If anything fails, the copy waits in the inbox. Docket never moves, renames, or edits the source file.
Document filing for the command line
Docket reads it.
Then files it.
Add a PDF, scan, or photo. Your model reads it. Jev classifies it, scores sensitivity and urgency, and flags what needs your attention. Docket files a copy without touching the original.
Open source · Bring your own OpenAI or OpenRouter key · Classification by Jev
- Category
- HOUSING
- Confidence
- 94%
- Sensitivity
- 1.8 / 3 · CONFIDENTIAL
- Urgency
- 0.3 / 3 · NO DEADLINE
- Needs action
- 12%
- Next step
- FILE FOR REFERENCE
- Handling
- LIMIT SHARING
- Review
- NOT REQUIRED
~/Documents/Docket/housing/
If it’s worth keeping, drop it in.
Record 001 / what happens to a file
Your original stays put. Docket files the copy.
Docket works from a copy, so the original stays where it is. Your model reads the document. Jev identifies what kind of document it is and whether it looks sensitive, urgent, or likely to need action.
Text files are read locally. PDFs and images go only to the provider and model you selected.
- legal
- 4%
- other
- 2%
- sensitivity
- 1.8 / 3
- needs_action
- 0.12
If Jev is unsure, Docket shows the other likely categories and marks the document for review.
SQLite saves each document’s name, category, sensitivity and urgency scores, action estimate, and copied-file location. API keys stay in the OS keychain. Docket discards extracted text after classification.
Made for the command line
Drag a path in.
Keep your hands on the keys.
Use OpenAI or OpenRouter to read documents, then Jev to classify them. Docket keeps the results and copied files on your machine.
Add files without dragging paths into the terminal. Docket will still leave every original untouched.
Read scans on your machine without an API key or upload.