Skip to main content
Linkly AI can pull up a document on your machine in a fraction of a second because of one thing it quietly does in the background: indexing. This page walks through every switch on the Index management page in the main window — what it controls, what it defaults to — and gives you a path to follow when indexing misbehaves. If you’ve just installed the app and haven’t added your own documents yet, start with the Quick Start.

​
How Indexing Works

Indexing in Linkly AI is layered — it isn’t a case of nothing working until everything finishes. Each layer that completes adds another way to search.
1

Layer 1 — filenames (seconds)

Paths and filenames are written to the index the moment a file is discovered. This layer is close to instantaneous, which is why you can search by filename right after adding a folder.
2

Layer 2 — full text (keywords)

Document bodies are parsed into a local full-text index that supports keyword matching. Once this layer is done, searching for “budget sheet” matches documents whose body contains those words.
3

Layer 3 — semantic (vectors)

An embedding model splits the text into chunks and turns them into vectors. Once this layer is done, searching for “how much did we spend last year” can surface a finance report whose title contains none of those words.
On top of those three layers sit a few optional extraction enhancements, whose job is to turn files that contain no text into files that do, and then feed the result into the three layers above:
  • Text recognition (OCR): reads text out of images and scans
  • Audio transcription and video transcription: turn spoken content in audio and video into text

​
Opening the Index Management Page

Any of these takes you to the Index management page:
  • Click the index status icon at the bottom of the main window’s left sidebar. In the Index status card that pops up on hover, Show more gets you there too
  • On the main window’s Home, in the My knowledge assets area, click the card that shows the index status (for example Search is ready or Processing files)
  • Right-click the L icon in the system tray and choose View Index
The page has four tabs:

​
Reading the Status Area

The top of the Index tab holds four cards:

​
Management: Search Capabilities and Extraction Enhancements

The Management area is split into two groups. Each row shows that capability’s status and doubles as its switch: The status tag on the right of each row can be one of: While OCR or transcription is working on a file, the row also shows the filename and progress (pages or duration). What it does: when on, lets you find documents using natural, descriptive language. Default: on With it off, search uses keywords only. Turning it back on backfills the semantic index automatically — no rebuild needed.

​
Text Recognition (OCR)

What it does: uses an OCR model to read text out of images, and to convert scanned PDFs into structured Markdown. Default: on With it off, images and scans can only be found by filename; they show up under Needs OCR on the Can’t extract tab, and they’re processed automatically once you turn it back on.

​
Audio Transcription and Video Transcription

What they do: use a speech recognition model to turn spoken content in audio and video into text, then feed that text into the full-text and semantic indexes. Default: both off ⚠️
Off doesn’t mean unindexed. Media files are still registered in the index — you can find them by filename, but not by what’s said inside them.
When to touch it: turn them on when you have meeting recordings, podcasts, or course videos you want to search.

​
Policy: Resource Usage

The three items in the Policy area only affect how indexing runs, not what gets indexed.

​
1. Performance Mode

What it does: controls how much compute background indexing is allowed to take. Default: Auto

​
2. Power Saving

What it does: pauses the power-hungry heavy work while a laptop is running on battery. Default: off
It only takes effect when the toggle is on and the machine is actually running on battery — both at once.
Paused while on battery:
  • Local OCR, including converting scanned PDFs to structured Markdown
  • Local audio and video transcription
  • Semantic indexing (vector generation)
  • Cloud library auto sync
  • Database housekeeping
Unaffected, still running: filename indexing and full-text (keyword) indexing. So even in Power Saving, new files remain searchable. When to touch it: laptop users should just turn it on and forget about it. On a desktop it makes no difference either way.

​
3. Sleep Models When Idle

What it does: releases the memory held by local AI models after a period of inactivity, and reloads them automatically when they’re needed again. Default: on Local models (semantic embedding, OCR, speech recognition) are unloaded after roughly 300 seconds (5 minutes) of inactivity to free memory. With it off, the models stay resident — around 1 GB of memory for all three combined. When to touch it: if memory is tight (8 GB or less), leave it on.

​
Rules: Which Files Get Indexed

Your folders often hold more than what you want AI to read: thousands of screenshots, gigabytes of screen recordings, exported intermediate files. The Rules tab keeps them out of the index. It has three parts.

​
File Types

Switch types on and off, in two groups: All are on by default. When you switch a type off, files of that type are removed from the index within about a minute, and search, the library and agents no longer see them; files on disk are not touched. If you switch it back on later, those files are scanned again, and their text may need to be recognized or transcribed again.

​
Exclusion Rules

Leave out some files or subfolders with wildcard rules, written the same way as .gitignore: Rules work at two levels: rules under All folders apply to every folder, and each folder can also have rules of its own. While you edit, the files a rule will leave out are listed live below it, and mistakes get a hint. The handier way in is from the library: right-click a file or folder and choose Add exclusion rule… to create a rule for it. In the folder tree on the left of the library, a folder with its own exclusion rules shows a funnel icon at the end of its row; right-click it and choose Edit exclusion rules… to change them.
Exclusion rules only take relative paths inside a folder. To stop indexing a whole folder, remove it under Manage sources in the library.

​
Built-in Rules

These always apply and can’t be turned off:
  • Hidden files and folders (names starting with ., such as .git and .obsidian)
  • Folders created by the system and sync apps
  • Linkly AI’s own app data
  • Temporary files (the ~$ files Office creates while a document is open, and Thumbs.db)
  • Files ignored by .gitignore in a Git repository, and whatever .ignore files in your folders list
  • Unsupported formats (archives, installers, code files and the like)
Notes, clips and canvases in the Linkly AI Space are always indexed, regardless of type switches and exclusion rules.

​
The Can’t Extract Tab

This tab lists files that can be found by filename, but whose content didn’t make it into the index. They can’t be found by full-text or semantic search, and agents can’t read their text. On the left, files are grouped by reason, with a count for each group; on the right is the file list for the selected group. Most groups offer Retry all, and you can also retry files one at a time to re-extract them and update the index. Files in the Unreadable group are also retried automatically once their content changes. Three groups have no retry: No text (extraction found no text in the file), plus Needs OCR and Transcription off. The latter two are processed automatically once you turn on the matching switch.

​
The Advanced Tab

​
Hardware Acceleration

What it does: runs local models on the GPU. With it off, everything runs on the CPU — slower but steadier. Default: on
  • Windows / Linux: the embedding model for semantic indexing uses the GPU when there’s a discrete GPU; on Windows with a discrete GPU, the layout model for scanned-document OCR uses the GPU too
  • Apple silicon Macs: it only controls whether the layout model for scanned-document OCR uses the Apple GPU
  • On Windows / Linux, if semantic indexing has switched to the cloud model, OCR can’t use the GPU either, so the switch has no effect and shows as unavailable
  • Intel Macs don’t have this switch
Which tasks use the GPU, and whether integrated graphics are involved, is covered in “Does the embedding model or OCR use integrated or discrete graphics?” in the FAQ.

​
Index Data

  • Location: the folder holding the index data; click the button on the right to open it in Finder / File Explorer
  • Disk usage: the combined disk space taken by the full-text index, vectors, and caches

​
Index Storage Optimization

Click Optimize to compact the vector index, reclaim the space held by deleted documents, and clear out extraction caches left behind by older versions. It doesn’t affect search results; disk usage drops noticeably, and search gets faster too. Quitting the app partway through is safe — nothing is corrupted, and you can continue optimizing after the next launch.
If the index is already in the optimal state, the button does nothing — it just reports Your index is already optimized, and there is no way to force a re-run.

​
Rebuild the Index

This is a destructive operation: the Rebuild… button is red, and you have to confirm once more after clicking it. It will:
  • Delete: the existing index data and the document content already processed (including OCR and audio/video transcription results) — this can’t be undone
  • Keep: your original files
Linkly AI then re-processes every document from the original files, redoing all OCR and transcription. With lots of documents this can take anywhere from minutes to hours. The rebuild runs in the background, and meanwhile search results are temporarily empty or incomplete (they don’t error).
In these three situations the button is refused with a message:
  • Indexing is paused (choose Resume indexing first)
  • A scan or processing is currently in progress
  • A rebuild is already running

​
Index Status

The first card in the status area, the index icon at the bottom of the sidebar, and the bottom of the Library’s left column all show the same set of statuses: Watching is the end state you want to see.

​
How Do I Pause Indexing?

There are two places to pause it:
  • The Index management page: the Pause button on the first card of the status area. Once paused, the button becomes Resume indexing; click it to resume
  • The system tray menu: right-click the L icon in the tray, and the menu has ⏸ Pause Indexing / ▶ Start Indexing (left-clicking opens the main window, not the menu)
While paused, indexing stops processing files, but the index already built stays searchable as usual.

​
FAQ

Because audio transcription and video transcription are off by default. While they’re off, media files are still registered in the index (hence the filename match), but nothing spoken inside them has been transcribed.Turn on Audio transcription and Video transcription in the Management area of the Index management page. The app downloads a 161 MB speech recognition model, then works through your existing files automatically — no rescan and no rebuild required.One exception: if a media file sits next to a matching .srt / .vtt subtitle file, or was transcribed earlier and the cache is still around, its content is searchable even with the toggles off.
First check whether Text recognition (OCR) is switched off in the Management area of the Index management page. This one switch governs text extraction from both images and scanned PDFs — turn it off and scanned PDFs become filename-only, with no body text.If the switch is on, look at that row’s status: Processing with N left means files are still queued or being recognized, so let it finish.
That means indexing has been paused. Click Resume indexing on the first card of the status area, or right-click the tray icon and click ▶ Start Indexing in the menu — either one resumes it.If it isn’t paused, double-check in the main window’s Library that you’ve actually added your own directories.
Work through these in order:
  1. Does the Semantic search row in the Management area of the Index management page read Off? If so, turn it on (it backfills automatically, no rebuild needed).
  2. Does it read Processing? Vectors are still being built; wait for it to finish.
  3. If it’s neither, go to Settings → AI Models → Embedding model and check whether the model is still downloading or failed to download — the search interface gives you no indication of this at all, and search silently degrades to keyword-only. If it failed, click Retry on the card.
Click Pause on the first card of the status area on the Index management page, or right-click the L icon in the tray → ⏸ Pause Indexing. Resuming happens in the same place: the button becomes Resume indexing, and the tray menu item reads ▶ Start Indexing.Rebuild the index can’t run while indexing is paused — resume first.
Go to Index management → Advanced → Maintenance and click Optimize in the Index storage optimization row. It compacts the vector index and reclaims the space held by deleted documents, without affecting search results.If it reports Your index is already optimized, there’s nothing left to reclaim, and a re-run can’t be forced.
Not yet. A rebuild is the last resort, and search results are temporarily empty or incomplete while it runs. Try this order instead:
  1. Confirm the relevant switches are on (OCR / Semantic search / Audio transcription / Video transcription)
  2. Confirm indexing isn’t paused
  3. Confirm the status has actually reached Watching rather than still sitting at Processing
  4. Check whether the file is on the Can’t extract tab, and retry it if retry is offered
Only once all of these are ruled out is it worth going to Index management → Advanced → Maintenance → Rebuild the index. A rebuild doesn’t touch your original files, but it does delete the text already extracted (including OCR and transcription results), so every document has to be processed again — with lots of documents this can take hours.
No. Performance Mode only controls how many threads indexing uses and how readily it yields — not whether it runs at all. Choosing Balanced just makes it slower and less intrusive.To actually stop it, use Pause on the first card of the status area, or Pause Indexing in the tray menu.Also worth knowing: Auto looks at keyboard and mouse idleness (over 120 seconds counts as idle); if Power saving is also on and the machine is on battery, it stays Balanced even when idle. On a pure Wayland Linux desktop the idle signal isn’t available, so Auto behaves identically to Balanced.
That’s expected. Power Saving takes effect only when the toggle is on and the machine is running on battery. Plugged in — or on a desktop with no battery at all — it does nothing.Which is exactly why you can leave it on year-round, with no need to flip it on before you head out and off when you get home.
From the gentlest measure to the strongest:
  1. Set Performance Mode to Balanced (takes effect immediately; jobs already running pick up the new budget)
  2. Temporarily turn off Audio transcription and Video transcription — media transcription is the heaviest work there is
  3. Still sluggish? Click Pause to pause indexing, and resume when the machine is free

​
Further reading

  • Quick Start — install, add documents, connect your AI tools
  • Using the Launcher — the full search interface
  • Using Libraries — scoping retrieval by library
  • FAQ — how long indexing should take, and how to grab the application log
  • Using the Chatbot — index status also shows at the bottom of the sidebar
  • Linkly AI Space — the folder that’s indexed by default
  • Index Scope — switch off file types you don’t need, leave out files and folders with rules