Mesa

项目介绍

Mesa is a spatial canvas for your files and media: instead of folders, you arrange files, notes, bookmarks and widgets on a zoomable surface inside named, nestable frames, so position itself keeps your thinking. Files open where they sit — a PDF renders inline page by page, a video plays in place, an audio file shows its waveform — and the built-in AI agent shares your view: ask it to find, summarize or rearrange content and it works on exactly what is on screen, with a single undo restoring any layout. You can connect Google Drive, which turns a folder hierarchy into a navigable space, or upload directly from computer or phone. The public demo needs no account and nothing is saved; the agent runs in the account version, which the author is currently sharing selectively. Mesa is a four-month solo project of over 100,000 lines of TypeScript, with a real Postgres database and end-to-end tests against a real database container, running in production on a Google Cloud VM. Note for readers: the no-account demo does not include the agent; the agent is shown in the author's videos and available on request.