Support.
Written by the person who wrote the app.
Three ways, all in the toolbar. Drop image or PDF files onto the window, choose Import to pick them from a file panel, or use Import from iPhone to photograph a page with a nearby iPhone or iPad through Continuity Camera. Whichever you use, the page arrives in the same review step before anything is filed.
Edge detection is a proposal, not a decision. Every imported page opens with draggable corner handles over the detected rectangle — move them and the perspective correction re-renders. If the page has no clear border against its background, detection will find nothing and hand you the full frame, which is the correct behaviour rather than a guess.
Recognition follows the pixels. A photograph taken at an angle, in poor light, or of small print will read worse than a flat, evenly lit page. Recapture rather than fight it: fill the frame with the page, avoid shadow across the text, and keep the camera parallel to the paper. Handwriting is not supported.
Yes. Silo Capture writes the page image with an invisible text layer positioned over the words, which is the same structure Preview, Spotlight, Acrobat and any other reader expect. Select, search and copy work in those apps without Silo Capture installed.
Search runs over recognized text, so a word that recognition misread is a word the index does not have. Open the document and check what was actually recognized. If the page reads badly, recapture it — re-running recognition on a better image is the only real fix.
Inside the app's own sandbox container, as ordinary PDF files alongside a local SQLite index. Export writes copies to any folder you choose. Nothing is stored anywhere else, because there is nowhere else — no account, no sync, no server.
It reads the recognized text for a date, a document type and an issuer, and proposes a filename from what it found. The proposal is always editable before filing, and Settings has a switch to skip it entirely and name every document yourself.
No, and it cannot use it. The app ships with no networking entitlement, so the sandbox refuses outbound and inbound connections outright. There is no account, no model download, and no telemetry — recognition uses the Vision framework already in macOS.
Include your macOS version, which Mac you are on, and what you were doing. There is no account to look up and no telemetry to consult, so what you tell me is all I have.