matDef.code,
            materialName: matDef.name,
            field,
            severity: 'WARNING',
            description: `材料【${matDef.name}】中要素【${field}】清晰度不足或识别置信度偏低（${(fieldData?.confidence ?? 0).toFixed(2)}）。`,
            suggestion: '请检查原图是否存在反光、遮挡或折角，建议重新上传高清原件翻拍件。'
          });
        }
      }
    }

    // 第二阶段：跨材料核心实体交叉核验（USCC、法人、住所）
    this.verifyCrossMaterialEntities(submissions, defects);

    // 第三阶段：有限状态机判定决策
    const hasFatal = defects.some(d => d.severity === 'FATAL');
    const hasTolerable = defects.some(d => d.severity === 'TOLERABLE');

    let status: PreAuditStatus;
    if (hasFatal) {
      status = PreAuditStatus.SUPPLEMENT_REQUIRED;
    } else if (hasTolerable) {
      status = PreAuditStatus.CONDITIONAL_ACCEPTED;
    } else {
      status = PreAuditStatus.ACCEPTED;
    }

    // 第四阶段：自动生成标准补正或告知通知书
    const noticeMarkdown = this.generateAuditNotice(status, defects);

    return { status, defects, noticeMarkdown };
  }

  // 跨材料实体交叉比对逻辑
  private verifyCrossMaterialEntities(submissions: ExtractedMaterialData[], defects: AuditDefect[]) {
    let baseUSCC: string | undefined;
    let baseLegalPerson: string | undefined;
    let baseAddress: string | undefined;

    // 优先从营业执照（假设编码为 MAT_BIZ_LICENSE）或申请书确立实体基线
    const license = submissions.find(s => s.materialCode === 'MAT_BIZ_LICENSE' || s.materialCode.includes('LICENSE'));
    if (license) {
      baseUSCC = license.extractedFields['unifiedSocialCreditCode']?.value;
      baseLegalPerson = license.extractedFields['legalRepresentative']?.value;
      baseAddress = license.extractedFields['enterpriseAddress']?.value;

      // 验证执照 USCC 本身有效性
      if (baseUSCC && !validateUSCC(baseUSCC)) {
        defects.push({
          type: 'FIELD_CONFLICT',
          materialCode: license.materialCode,
          materialName: '营业执照',
          field: 'unifiedSocialCreditCode',
          severity: 'FATAL',
          description: `统一社会信用代码【${baseUSCC}】校验码不符合国家 GB 32100-2015 规范。`,
          suggestion: '请核实是否录入错误，或联系市场监管部门换发最新证照。'
        });
      }
    }

    // 遍历其他材料，核对与基线要素的一致性
    for (const sub of submissions) {
      const uscc = sub.extractedFields['unifiedSocialCreditCode']?.value;
      if (baseUSCC && uscc && uscc !== baseUSCC) {
        defects.push({
          type: 'FIELD_CONFLICT',
          materialCode: sub.materialCode,
          materialName: sub.materialCode,
          field: 'unifiedSocialCreditCode',
          severity: 'FATAL',
          description: `统一社会信用代码在【${sub.materialCode}】（${uscc}）与营业执照基线（${baseUSCC}）不一致。`,
          suggestion: '请确保申报主体代码一致，不得使用关联主体或其他关联方凭据。'
        });
      }

      const legalPerson = sub.extractedFields['legalRepresentative']?.value;
      if (baseLegalPerson && legalPerson && legalPerson !== baseLegalPerson) {
        defects.push({
          type: 'FIELD_CONFLICT',
          materialCode: sub.materialCode,
          materialName: sub.materialCode,
          field: 'legalRepresentative',
          severity: 'FATAL',
          description: `法定代表人姓名在【${sub.materialCode}】（${legalPerson}）与基线（${baseLegalPerson}）不符。`,
          suggestion: '若近期发生法定代表人变更，请附交市场监管核准变更通知书及授权证明。'
        });
      }

      const address = sub.extractedFields['enterpriseAddress']?.value;
      if (baseAddress && address) {
        const { match, similarity } = compareAddressWithTolerance(baseAddress, address);
        if (!match) {
          defects.push({
            type: 'FIELD_CONFLICT',
            materialCode: sub.materialCode,
            materialName: sub.materialCode,
            field: 'enterpriseAddress',
            severity: 'FATAL',
            description: `企业住所地址在【${sub.materialCode}】（${address}）与基线（${baseAddress}）差异过大（匹配度 ${(similarity * 100).toFixed(1)}%）。`,
            suggestion: '请核实是否属于跨区域迁移，或统一修正申报表单住所字段。'
          });
        }
      }
    }
  }

  // 格式化输出政务标准告知通知单
  private generateAuditNotice(status: PreAuditStatus, defects: AuditDefect[]): string {
    const dateStr = this.referenceDate.toISOString().slice(0, 10);
    if (status === PreAuditStatus.ACCEPTED) {
      return `### 政务服务智能预审回执单（预审通过）\n**预审日期**：${dateStr}\n**审查结论**：申报材料齐全，要素完全吻合，符合法定形式，已自动推入正式行政审批内网系统。`;
    }

    if (status === PreAuditStatus.CONDITIONAL_ACCEPTED) {
      const items = defects.filter(d => d.severity === 'TOLERABLE').map((d, i) => `${i + 1}. **【${d.materialName}】**：${d.description}（建议：${d.suggestion}）`).join('\n');
      return `### 政务服务“容缺受理”告知承诺书\n**办理批次**：PR-${Date.now()}\n**预审日期**：${dateStr}\n**审查结论**：核心法定条件已满足，根据“容缺受理”规范，以下辅助材料允许后补：\n${items}\n\n**承诺承诺期限**：请于本受理单出具之日起 5 个工作日内补齐上述材料，逾期未补齐将依法中止审批程序。`;
    }

    const fatalItems = defects.filter(d => d.severity === 'FATAL').map((d, i) => `${i + 1}. **【${d.materialName}】${d.field ? '要素：' + d.field : ''}**\n   - 问题：${d.description}\n   - 补正要求：${d.suggestion}`).join('\n\n');
    return `### 行政审批一次性补正告知书\n**告知书编号**：BZ-${Date.now()}\n**下发日期**：${dateStr}\n**审查结论**：经 AI 智能预审与要素校验，您的申报材料存在以下必须补正的法定事项，暂不予正式受理：\n\n${fatalItems}\n\n**经办提示**：请根据上述要求逐项修正后重新提交预审。`;
  }
}