PriviPaste is a privacy first VS Code extension that redacts Personally Identifiable Information (PII) locally on your machine. Ensure your sensitive data (names, emails, keys, etc.) is sanitized before it is shared with AI coding assistants or external services.
Key Features
Zero Data Leakage: Redaction is performed entirely on your local machine using client-side AI.
AI-Powered Precision: Leverages the openai/privacy-filter model via Transformers.js for accurate entity detection.
Descriptive Redaction: Replaces sensitive data with descriptive tags like [EMAIL_1] or [NAME_1] to preserve semantic meaning for AI models.
Fast Workflow: Sanitize your clipboard with one shortcut (Alt+P) before pasting.
Hardware Accelerated: Uses WebGPU for near-instant redaction where available.
How to Use
Copy any text containing sensitive data.
Place your cursor wherever you want to paste the text (e.g., your code editor, VS Code Copilot Chat, or an AI Assistant input box).
Press Alt+P (or Cmd+Alt+P on Mac).
The extension will automatically redact the PII and instantly paste the safe version right where your cursor is!
FAQ
Q: Does my code leave my machine?
A: No! All redaction is processed 100% locally on your own hardware.
Q: Why does it need to download a model?
A: To keep the extension size small, the AI engine downloads the necessary language model directly from Hugging Face on your very first run. After that, it operates entirely offline.
[!IMPORTANT]
First Run Model Setup: On your first use, PriviPaste will download the openai/privacy-filter AI model from Hugging Face. This only happens once. You can track the progress in the status bar at the bottom of your window. Subsequent redactions will be near-instant!
Privacy & Security
PriviPaste is built with a "Local First" philosophy:
No Cloud Calls: The AI model runs entirely in your VS Code process.
No Telemetry: We do not collect or transmit any data from your files or clipboard.
Open Integrity: Built using open-source models and libraries.
Requirements
VS Code 1.80.0+
(Recommended) A WebGPU-compatible system for best performance.
