1.5.0

Added

• Semantic link suggestions in the editor (#35): Type a trigger (default ;;) followed by a rough description — e.g. ;;book with zombie — to get inline semantic suggestions and insert a [[link]], even when you've forgotten the exact title. Configurable (or disable) under Display settings.
• Errored notes are visible and retryable (#45, #46): Notes that fail to index now show in an "Errored files" list with the reason, plus a "Retry errored" button and command. Editing a note retries it automatically.

Changed

• Index settings reorganized: The crowded Index section is split into three focused sections — Index, Exclude folders, and Exclude content.
• Excalidraw excluded by default (#46): New installs skip the Excalidraw/ folder (its drawings can't be meaningfully embedded). Existing exclusion settings are unchanged.

Improved

• Faster Ollama indexing (#46): A note's chunks are now embedded in batched requests instead of one round-trip per chunk — most noticeable on large vaults and small-context models.

Fixed

• Non-English & token-dense notes failing to index on Ollama (#46): CJK/multi-byte content and table/code/number-heavy notes no longer fail with "input length exceeds the context." Chunk sizing now uses UTF-8 byte length and is capped to the model's real context, with /api/embed truncation as a backstop. bge-m3 (8K context) has the most headroom.
• Honest indexing status (#45, #46): Indexing failures are no longer hidden behind a "fully indexed" report. Settings now show separate Indexed / Errored / Excluded counts, and terminally-errored notes are no longer re-attempted on every restart.