SaiLoR v1.9.1

New

• AI assistance (preview). SaiLoR can now fill annotation fields and screening decisions with a language model.
  • Use a hosted model or a local one (Ollama, llama.cpp).
  • Use prompt mode, or agent mode with a judge that checks the evidence quotes.
  • Let the AI take its own reviewer seat. Its work stays separate from the human seats.
  • Check the cost estimate and set a spending cap before you run all papers.
  • Run AI screening, re-check answers, and get schema feedback.
  • Each paper keeps a record of the AI runs that wrote into it.
  • The AI windows are marked as a preview feature.
• View-only mode. Tick View only in the toolbar to look at annotations without changing anything.
  • The annotation form, screening panel, PDF highlights and notes are locked.
  • Undo, redo and AI runs are blocked.
  • On the Consolidation seat, the reviewer comparison does not start. SaiLoR shows only the stored consolidation.
  • The setting lasts for the session. It is off after a restart.
• New toolbar layout. The toolbar now groups file and Git actions, review tools, and status.

Fixed

• Fixed some tests with dependencies on the system locale.

Release and archive

• Each release is now archived on Zenodo as a new version of the SaiLoR record. The following DOI points to the latest version: 10.5281/zenodo.23804729.

Notes

• The installers are not code-signed. macOS and Windows show a warning on first start.
• Auto-update works on Windows and Linux only.