ClientVault：離線移除合約及發票中的敏感客戶資料，生成案例與税務帳簿
ClientVault
ClientVault 是一款離線優先工具，可從合約及發票移除敏感客戶資料，並生成公開案例與税務帳簿。它以 Python、FastAPI/Gradio 開發，透過 Ollama 在本機執行開放權重 Gemma 2 (2B) 推理，資料不會傳出裝置。程式碼採 MIT 授權，每次執行毋須費用。
Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
What I Built
ClientVault is an offline-first tool that strips sensitive client data from contracts and invoices to generate public case studies and tax ledgers. It helps freelancers under strict NDAs safely market their work and organize finances without leaking proprietary client info to the cloud.
Demo
- Live App: https://clientvault.onrender.com
- Local run:
git clone https://github.com/mananmeemroth/ClientVault.git && pip install -r requirements.txt && python app.py
Code
Source code licensed under MIT: https://github.com/mananmeemroth/ClientVault
How I Built It
Built with Python and FastAPI/Gradio, powered by Google's open-weight Gemma 2 (2B) model running local inference via Ollama. Custom system prompts handle entity redaction ([CLIENT], [REDACTED]) and structured financial table extraction entirely in local memory.
Why Does Open Innovation Matter?
Uploading NDA-protected documents to closed cloud APIs breaches client confidentiality and carries per-token costs. Running Gemma locally guarantees zero data leakage outside the device and costs nothing per run.
My Agent Session
Used an AI pair-programmer to craft the entity-redaction prompts for Gemma and generate the turnkey render.yaml configuration for one-click deployment.
My First User
A freelance software consultant tested it on a confidential fintech SOW to draft an anonymized portfolio case study. He confirmed he would pay a one-time $20 fee for an offline tool that protects him from NDA breach disputes.
Prize Categories
- Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle
- Best Use of Gemma
- Best Use of Render
來源：DEV Community · dev.to