// how it works
a private folder scan first, optional AI review second.
KeplerDocs is being built as a folder secretary for people whose work lives in local project, client, finance, legal, research, and operations folders. The web app is a preview of the workflow: map a folder, understand its risk, and decide what needs attention.
// scan workflow
01
choose a folder in the browser preview
The current KeplerDocs app is a browser preview. You choose a local folder from your device, and the browser grants temporary read access for that folder. KeplerDocs does not get broad access to your computer.
02
the folder stays local
By default, your files are not uploaded. The scanner reads names, paths, extensions, sizes, modified dates, and other local metadata so it can build a private map of the folder.
03
KeplerDocs builds a file map
The scan turns the folder tree into structured entries with stable file IDs, categories, fingerprints, warnings, and timestamps. That gives the app a reliable basis for reports, rescans, and change detection.
04
the health score explains folder risk
KeplerDocs calculates a deterministic folder health score from signals such as stale files, duplicate groups, unusually large files, sensitive filenames, unreviewed files, and scan warnings.
05
AI analysis is optional
AI review only runs when you request it. The app can send limited text previews for supported files to generate summaries, sensitivity ratings, and review guidance. This is separate from the default folder scan.
06
native folder workflows are planned later
A future native workflow can make recurring scans and long-running folder monitoring more dependable than browser folder access. The current web preview remains read-only.
// what the report looks for
stale files that have not changed in 90, 180, or 365 days
duplicate groups based on exact, likely, and variant file names
large files using category-aware thresholds
new and recent files from the last 7 and 30 days
sensitive filenames such as credentials, legal, identity, and finance material
scan warnings for skipped files, hidden folders, aborted scans, and browser limits
// trust model
no default file uploads
A normal scan works from local folder metadata. Files are not uploaded by default.
read-first product model
The current app reports on folders. It does not move, rename, archive, or delete files.
visible scan limitations
The browser preview shows scan warnings and max-file caps so users know when a report is incomplete.
built for a native future
The architecture is being prepared for a later native workflow where recurring folder reviews can be more dependable.
This page describes the current product foundation: local folder mapping, metadata extraction, deterministic detection, folder health scoring, optional AI analysis, and a planned native workflow later.