Skip to content

Find related

Rank notes worth linking by shared graph context (co-citation and coupling).

Called find-related in the step's YAML.

A Relations action. It asks nothing in the wizard. It reads only the knowledge model: relatedness is purely structural, with no text or AI inference. Deterministic and offline.

Settings

  • Result property — where the ranked list of links is written. Default: related.
  • Write to — Frontmatter (a property on the note) or Context (only for the next steps). Default: Frontmatter. Either way the next steps can use it as {{<result property>}}, e.g. {{related}}.
  • Source note (optional) — the note to rank against. Empty uses the note being built.
  • Max results — how many suggestions to keep, strongest first. Default: 10.

What it writes

---
related:
  - "[[Spaced repetition]]"
  - "[[Writing to think]]"
---

A notice reports the count. An empty list is a valid result — in particular, the note being built is not indexed yet, so ranking it before it exists finds nothing; point Source note (optional) at an existing note to get suggestions. If the index isn't ready, the action does nothing.

How the ranking works

For each candidate note the score combines two graph signals:

  • Co-citation (weight 2) — how many notes link to both the source and the candidate.
  • Bibliographic coupling (weight 1) — how many notes the source and the candidate both link to.

score = 2·co-citation + 1·coupling. The source itself, notes already linked to it (either direction) and candidates with a score of 0 are left out. Ties are broken by path.

Capabilities

File-system read + the disclosed result-property write. No network, no AI.