Show HN: 用确定性 .py 引擎取代 5 万美元的手动取证审计
2 分•作者: cd_mkdir•5 个月前
我是一名软件架构师，最近构建了 Exit Protocol (<a href="https://exitprotocols.com" rel="nofollow">https://exitprotocols.com</a>)，这是一个针对高冲突诉讼的自动化法务会计引擎。
问题：
如果你离婚了，需要证明在高度混合的联名银行账户中的 25 万美元是你的“单独财产”（例如，来自婚前的创业公司退出），那么举证责任完全是数学上的。过去，这意味着要支付给法务注册会计师每小时 500 美元，将多年模糊的银行 PDF 文件导入 Excel，并手动追踪每一美元。这需要数周时间，通常花费超过 5 万美元。
我研究了法院用于此的法律标准——最低中间余额规则（LIBR），并意识到这并不是一个会计问题。它是一个分布式系统状态机问题。
为什么我们没有直接“用 AI 解决”？
目前有一百家法律科技初创公司正在尝试使用 LLM 来总结银行数据。在法庭上，生成式 AI 是致命的负债。如果 LLM 幻觉出单个交易，整个账本在 Daubert 标准下将不被采纳。
为了使其适用于法庭，我们必须构建一个严格的确定性流程：
1. 视觉原生摄取（超越 Tesseract）
银行对账单是 OCR 的最终 Boss（合并的单元格、重叠的借方/贷方列）。标准的线性 OCR 会灾难性地失败。我们构建了一个空间网格 OCR 流程（使用 Azure Document Intelligence 和本地 Surya OCR 备用方案），它映射了页面的几何结构。即使是来自多代“地狱 PDF”的文件，它也能完美地重建表格账本。
2. 确定性引擎（LIBR）
LIBR 算法充当单向棘轮。如果账户余额降至低于你的单独财产索赔金额，你的索赔将永久限制在该新下限。随后的婚姻存款不会补充它（“补充谬论”）。引擎按时间顺序重放数千笔交易，持续评估 S\_t = min(S\_t-1, B\_t)。
3. 解决时间戳歧义
银行 PDF 文件提供日期，而不是时间戳。如果 1 万美元的存款和 1 万美元的提款发生在同一天，顺序很重要。我们构建了一个模拟切换，强制“最坏情况”（先处理提款）与“最佳情况”排序，为和解谈判建立了数学上无可辩驳的“真相区域”。
4. 密码学监管链和主权模式
律师们害怕云 SaaS 泄露。我们通过 Docker 将整个单体应用（Django 5.0/Postgres/Celery）容器化，以便企业公司可以在自己的硬件上以气隙方式运行它（主权模式）。此外，每个生成的 PDF 档案都使用底层数据快照的 SHA-256 哈希值进行密封，向法官证明自生成以来输出未被篡改。
如果你想看看实际的数学运算，我们设置了一个“演示沙盒”，其中填充了一个合成的、高度复杂的 3 年混合账本。你可以在这里自己运行引擎（推荐桌面端）：<a href="https://exitprotocols.com/simulation/uplink/" rel="nofollow">https://exitprotocols.com/simulation/uplink/</a>
这是它从原始 PDF 或法务审计档案中生成的确切的“律师工作成果”——<a href="https://exitprotocols.com/static/documents/Forensic_Audit_Sample_Vinay_MKT2026.pdf" rel="nofollow">https://exitprotocols.com/static/documents/Forensic_Audit_Sa...</a>
我希望从 HN 社区获得关于架构的反馈——特别是处理边缘情况数据摄取和在 B2B 企业部署中维护密码学完整性。
谢谢！
查看原文
I’m a software architect, and I recently built Exit Protocol (<a href="https://exitprotocols.com" rel="nofollow">https://exitprotocols.com</a>), an automated forensic accounting engine for high-conflict litigation.<p>Problem:
If you get divorced and need to prove that a specific $250k in a heavily commingled joint bank account is your "separate property" (e.g., from a pre-marital startup exit), the burden of proof is strictly mathematical. Historically, this meant paying a forensic CPA $500/hour to dump years of blurry bank PDFs into Excel and manually trace every dollar. It takes weeks and routinely costs over $50,000.<p>I looked at the legal standard courts use for this—the Lowest Intermediate Balance Rule (LIBR)—and realized it wasn’t an accounting problem. It is a Distributed Systems state-machine problem.<p>Why we didn't just "Throw AI at it"?<p>There are a hundred legal-tech startups right now trying to use LLMs to summarize bank data. In a courtroom, GenAI is a fatal liability. If an LLM hallucinates a single transaction, the entire ledger is inadmissible under the Daubert standard.<p>To make this court-ready, we had to build a strictly deterministic pipeline:<p>1. Vision-Native Ingestion (Beating Tesseract)
Bank statements are the final boss of OCR (merged cells, overlapping debit/credit columns). Standard linear OCR fails catastrophically. We built a spatial-grid OCR pipeline (using Azure Document Intelligence with a local Surya OCR fallback) that maps the geometric structure of the page. It reconstructs tabular ledgers perfectly, even from multi-generational "PDFs from hell."<p>2. The Deterministic Engine (LIBR)
The LIBR algorithm acts as a one-way ratchet. If an account balance drops below your separate property claim amount, your claim is permanently capped at that new floor. Subsequent marital deposits do not refill it (the "replenishment fallacy"). The engine replays thousands of transactions chronologically, continuously evaluating S_t = min(S_t-1, B_t).<p>3. Resolving Timestamp Ambiguity
Bank PDFs give you dates, not timestamps. If a $10k deposit and $10k withdrawal happen on the same day, order matters. We built a simulation toggle that forces "Worst Case" (withdrawals process first) vs "Best Case" sorting, establishing a mathematically irrefutable "Zone of Truth" for settlement negotiations.<p>4. Cryptographic Chain of Custody & Sovereign Mode
Lawyers are terrified of cloud SaaS breaches. We containerized the entire monolith (Django 5.0/Postgres/Celery) via Docker so enterprise firms can run it air-gapped on their own hardware (Sovereign Mode). Furthermore, every generated PDF dossier is sealed with a SHA-256 hash of the underlying data snapshot, proving to a judge that the output hasn't been tampered with since generation.<p>If you want to see the math in action, we set up a "Demo Sandbox" populated with a synthetic, highly complex 3-year commingled ledger. You can run the engine yourself here (Desktop recommended): <a href="https://exitprotocols.com/simulation/uplink/" rel="nofollow">https://exitprotocols.com/simulation/uplink/</a><p>Here is the exact "Attorney Work Product" it generates from raw PDF or Forensic Audit Dossier our system generates- <a href="https://exitprotocols.com/static/documents/Forensic_Audit_Sample_Vinay_MKT2026.pdf" rel="nofollow">https://exitprotocols.com/static/documents/Forensic_Audit_Sa...</a><p>I'd love feedback from the HN crowd on the architecture—specifically handling edge-case data ingestion and maintaining cryptographic integrity in B2B enterprise deployments.<p>Cheers!