Claude API のドキュメント抽出が「確信を持って間違える」とき — 検証層の設計メモ

// src/arithmetic.ts
import type { Invoice } from "./schema";
 
export type Discrepancy = { field: string; expected: number; got: number; diff: number };
 
// 通貨ごとの許容誤差（最小単位の丸め用）。JPY は端数なし、USD は 1 セント。
const TOLERANCE: Record<string, number> = { JPY: 0, USD: 0.01, EUR: 0.01 };
 
export function auditInvoiceMath(inv: Invoice): Discrepancy[] {
  const tol = TOLERANCE[inv.currency] ?? 0.01;
  const issues: Discrepancy[] = [];
 
  const lineSum = inv.lineItems.reduce((s, li) => s + li.amount, 0);
 
  // 明細合計と小計の整合
  if (inv.subtotal !== undefined && Math.abs(lineSum - inv.subtotal) > tol) {
    issues.push({ field: "subtotal", expected: lineSum, got: inv.subtotal, diff: inv.subtotal - lineSum });
  }
 
  // 小計＋税＝合計の整合（小計が無ければ明細合計で代用）
  const base = inv.subtotal ?? lineSum;
  const expectedTotal = base + (inv.tax ?? 0);
  if (Math.abs(expectedTotal - inv.total) > tol) {
    issues.push({ field: "total", expected: expectedTotal, got: inv.total, diff: inv.total - expectedTotal });
  }
 
  // 各明細の 数量×単価＝金額（両方そろっている行のみ）
  for (const [i, li] of inv.lineItems.entries()) {
    if (li.quantity !== undefined && li.unitPrice !== undefined) {
      const expected = li.quantity * li.unitPrice;
      if (Math.abs(expected - li.amount) > tol) {
        issues.push({ field: `lineItems[${i}].amount`, expected, got: li.amount, diff: li.amount - expected });
      }
    }
  }
  return issues;
}

// src/agreement.ts
import Anthropic from "@anthropic-ai/sdk";
import { InvoiceSchema, type Invoice } from "./schema";
 
const client = new Anthropic();
 
// 正規化してから比較する（全角空白・前後空白・大文字小文字を吸収）
const norm = (v: unknown) =>
  typeof v === "string" ? v.normalize("NFKC").trim().toLowerCase() : v;
 
export function fieldAgreement(a: Invoice, b: Invoice): string[] {
  const mismatches: string[] = [];
  const keys: (keyof Invoice)[] = ["invoiceNumber", "issueDate", "dueDate", "total", "currency"];
  for (const k of keys) {
    if (norm(a[k]) !== norm(b[k])) mismatches.push(String(k));
  }
  if (norm(a.vendor.name) !== norm(b.vendor.name)) mismatches.push("vendor.name");
  return mismatches;
}
 
// 視点を変えた二次抽出。一次と独立性を持たせるため指示の枠組みを変える
export async function secondPass(content: string): Promise<Invoice> {
  const res = await client.messages.create({
    model: "claude-sonnet-4-6",
    max_tokens: 2048,
    system:
      "あなたは監査担当です。請求書を一字一句確認し、記載されている値だけを JSON で写し取ってください。" +
      "推測・補完・四捨五入は禁止。読み取れない項目は null。有効な JSON のみを返すこと。",
    messages: [{ role: "user", content: `次の請求書を転記してください:\n\n${content}` }],
  });
  const text = res.content[0].type === "text" ? res.content[0].text : "";
  const json = text.match(/\{[\s\S]*\}/)?.[0] ?? text;
  return InvoiceSchema.parse(JSON.parse(json));
}

// src/verify.ts
import { InvoiceSchema, type Invoice } from "./schema";
import { auditInvoiceMath } from "./arithmetic";
import { fieldAgreement, secondPass } from "./agreement";
 
export type Verdict =
  | { status: "accept"; data: Invoice }
  | { status: "review"; data: Invoice; reasons: string[] }
  | { status: "reject"; reasons: string[] };
 
export async function verifyInvoice(raw: unknown, sourceText: string): Promise<Verdict> {
  // 1. スキーマ：形が壊れていれば即 reject
  const parsed = InvoiceSchema.safeParse(raw);
  if (!parsed.success) {
    return { status: "reject", reasons: parsed.error.issues.map((i) => `${i.path.join(".")}: ${i.message}`) };
  }
  const inv = parsed.data;
  const reasons: string[] = [];
 
  // 2. 算術：内部矛盾は review（自動 accept させない）
  const mathIssues = auditInvoiceMath(inv);
  for (const d of mathIssues) {
    reasons.push(`算術不整合 ${d.field}: 期待 ${d.expected} / 抽出 ${d.got}（差 ${d.diff}）`);
  }
 
  // 3. 二重抽出：重要フィールドの不一致は review
  try {
    const second = await secondPass(sourceText);
    const diff = fieldAgreement(inv, second);
    if (diff.length > 0) reasons.push(`二次抽出と不一致: ${diff.join(", ")}`);
  } catch {
    reasons.push("二次抽出に失敗（独立検証なしのため要確認）");
  }
 
  if (reasons.length === 0) return { status: "accept", data: inv };
  return { status: "review", data: inv, reasons };
}