Finding & Navigating

Reaching a Rem, a source, or a figure that RemNote's own search will not surface.

Find Rem — Reference or Open

A floating picker that finds a Rem by name even when RemNote's own reference search can't, then either inserts a reference to it at your cursor or opens it in a new pane. Built for the frustrating case where a perfectly normal Rem — even a Concept referenced dozens of times — simply never appears when you type its name in the [[ reference search.

How to invoke

Run Find Rem (insert reference / open in pane) (quick code fir) or press Opt+Shift+F / Alt+Shift+F. A compact box opens at your cursor.

• Type a name → results appear as you type, with the best matches floated to the top (an EXACT badge marks an exact-name match; an ALIAS badge marks a match found through one of the Rem's aliases). Each row shows the Rem's type badge (see Spotting PDF highlights), a 🖼️ icon when it holds an image (see Spotting figures), its back text (the definition side of a Concept↔definition card), and a short breadcrumb (root / … / parent) so you can tell which document it lives in when names collide.
• Enter or click → inserts a reference to the selected Rem at your cursor.
• Ctrl+Enter / Cmd+Enter or Ctrl/Cmd+click → inserts the reference as a pin (the link chip without the referenced text). See Insert as a pin below.
• Opt+Enter / Alt+Enter or Opt/Alt+click → inserts the Rem's text followed by a pin — the readable text plus a link chip (RemNote's paste "Text with Pin"). See Insert text with a pin below.
• Shift+Enter or Shift+click → opens the selected Rem in a new pane beside your current one (without inserting anything).
• Ctrl/Cmd+Shift+Enter or Ctrl/Cmd+Shift+click → appends a pin at the end of the Rem you are editing, keeping your selected text. See Pin a source at the end of a Rem below.
• ↑/↓ navigate · Esc closes.
• Concepts only checkbox narrows results to Concept-type Rems.

The Rem you triggered the picker from is excluded from results — a Rem can't reference itself.

Spotting PDF highlights

A PDF highlight carries the same DEFAULT_TYPE as any plain Rem, so the type badge alone couldn't tell a highlight apart from your own note of the same sentence — a common collision, since a note is often worded exactly like the passage it came from. Highlights are badged PDF HIGHLIGHT in amber instead, so you can see at a glance which result is the source passage and which is your note, and reference the one you meant.

Spotting figures

A result whose Rem holds an image is marked with a 🖼️ before its name. An image contributes no searchable text, so a figure Rem is identified only by its caption — and captions are often near-identical to the prose that discusses them (Figure 6.4: Definitions used on turning circle beside a highlight reading "… consists of (Fig. 6.4)"). The icon tells you which row is the actual figure.

This reads the Rem's own text directly, so it is accurate whether or not you have ever run Tag Rems With Images — it does not depend on the HasImage tag being present or up to date.

Why it finds Rems the normal search can't

RemNote's reference search builds its candidate list per token, with a cap. When every word in a Rem's name is high-frequency in your knowledge base (e.g. Navegação Interior, mar territorial — where both navegação/interior and mar/territorial appear in hundreds of Rems), the exact-name Concept never makes any token's candidate cut, so typing its full name returns a flood of partial matches but not the Rem itself. This is a property of the search ranking — not a corruption of the Rem — so "Reload Search Cache", retyping the name, or changing its type do not fix it. (You can confirm all of this on a specific Rem with the Search / Linkage Diagnostics tool in the Debug Widget.)

This picker sidesteps the limitation: it searches each word of your query separately, unions the results, keeps only the Rems whose name contains all your words, and floats exact-name matches to the top. Because a distinctive word (e.g. interior) does return the Rem, it reliably surfaces — then ranking puts the exact match first.

It also asks for the whole query as a phrase, which is the one thing that pulls back a small, precise set instead of a truncated flood. RemNote only looks for an exact run of adjacent words when the phrase it is handed is at least two words long, and it matches those words literally apart from the last one. That matters for names built from common words plus a number: a query like fig. 6.4 seeds the word fig, which in a knowledge base full of figures returns thousands of Rems and is cut back long before yours is reached — whereas the phrase figure 6.4 matches only the handful of Rems that actually contain those two words side by side. Since the literal spelling decides whether the phrase matches, both spellings are asked for (see Figure = Fig = Fig.).

Find by alias

Like RemNote's native [[ search, this picker also matches a Rem by its aliases (the built-in Aliases powerup — the alternate names you add via Edit or Add Alias). Earlier versions only matched the Rem's primary name, so a Rem named Via navegável with an alias vias navegáveis would not appear when you typed the alias.

When a result's primary name doesn't contain every word you typed, the picker consults that Rem's aliases and matches against them too. An alias match is shown with the alias text as the row's title, an ALIAS badge, and the Rem's real name underneath (↳ Via navegável) so you know which Concept it links to.

Picking an alias match inserts a reference to the owning Rem that renders the alias text — exactly the shape RemNote uses for its own alias references (the reference's aliasId points at the matched alias). So inserting the vias navegáveis result links to the Via navegável Concept while displaying "vias navegáveis". Alias matches work with every insertion mode — normal reference, pin, and cloze-aware insertion.

Recommended use cases

• When RemNote's [[ search won't find a Rem you know exists (common-word names, heavily-referenced Concepts). Use Opt+Shift+F instead of [[ to insert the reference.
• To open an "invisible" Rem. Shift+Enter / Shift+click jumps to the Rem in a new pane — the practical way to reach Rems the normal search buries.
• To reference a Rem inside a cloze deletion (see below).

Insert as a pin

A pin is a Rem reference that renders as just the link chip without the referenced Rem's text. In native RemNote, creating one is fiddly: you insert the reference with [[, then right-click it → Edit or Add Alias, clear the text, and press Enter. This picker does it in one keystroke — press Ctrl+Enter / Cmd+Enter (or Ctrl/Cmd+click a result) instead of plain Enter, and the reference is inserted with pin: true.

Pins are cloze-aware and selection-aware just like normal references: inside a cloze the pin stays inside the cloze, and a pin can replace selected text.

Insert text with a pin

Press Opt+Enter / Alt+Enter (or Opt/Alt+click a result) to insert the Rem's text spelled out, followed by a pin — the same result as RemNote's paste dialog option "Text with Pin", but in one keystroke and without copying first. Use it when you want the reference to read as normal prose and carry a link back to the source — a flow that's handy when you're weaving a referenced Rem into a sentence.

It brings across the source's full rich text, not just a plain label:

• Formatting and images are preserved — bold, italic, inline colours, LaTeX, embedded images, and so on come along exactly as they appear in the source Rem.
• Front/back cards bring the back too. If the source is a Concept↔definition (or any front/back) card, the back text is included after the front, joined by a practice-direction arrow (⇒, ⇐, or ⇔) instead of RemNote's card delimiter.
• The source's clozes are marked, not re-clozed. Any cloze deletions in the source are inserted as highlighted text (yellow background + reference-coloured font) — the same visual mark the plugin's cloze command (Opt+Z) leaves behind — so they still read as clozes without becoming functional clozes in your target (no "clozes out of clozes").

Alias matches use the alias Rem's text; the trailing pin still links to the owning Rem.

Pin a source at the end of a Rem

The other three modes all insert at your cursor, and — like RemNote's own [[ — a text selection is consumed: the words you highlighted are replaced by the reference. That is wrong for one very common move: linking a flashcard back to where it came from. There, the selected text is the card, and the reference belongs at the end of the line, out of the way.

Press Ctrl/Cmd+Shift+Enter (or Ctrl/Cmd+Shift+click a result) to do exactly that: a pin is appended at the end of the Rem, and your selected text is left untouched. The pin lands at the end of the field the selection is in, so highlighting something on a card's back appends to the back, not the front. When the selection sits inside a cloze, the pin stays outside it — a pin at the end of the line is a source marker, not part of the answer.

The queue flow this is built for. You are reviewing a card and want to check where it came from — say the PDF highlight behind it:

1. Peek at the source while the card is in front of you.
2. Select part of the card's text — the phrase you would search for.
3. Press Alt+Shift+F. The picker opens with that text already in the search box, so the source Rem (its PDF HIGHLIGHT badge makes it easy to spot — see Spotting PDF highlights) is usually the first result.
4. Press Ctrl/Cmd+Shift+Enter.

The card keeps reading exactly as before, and now carries a small chip at the end that takes you straight back to the highlight next time the card comes up — no re-searching, no broken wording.

Cloze-aware insertion

Both RemNote's native [[ and this picker would normally break a cloze deletion if you inserted a reference inside it — the new reference would land outside the cloze. This picker is cloze-aware: when your cursor (or selected text) sits inside a cloze, the inserted reference is stamped with that cloze's id, so it stays inside the cloze instead of splitting it.

Accent-insensitive & selection-aware

• Accent/diacritic-insensitive: typing navegacao interior matches Navegação Interior.
• Figure = Fig = Fig.: figure abbreviations are treated interchangeably, so typing fig 4.3 lists a Rem named Figure 4.3, and typing figure 4.3 finds one named Fig. 4.3 or Fig 4.3. Any capitalisation works and the trailing dot is optional. It's folded the same way accents are — the standalone word fig/fig. is canonicalised to figure in both your query and each Rem's name (and alias) before matching, so an exact match still ranks first with its EXACT badge. Only the whole word is affected: figs, configure, etc. are left alone. The search itself is run in both spellings too, not just the matching: fig. 6.4 is looked up as fig. 6.4 and as figure 6.4, because RemNote's phrase lookup matches the words literally and would otherwise never return Figure 6.4: … for a query written Fig. (see Why it finds Rems the normal search can't).
• Selected text seeds the search: if you select text before invoking, the box opens pre-filled with it (and selected, so you can refine or overwrite). On insert, the selected text is replaced by the reference — exactly like RemNote's [[ behaviour where selected text becomes the link.

The reference is inserted into the editor that was focused when you opened the picker. In the rare case RemNote has no active editor caret at insertion time, the picker copies the reference to your clipboard instead and tells you to paste it.

Stays on-screen near the edges

The picker opens at your cursor, but it now keeps itself fully visible instead of being clipped by the window edge:

• Near the right edge it flips to open to the left of the cursor.
• Past the vertical midpoint it flips to open above the cursor (like RemNote's own selection search), and the results list is capped to the space available on whichever side it opens, so a long list scrolls rather than running off-screen.

This all happens before the box appears, so you never see it jump.

---

Open Source in Popup

Opens the PDF or web article behind a reference pin without leaving the queue. Built for the moment in review when a flashcard (or any Rem) carries a pin to a PDF highlight and you want to glance at the surrounding source — clicking the pin directly navigates away and tears down the queue (you lose your position and the ability to rate the card). This shows the source on top of (or beside) the queue instead, so you read it and dismiss it without interrupting the session.

It comes in two variants — a centered modal popup and a non-blocking floating window — that share the same reader; pick whichever fits the moment.

Two ways to open it

Both are triggered the same way — hover the reference pin, then press a shortcut — and both only act on genuine Reader sources (see What it opens).

	Modal popup	Floating window
Command	Open Hovered Source in Popup	Open Hovered Source in Floating Window
Shortcut	Opt+O / Alt+O	Opt+Shift+O / Alt+Shift+O
Placement	Centered, large	Right portion of the screen (≈48% width)
Blocks the UI behind it	Yes — a backdrop covers/dims the card	No — the card/editor stays visible beside it
Best for	"Open → read → close → rate" in one focused glance	Peeking back and forth between source and card without close/reopen
Closes on	✕, Esc, or clicking outside	✕, Esc, or advancing to the next card (does not close on outside click, so you can highlight/select in the PDF)

Why hover + a shortcut, instead of right-click? RemNote exposes a hover event for references to plugins but no right-click/context-menu event, and the plain left-click navigation can't be intercepted. So the queue-safe path is: hover to identify the reference, then a shortcut you own to open it. The plugin tracks only the last-hovered reference and resolves it when you press the key.

How to invoke

1. Hover the reference pin (the link chip) you want to open. Hovering is the signal the plugin listens to — no click needed.
2. With the pin still hovered, press Opt+O for the modal popup, or Opt+Shift+O for the floating window.
3. The source renders inside RemNote's own PDF/web reader. The queue stays live; close it to return exactly where you were.

What it opens

The command inspects the hovered reference's target and only acts on Reader sources (identical for both variants):

Hovered reference points to…	Behavior
PDF highlight	Opens the host PDF and auto-scrolls to the highlight
HTML / web-article highlight	Opens the article (Reader Mode) at the highlight
PDF source document (the uploaded file Rem)	Opens the PDF at the top
HTML source (a non-YouTube Link Rem)	Opens the article
A plain Rem (no PDF/HTML source)	Nothing happens — you get a toast and default behavior is left untouched

So you can hover any pin and press the shortcut safely; only genuine sources open a viewer.

Scroll to Highlight

For PDF/HTML highlights, the viewer auto-scrolls to the highlighted passage once the reader finishes mounting (the embedded PDF engine takes a few seconds to initialize, so the scroll is retried a few times). If you then scroll around the document and want to jump back, click the 🔖 Scroll to Highlight button in the header to re-center on the highlight at any time.

Floating window — interaction & closing

The floating variant is designed so the source sits beside your card while you study, which is why it behaves a little differently from the modal:

• Stays open while you use the PDF. Unlike a modal, clicking into the reader to highlight, select text, or click existing highlights will not dismiss it.
• Closes itself when you advance the card, so a previous card's source never lingers over the next one.
• Esc closes it — without closing the queue. The plugin "steals" the Esc key while the float is open, so RemNote's queue doesn't act on it; Esc closes the float instead. (When focus is inside the PDF itself, the browser handles Esc within the reader; use the ✕ button there.)
• Not user-resizable. RemNote floating widgets have a fixed registered size; the window opens at ≈48% of the screen width on the right.

Recommended use cases

• Mid-review context check. While rating a flashcard that references a PDF highlight, open the source to re-read the paragraph it came from — then close and rate, without losing your queue position. (Modal is ideal for a quick look; floating if you need to glance repeatedly.)
• "What was I trying to recall?" Keep the floating window open beside the card so you can look from the source back to the card and forth, without the close/reopen churn a modal forces.
• Reviewing extracts and clozes that carry a bridging pin back to the original PDF source — see the figure or surrounding text without navigating into the document.
• Any pin to a long source where the inline hover-preview is too small but opening the full document in a pane is too disruptive.

Implemented as the Source Popup widgets — modal and floating — triggered by the Open Hovered Source in Popup / Floating Window commands. The default shortcuts Opt+O and Opt+Shift+O can be rebound in RemNote's keyboard settings.

---

Filter a Document by Images

RemNote's search indexes text. An image carries no searchable token, so neither Ctrl+F, nor the query language, nor a Search Portal can answer "show me the figures in this chapter" — the Filters section of RemNote's document search only lists tags, and an image is not one.

Tag Rems With Images (quick: img) closes that gap. It scans a scope for images and marks every Rem holding one with the HasImage tag — which the native document filter can isolate.

How to use it

1. Put your cursor in the Rem you want to scan — or simply open the document — and run Tag Rems With Images from the Omnibar (Cmd+/).

2. The Image Scan popup opens with two scopes to choose from:

   • Scan this Rem and its descendants — the button names the exact Rem, so you can be sure of the target before anything is written. The scope is the focused Rem when your cursor is in one, and the open document otherwise. (With neither, this button is disabled.)
   • Scan the whole knowledge base — every Rem, every document. The first such run is slow in proportion to how many images it finds (how long it takes), so reach for it when you want the tag applied everywhere once, and use the scoped run for day-to-day work.

   The popup is fully keyboard-driven: ↑/↓ move between the two scopes, Enter runs the selected one, Esc cancels. (Esc is ignored while a scan is running, so a reflex press can't abort a long run.)

   

3. Progress is reported live while it runs. Keep the popup open until it finishes — the scan runs inside it, so closing it stops the walk. Nothing is corrupted if you do: whatever was already tagged stays correct, and running the command again picks the work up.

4. When it finishes, the same popup reports the work done — Rems scanned, how many hold an image, how many were newly tagged, how many had the tag cleared — and repeats the two ways to use it. Scan again goes back to the scope choice; ? in the header opens this page.

Seeing the result

Filter one document. Open it, press Cmd/Ctrl+Shift+F (or Cmd/Ctrl+F and switch the search mode to Filter), then pick HasImage. The document collapses to just the Rems that carry an image. The count next to each filter tells you how many Rems it would leave — 3 · ⚡ HasImage below.

Collect them anywhere. A Search Portal on the HasImage tag gathers every tagged Rem into one place — useful for building a figure index across documents, and combinable with another tag or a document in the query to narrow it down. This is the whole-KB scan's payoff: with the tag applied everywhere, one portal is a live index of every image in the knowledge base.

What counts as an image

Any image element in a Rem's front text or back text — pasted, dragged, added with /image, or extracted from a PDF. Image-occlusion Rems count too, since the occlusion is drawn on an ordinary image element. An image sitting only on the back of a flashcard is found, which a purely visual scan of the outline would miss.

Re-running it

The command is idempotent and self-correcting. On every run it also removes the tag from Rems inside the scope that carry it but no longer hold an image — so deleting a figure and re-running leaves no stale mark behind. Rems outside the scanned scope are never touched, so a scoped run cannot disturb tags applied in other documents (only a whole-KB run reaches them).

Only Rems whose state actually changes are written to, which is what makes a re-scan of a large document cheap.

How long it takes — the first whole-KB run is slow

Finding the images is fast. Applying the tags is not. Reading every Rem in a large knowledge base takes seconds; writing a tag costs a round trip to RemNote, and those happen one at a time. So the cost of a run is set almost entirely by how many tags it has to write, not by how many Rems it looks at.

Measured on a knowledge base of 413,000 Rems holding 23,000 images:

Run	Writes	Time
Reading every Rem, before any tagging	—	~10 seconds
First whole-KB run, tagging everything it finds	23,000	~30 minutes
Every whole-KB run after that	~0	~10 seconds
One document (12,000 Rems, 1,200 images)	1,200	~90 seconds

So the whole-KB scan is a one-time cost, and only on a knowledge base of that size — the write cost per tag also grows with the knowledge base, so a smaller one is disproportionately quicker. After the first pass there is nothing left to write, and a re-scan only has to write the handful of Rems whose images changed since. Day-to-day, run it on a document and it finishes while you watch.

If you would rather not sit through the first pass, run it per document as you go; the tag accumulates, and a whole-KB run afterwards finds most of the work already done.

Leave it running

The scan lives inside the popup, so keep it open. If you do close it, nothing breaks — the tags already written stay correct, and running the command again picks up where it left off.

Clearing the tag

Remove HasImage Tags (quick: rmimg) takes the tag off every Rem that carries it, in the focused Rem's subtree or across the whole knowledge base. It exists because a whole-KB scan can mark tens of thousands of Rems and RemNote offers no way to take a tag off in bulk.

Nothing is lost. The tag is derived from the images themselves, so Tag Rems With Images rebuilds it exactly — the same relationship Remove All Priority Band Tags has to Refresh Priority Badges. That is why there is no undo: re-running the scan is the undo.

The scopes and keys match the scan's, and it costs the same per tag — clearing 23,000 tags takes about as long as applying them, so prefer the document scope unless you really do want the tag gone everywhere.

The tag is invisible in the outline

HasImage is bookkeeping for the filter, not something to read. Its chip is hidden from the editor tag bar — precisely, by targeting that pill alone, so your own tags on the same Rem stay visible. You will still see HasImage where it matters: in the document's Filter list.

Pins that lead to an image are ringed

A pin is drawn with a hairline ring saying where it leads, so you can pick the right one without following any of them — useful in an extract, where the plugin leaves both a reference to the parent it came from and a pin to the PDF highlight:

Ring	The pin leads to
Blue	a Rem holding an image — a figure in your own notes
Yellow	a text highlight in a PDF or web article — the source passage
Yellow + blue	a PDF area highlight — a clipped figure from the source
none	an ordinary Rem

Yellow means "this leads into a source document" and blue means "you will land on an image", so an area highlight — which is both — carries both colours, one per edge. The rings brighten when you hover one or edit the Rem.

Off by default. Turn on Enable Pin Reference Colour Rings (Plugin Settings → Editor Indicators) to get them. Two of the three states depend on tags this scan writes, so the feature is opt-in rather than something a knowledge base wakes up wearing. With the setting off, pins are left completely unmarked — including the priority-band border the highlight styling would otherwise draw on them. Every colour the plugin uses is catalogued in the Colour Coding Reference.

This is the second thing the image scan buys you, and it needs the scan to have run: the ring keys on the tag, so it appears on a pin only once its target has been tagged, and disappears when a re-scan clears the tag from a Rem whose image is gone.

The pin next to it in that screenshot shows the two markers are independent: the orange dotted box is the priority band of the linked highlight, the blue ring is "leads to an image". A pin can carry both.

The ring shares its box with the priority band. Pins to highlights also carry the priority band marker of the linked highlight, which sets the bottom and right edges. Rather than avoid that, the ring takes the top and left — so an extracted highlight's pin shows one box, half band colour and half ring colour, instead of two nested ones. On an area highlight the top edge is yellow and the left is blue for exactly this reason: those are the two edges that survive when a band is present. All three rings are solid, keeping them clear of the band marker's dotted and dashed vocabulary.

The blue ring is drawn in RemNote's accent colour — the same one the app uses for links and selection, which reads correctly on something that is a navigation target. It was neutral grey at first, on the theory that hue is already spoken for in this plugin (priority bands own the red→green ramp, #pdfextract is blue, #incremental green); in practice a grey hairline around an 18px icon in running text was invisible until hovered, which is no marker at all. The accent can't be confused with any of those, because this ring is an outline on an icon — never a background fill, never a left border — and it appears on nothing but pins. There is no fill either, so a pin sitting inside a highlighted extract never paints over the highlight's own colour, and both colours come from RemNote's own border tokens, so the ring follows light and dark mode.

Why an area highlight is worth its own state. When you clip a region of a PDF instead of selecting its text, RemNote stores the image on the highlight Rem in place of the text — an area highlight. It is a highlight and an image, which is why it wears both colours rather than a third one of its own.

What identifies one is that the Rem holds an image and nothing else — no caption, no prose. That matters, because adding a figure to an ordinary text highlight is common, and such a Rem holds an image and is a highlight, yet is not an area highlight; it keeps the blue ring. The scan decides this and records it as PdfAreaHighlight, since "image and nothing else" is a property of the Rem's text that no filter or stylesheet can inspect on its own.

The two tags are mutually exclusive. An area highlight carries PdfAreaHighlight and not HasImage; every other Rem holding an image carries HasImage. Tagging area highlights with both would make HasImage a complete list of every figure, which is tidier to filter — but RemNote collapses two or more tags into a "N tags" chip that no rule can hide without also hiding your own tags, and that clutter is not worth a filter you would rarely run on a highlights document. So: filter HasImage for figures in your notes, PdfAreaHighlight for clippings from a source. PdfAreaHighlight's chip is hidden from the tag bar exactly like HasImage's.

Two consequences worth knowing: caption an area highlight and the next scan swaps its tags — PdfAreaHighlight off, HasImage on — and the ring turns blue; and a Rem holding only an image that did not come from a PDF, a figure pasted into your own notes, gets HasImage and stays blue, since the yellow means "this leads into the source document".

Only image pins. Pins to ordinary Rems are left alone. Ringing every pin was tried and says nothing — a marker that appears on all of them carries no information.