Skip to main content
LLM Matrix Lab AI Workbench
Legal Document • Last Updated: August 14, 2026

Privacy Policy

LLM Matrix Lab is committed to protecting your privacy. This document outlines how our client-side tokenizer application handles data and user inputs.

1. Overview & Commitment

LLM Matrix Lab is built with privacy as a core principle. All text encoding, token segmentation, byte representation calculations, image patch grid analysis, audio spectrogram tokenization, 3D WebGL rendering, and neural network training occur entirely within your web browser session. We do not transmit, log, or store your text prompts, uploaded images, audio samples, or drawn canvas data on remote servers.

2. Data Collection & Processing

We do not collect personal identifying information. Here is how your inputs across our tools are handled:

  • Text Inputs: Processed in real-time via WebAssembly (Tiktoken WASM) or local JavaScript tokenizer algorithms.
  • Image Uploads: Processed strictly using client-side HTML5 Canvas API elements for spatial patch calculation.
  • Audio Samples: Analyzed 100% in-browser using HTML5 Web Audio API for acoustic codebook and spectrogram breakdown.
  • Neural Network & MNIST Drawing Canvas: Neural network weights, backpropagation steps, and user-drawn digits execute strictly within browser RAM memory without remote logging.
  • Prompt Efficiency & Visual Diffs: Evaluated locally with zero external API calls.
  • 3D WebGL Transformer Engine: Rendered on your local device GPU via WebGL shaders.

3. Local Storage & Cookies

LLM Matrix Lab uses browser local storage (`localStorage`) solely to store user preferences, such as:

  • Selected model choice (e.g., `gpt-4o`, `deepseek-ai/DeepSeek-R1`)
  • Whitespace visibility preference toggle
  • Neural Network architecture configuration presets (learning rate, activation function, hidden layers)
  • Editor and 3D visualizer display preferences

No tracking cookies or third-party analytics cookies are set by LLM Matrix Lab.

4. Third-Party APIs & Libraries

LLM Matrix Lab uses open-source libraries and CDN resources for tokenizer vocabulary files (such as HuggingFace tokenizer JSON configs). These static assets are fetched over secure HTTPS connections. No input data is sent to external API providers during tokenization.

5. Your Rights & Control

Since all data remains on your local machine, you have complete control over your data. You can clear your browser storage or cache at any time to remove saved preferences.