AutoReceipt

Language / 语言: English primary · 中文概览如下。

中文概览

面向员工工作流的原生 AI 工具，将收据识别、标准化和报销准备连接成可复用流程。

A Native AI employee skill for receipt recognition, normalization, and reimbursement preparation.

NATIVE AI · EMPLOYEE WORKFLOW · INTERNAL USE

Employee workflow: receipts and travel documents → structured expense records → standardized files → reimbursement-ready workbook.

---

Why this exists

Native AI should not only help engineers write code. It should remove repetitive knowledge-work steps across the company.

AutoReceipt is a concrete employee workflow:

Receipts / invoices / itineraries
              ↓
document recognition
              ↓
expense classification
              ↓
policy / rule application
              ↓
file normalization
              ↓
reimbursement workbook
              ↓
human review

---

What it does

• recognizes supported travel-expense documents;
• extracts structured fields;
• applies company reimbursement rules;
• renames / organizes attachments;
• generates an import-ready expense workbook;
• keeps a human review step before submission.

---

Provider-neutral design

Do not define the project by one VLM provider.

README:

Any supported vision-language model may be configured.

Move specific provider recommendations, pricing, API-key steps, and currently preferred models to docs/SETUP_GUIDE.md.

Model availability and pricing change faster than the workflow itself.

---

Privacy

This workflow may process personal, financial, travel, or tax-related documents. Deployment owners must review model-provider data handling, retention, logging, and local-storage policies before use.

A PRIVACY.md documenting the exact behavior of any concrete deployment (what files are sent to a model provider, whether local-model execution is supported, what is logged, where temporary files are stored, deletion / retention behavior, secrets / API-key handling, what must never be committed to Git) is the deployment owner's responsibility and must be produced at deployment time. This repository does not publish such a PRIVACY.md because the policy depends on the chosen model provider, the company's retention obligations and the deployment jurisdiction.

This project is classified as Internal Utility. It is not a production-ready product.

---

Evidence

Classify current evidence as Internal Use, not a public benchmark.

Useful future metrics:

• field extraction accuracy;
• manual correction rate;
• processing time saved;
• policy-rule error rate;
• reimbursement rejection rate.

---

TopPrism metadata

topprism:
  purpose: native-ai
  capability: employee-expense-workflow
  platform_layer: organizational-intelligence
  maturity: internal-utility
  evidence:
    type: internal-use