Skip to content

The graph

The Graph draws your notes as dots and the links between them as lines.

Clicking a dot selects it, highlighting that note and its links so they stay picked out while you pan around; clicking it again clears the selection. Double-click a dot to open the note. -click (Ctrl-click on Windows and Linux) opens it without disturbing the selection, so you can work through a highlighted group.

Vault shows everything. This is the impressive one and the less useful one — past a few hundred notes it becomes a dense cloud that tells you little beyond “yes, things are connected”.

File shows one note and its immediate neighbors, labeled. This is the one worth using. Open a note, switch to File, and you can see at a glance what it draws on and what draws on it.

Not decoration, though it is decorative. Three real uses:

  • Finding the orphans. Notes with no links are usually notes you’ll never find again. The graph makes them obvious.
  • Spotting the hubs. A note with a lot of connections is usually an important one — often one you didn’t plan.
  • Wandering. Sometimes you don’t know what you’re looking for, and following links visually turns up things search wouldn’t, because you didn’t know the word to search for.

By default a line is a link you wrote. Switch Links to Semantic in the settings drawer and the lines become notes that are about the same things — found by Semantic search, so it has to be turned on first. Both keeps your links and draws the semantic ones dashed beside them.

Two sliders shape it: Neighbours per note (how many similar notes each note reaches for) and Minimum similarity (how alike two notes must be before a line is drawn). Nudging them is instant. The first time you switch, Pellet works out every note’s nearest neighbours; a few seconds for a few thousand notes, a minute or two for a very large vault, with a progress line and a Cancel. After that it remembers, and only re-checks the notes you change.

Semantic links are a desktop feature for now.

Groups colour dots by a rule, the way Obsidian’s do. Open the drawer, add a group, and type a query:

  • path:Projects — everything under a folder (path:/Projects for the folder at the top of the vault only)
  • tag:#work — notes with a tag, including nested ones like #work/deep
  • file:Draft — notes whose name contains a word (a bare word does the same)
  • path:Projects -tag:#done — several terms must all hold; - leaves something out

Each group has a colour and a switch. The first group that matches a note wins, so drag the important ones up. Groups travel with your vault settings, so every device draws the same colours. A grouped note keeps its colour when you hover or open it — it still stands out, by growing while everything else fades back.

You can narrow what’s drawn with conditions on four things: Folder, Title, Tags and Kind — the same kinds the all-files table uses, so “Kind is Canvas” means the same in both. Dates and status aren’t offered: those are questions for the task views, not for a map of links.

Conditions apply to attachments too, so filtering to a folder no longer leaves its images behind. Asking for an attachment kind on its own, though, draws nothing: an attachment only reaches the graph through a note that links it, and a note is never an image.

Alongside them are switches for what to include at all: attachments, unresolved links, and Orphans — notes with no links either way. Turning orphans off is a quick way to see the structure rather than the fringe.

Dots grow with how many links they have, so the hubs stand out without a label. Hovering a dot lifts it and its neighbours out of the crowd, names it, and says how many links it has; the neighbours are named too, as many as fit without the words landing on each other. Zoom in for the rest. Zoomed out, names fade away until you hover or zoom back in. Attachments and links to notes that don’t exist yet are drawn hollow when you choose to show them.

Names are only drawn where they fit. Pellet shows the most-linked notes first, skips any name that would land on one already drawn, and stops at 120 — so a big vault reads as a map with its landmarks labelled rather than a wall of text. Zoom in and more names appear as the room for them opens up.

Adjust dot size, link thickness, labels, and how hard the layout pushes things apart. Worth a minute of fiddling: the default suits a medium vault, and a very large or very small one reads better with a nudge. The four force sliders run from 0 to 10 with the default in the middle at 5, so there is always a way back to where you started.

The layout runs separately from the interface, so a few thousand notes stay draggable instead of freezing the window while it thinks.