Search

Search looks across the whole graph at once: names and the text inside every note. It is the bigger sibling of Quick Find, which matches names only and lives in the sidebar (The Graph Sidebar).

Open it

Press Alt+F or Ctrl+K (Cmd+K on a Mac), select Search everything under the Quick Find box, or type / in the editor and choose Search. If you had typed something into Quick Find, it comes across selected, so you can replace it by typing or keep it with the right arrow.

Two groups of results

Search runs two kinds of matching and shows them stacked, each with its own paging, so a long list of one never buries the other.

Names matches document names the way Quick Find does: pages, journal entries, aliases and names you have linked to but not written yet, exact matches first. It answers at once. If nothing matches, the last row offers + New page with what you typed.

Text finds every mention of your words inside your notes, grouped by the document they are in, so you can see which pages talk about something most. Each document shows how many matches it has, and each match shows the bullets above it and the line it is on with your words picked out. Documents with the most matches come first.

Text matching follows these rules:

• It matches whole words, and treats the last word you type as unfinished: orph finds orphaned while you are still typing, but graph does not find subgraph.
• Put a phrase in quotes to match the words together: "quantum mechanics". An unclosed quote is treated as closed at the end, so a half-typed phrase still works.
• Links count as words. Searching physics finds it whether you wrote it as plain text or as [[Physics]].
• There is no search syntax. Punctuation is ignored and separates words, so C++, don't and orphan - asset all search for the words in them.
• Text results start at two typed characters. Name results start from the first.

Search covers the graph you have open, and only that one. Opening a result closes Search; reopen it and your query is still there.

Keys

Up and Down move one highlight through both groups as one list, stopping at the ends. Enter opens the highlighted row. Escape closes Search without clearing what you typed.

Opening a text result takes you to the matched line, scrolling there if it is further down the page, even when the document is already open in another tab.

While the graph is still indexing

When a large graph has just opened, the index that text search reads is still being built. Name results appear as it goes. Text results are withheld until the index is complete, with a note saying how far it has got, so a search can never wrongly say "not found".

What search never reaches

A protected document stays out of the text results completely. Neither its words nor its encrypted form is put in the search index, so there is nothing to match even while it is unlocked in front of you. Its name is still found. When a graph holds protected content, Search says "Encrypted content is not searched." beneath the text results, so an empty result never misleads you. See Protected Documents.

Search matches words, not meaning: it does not find "due 31 January" for tax deadline. An AI agent connected to a graph through the Headless Client can search by meaning as well (The Headless Client).

Rebuilding the index

Search, backlinks and the Tasks view all read one index that EtherPK builds from your documents and keeps in the browser. It is never written into your graph and can always be rebuilt. If a sidebar disagrees with a document, Settings > Maintenance > Rebuild search index re-reads every document from scratch (Graph Settings).