paperless-classify

Paperless-ngx document classification supports bulk batch updates from stdin using pipe-delimited fields. This allows you to process multiple documents at once, removing the inbox tag and preserving existing metadata when optional fields are omitted from the input.

For ambiguous documents during taxonomy assignment, use a compact-to-full content strategy. Run the classifier initially with brief content for most docs, then selectively re-run with full content analysis on ambiguous cases to improve classification quality and accuracy.

Yes, you need a prepopulated taxonomy and the inbox tag workflow enabled before classifying documents. The classifier aligns assignments to your existing taxonomy conventions, prompting for creation only when taxonomy gaps require it, while correspondents are always created automatically.

Yes, existing metadata is preserved when fields are omitted during updates. The classifier CLI interface ensures that bulk batch updates only modify the fields you explicitly provide, removing the inbox tag while leaving other correspondent, type, and tag assignments intact.