Disclaimer
Is this project open source? Yes
Is this project completely free? Yes
Is this project vibe-coded beyond the author’s ability to comprehend how it works? No
Community Directory: Smart Related Notes
Hi all,
I wanted to share a plugin I have been building lately. Smart Related Notes adds a sidebar panel that shows the notes most related to the one you are reading, ranked by semantic meaning rather than by shared words or folders. Open a note and it surfaces the rest of your vault that actually connects to these concepts. It particularly suits the atomic note taking style where you create notes for concepts and don’t rely so much on folders. I switched recently to that, which caused me to begin building this.
It runs a small embedding model entirely on your machine. No cloud, no API key. After a one-time model download it works fully offline, and your notes never leave your computer. It understands German, English and a bunch of other languages, so it matches across languages too.
A few things it does:
- Ranks related notes in the sidebar as a stack of cards you can click.
- Suggests inline links while you write, when a concept already has a note. (If you have unrelated notes like analysis for math that do not fit the context it does not show the suggestion)
- Finds related-but-unlinked notes, orphans and near-duplicates so you can tidy a messy vault (still work in progress not out at the moment)
- Lets you wall off an area (say a novel) so it stays self-contained and never bleeds into the rest.
As part of the 2.0 release the plugin now works out the structure of your vault from meaning alone, with no tags and no links required (of course having them will improve edge case performance and overall raise the bar).
For example: how does it know that a note about the One Ring is an item and not a character? Nothing tells it. The note is about an object with object language (forged, golden, an artifact), so its vector lands near the other objects, while a character note lands near the other characters. The categories fall out of the way you write itself.
Here is the full concept for anyone interested.
Unfortunately it reverts back to the small display no matter what size I set 
What is new in 2.0:
- Multi-granularity embeddings: each note is understood as a whole, as ideas, and as passages, so a link by a single idea still surfaces.
- A smarter ranking funnel and a much better sidebar search.
- An optional long-context model (jina-embeddings-v5) for people who want maximum quality. It is opt-in because it is a bigger 300mb, non-commercial download; the default stays small, fast and offline.
- A fresh install now asks you to pick a model before it embeds anything, and there is a model manager to remove ones you no longer want.
I would love feedback, and I am especially curious whether the meaning-based matching surfaces useful connections in your own vaults. Happy to answer anything about how it works under the hood.
Thanks for reading.