The graph
What a knowledge graph of your notes actually shows you
A knowledge graph of your notes draws every note as a node and every link as an edge, then lets physics arrange it. It is useful for three specific things: finding clusters of thinking you did not plan, spotting orphaned notes nothing references, and seeing which two areas of your work have never once been connected.
Why a graph and not a folder tree
A folder tree encodes one relationship: containment. It answers "what is inside what", which is a question you rarely have. A graph encodes the relationship you actually care about — "what is connected to what" — and because a note can link to any number of others, it does not force the single-category decision that folders do.
The tree is still useful for storage. The graph is what you use for thinking.
Three things a graph tells you that search cannot
- Clusters you did not plan. After a few hundred notes, dense regions appear where you have been thinking hardest. They are usually not the topics you would have listed if asked.
- Orphans. Notes nothing links to and which link to nothing. Some are fine — a shopping list is not supposed to be connected. Others are ideas you captured and then abandoned, and seeing them is often the prompt to do something with them.
- Structural holes. Two dense clusters with no edge between them means two bodies of your own knowledge that have never been brought into contact. That gap is frequently where the interesting work is.
Reading a graph properly
A few conventions make one legible rather than pretty. Node size by degree — how many notes link to it — puts your hub notes visually forward. Colour by folder or tag turns the graph into a map of your own categories rather than an undifferentiated hairball. And a local graph, showing only the neighbours of the note you are in at an adjustable depth, is often more useful day to day than the global one.
Ghost nodes are the underrated part: links to notes that do not exist yet, drawn as hollow nodes. They are a to-do list made of your own intentions, and in Nodum clicking one creates the note with the backlink already resolved.
The engineering, briefly
Force-directed layout is expensive — every node repels every other node — and most graph views quietly stop being usable somewhere in the low thousands. Nodum renders the graph on WebGL2 through @cosmos.gl/graph, running the force simulation on the GPU, which is what keeps it smooth into the tens of thousands of notes.
The interaction matters as much as the frame rate: draggable nodes, hover highlighting of neighbours, live force sliders for centre, repulsion, link force and link distance, search and tag filters, and labels that fade with zoom. There is also a guard so that a lost WebGL context — which browsers do, unprompted — cannot take the workspace down with it.
Graph data is cached in Redis and invalidated on note and link writes, so opening the graph on a large vault does not mean recomputing it.
Questions people ask
What is a knowledge graph in a note-taking app?
It is a visualisation where each note is a node and each link between notes is an edge, arranged by a force simulation so connected notes pull together. It shows clusters, orphaned notes and gaps between topics — structural information about your notes that search cannot surface.
Is the graph view actually useful, or just decorative?
It is genuinely useful for three things: finding clusters of thinking you did not plan, spotting notes nothing references, and noticing two areas of your work that have never been linked. It is not useful for navigation — search and a quick switcher are faster for getting to a specific note.
How many notes can a graph view handle?
Most implementations degrade in the low thousands because force-directed layout is quadratic in the number of nodes. Nodum runs the simulation on the GPU via WebGL2, which keeps it interactive into the tens of thousands of notes.
Related reading
Comparing specific tools
- Nodum vs ObsidianNodum is a free, MIT-licensed Obsidian alternative for the web: same [[wikilinks]], same backlinks, same graph — no install, and you can read the source.
- Nodum vs NotionNodum is a free, open-source Notion alternative built on plain markdown, [[wikilinks]] and a knowledge graph — self-hostable, exportable, no per-seat pricing.
- Nodum vs LogseqNodum and Logseq are both open source and both link notes. Nodum is a document editor in the browser; Logseq is a local-first outliner. Here is the honest split.
All 18 are on the alternatives index. Terms used on this page are defined in the glossary.
Start tying things together.
A vault takes about ten seconds to make. Bring notes with you, or start with one.