Your files. Your device.
The portfolio, tools, and synthetic systems have different jobs—and different data boundaries. This page describes how this implementation handles them.
Files stay in the local tool
Audio, images, PDFs, CSV files, and their results are processed in your browser. Tools do not upload their contents to this site’s server or a model API. Files and outputs stay in working memory unless you download them. Clearing or closing a tool discards its in-memory session.
Engines still need a download
Speech and OCR models are downloaded only after you explicitly start the relevant action. Model hosting providers receive ordinary asset requests, including your IP address. Those requests do not contain your audio, document contents, or filenames. Downloaded models can be cached separately from your files.
A synthetic workspace, stored locally
Systems demo records are synthetic and live in this browser’s IndexedDB database. Browser profiles do not synchronize automatically. Export provides a portable backup; import or reset is an explicit action. Browser data can be evicted or cleared by your device. Please do not enter real confidential records into the demo.
A small, local assistant
The visible assistant searches public portfolio information in the browser. Its conversations stay in memory and it never reads tool files, transcripts, or the systems database. The separately documented portfolio API processes requests only when called explicitly.
Hosting and ordinary requests
Vercel serves the website and can process ordinary request metadata for delivery and security. This implementation does not install analytics, advertising, session replay, or tracking pixels. External links open the named third-party service, whose own practices apply.
Questions
Contact Hugo if you find behavior that does not match this page.