28 releases 32 tags

RSS feed

• 0.9.0 e0f0f35f9c

  Compare

  0.9.0 Stable

  jkaindl released this 2026-07-07 17:44:04 +00:00 | 240 commits to main since this release

  Added

  • Non-deterministic Smart Apply mode ("additive") — a per-application mode toggle in the Smart Apply cockpit. Alongside the default deterministic mode (the LLM only assigns your original blocks and never invents), the additive mode lets the LLM infer frontmatter values that aren't verbatim in the text and insert marked addition blocks — each carrying an ordinal confidence (high/medium/low) that you accept or reject individually in the diff gate. Low-confidence items are unchecked by default. Your original blocks stay byte-exact either way, and the deterministic path is bit-identical to before. A template can declare its preferred mode via a reserved smartapply_modus frontmatter key (never leaked into target notes), with a global default in settings. An optional "keep provenance" audit toggle records inferred keys as a smartapply_erschlossen field and marks additions with %%erschlossen: …%% comments — or leaves a clean document. Confidence is encoded via icon shape and text, not colour alone (WCAG 1.4.1).
  • Undo ↔ Redo toggle in the Smart Apply diff gate — after "Undo" the button becomes "Redo" and restores exactly what was applied (the final selection plus audit state), so an accidental undo is one click to reverse.

  Downloads

  • Source code (ZIP)
    0 downloads
  • Source code (TAR.GZ)
    0 downloads

  ---

  • main.js
    0 downloads · 2026-07-07 17:45:08 +00:00 · 172 KiB
  • manifest.json
    0 downloads · 2026-07-07 17:45:08 +00:00 · 328 B
  • styles.css
    0 downloads · 2026-07-07 17:45:08 +00:00 · 16 KiB