Engine
How this workspace processes your files, and what is held on this device.
Offline OCR
The Tesseract engine and its WebAssembly core are served from this app. The English language pack is downloaded once and then kept on the device. Do this while you have a connection so scanning works in court with none.
Pipeline
- 1. Extraction. PDFs are read through their text layer; pages with no text layer are rasterised and passed to Tesseract. Images always go to Tesseract. Workbooks, Word files and text files are parsed directly.
- 2. Layout analysis. Word positions from either engine are run through a projection-profile column detector and a gap-alignment table detector, so multi-column decisions and tabular exhibits keep their reading order.
- 3. Retrieval index. Blocks are chunked per page and indexed with BM25 in this browser's database.
- 4. Analysis. The four case outputs are produced by quoting the record and applying named rules of court — no model is involved, so nothing can be invented.
- 5. Assistant. Offline it answers with retrieved passages only. Online it reasons over the same passages through the Legal MCP tool surface, under a strict no-hallucination instruction.
On this device
- Documents in workspace
- 0
- Retrieval chunks indexed
- 0
- Law library provisions stored offline
- 0
- Archived sessions
- 0
- Pages extracted
- 0
- Words extracted
- 0
Connection status: offline — local engine only. Install this app from your browser menu to keep it on the home screen and open it without a connection.