Our summary
Folders of vaguely named scans are difficult to search. This recipe reads text documents, asks Jev to choose a document type, and identifies the document's own date. It first prints a proposed filing plan rather than immediately changing the files.
Ordinary software finds possible dates and keeps nearby words that explain them. Jev chooses from those candidates, separating an issue date from a deadline or birthday. Applying the plan copies files into new folders; low-confidence classifications go into a review folder and originals remain untouched.
The recipe directly accepts text and Markdown files, so scans need a separate text-reading step first. It reads only a limited opening portion of each document and assumes day-first slash dates. Those choices can miss useful information or misread documents from another country.
Key takeaways
- Inspect the proposed plan before making copies.
- Keep surrounding words when asking which date matters.
- Check date conventions and text extraction before trusting the filing result.