Your documents,
found by meaning

Drag a PDF or TXT file into the app. CtrlSense saves it locally and queues it for processing.

The extractor reads every page of the PDF and returns one clean string of all content — ready to be embedded.

The text is fed into all-mpnet-base-v2. The gold-standard 420MB model outputs 768 numbers encoding the document's deep semantic meaning with far greater accuracy than smaller models.

The vector is saved in ChromaDB — a local embedded database. No server needed. Every dot on the map is one of your documents, placed by meaning.

Your query is embedded the same way. ChromaDB finds the nearest vectors by cosine similarity and returns ranked results in under 100ms.

When you click a result, the original file opens instantly. The app scrolls straight to the matched paragraph and applies a pulsing golden glow to the exact keywords — no hunting required.

A background thread watches your dedicated folder at Documents\Ai Assistant Input. The moment you drop a file in — even from another app — CtrlSense wakes up, extracts text, and injects it into the vector brain. No button clicks needed.

Press Ctrl + Shift + D from anywhere on your computer — mid-game, mid-browser, mid-document — and the CtrlSense overlay instantly snaps to the front, ready for a semantic query. No minimizing windows needed.