sist2app/sist2 v4.1.0

on GitHub

Breaking changes:

• Elasticsearch mappings changed. Run index with --force-reset.

New features:

• Email messages (.eml) and mbox mailboxes are now supported
• Support for chunked embeddings (SQLite only for now)
• The media type filter has select all / none / invert buttons. Search can be filtered on several paths at once
• scan --no-stats skips statistics generation, and web can pick the default UI theme for new visitors

Bug fixes:

• WordPerfect files are parsed again, and media inside an ogg container is no longer skipped
• An Elasticsearch request that stops answering is given up on instead of hanging forever
• A search with an unknown embedding model no longer kills the web server
• A document is ranked by its best embedding instead of once per embedding, so a long document no longer floods the results
• Paging through results that tie on a descending sort no longer repeats or skips documents
• Tagging a document whose tag name contains special characters no longer fails
• sist2-admin lists the OCR languages found in the tessdata folder, so they match what sist2 can actually use
• A task appears in the sist2-admin history when it starts, and its page shows which index file it uses

docker pull sist2app/sist2:4.1.0

Full Changelog: v4.0.5...v4.1.0