Your library doesn't update itself when your content changes. A file is indexed as uploaded, and a website source is a snapshot from the day it was crawled. This guide covers the three habits that keep a Clone answering from current facts: replacing uploaded files, re-syncing web sources, and confirming every change in the Activity feed.
It depends on the source type. Static files — documents, images, audio, video and other uploads — are replaced. Re-syncable sources — websites, YouTube channels and other Web & Media sources — are refreshed from the Synchronization sub-tab without re-adding them.
Updating an uploaded file
There's no in-place edit for a static file: what you uploaded is what's indexed. Updating one is a quick replace.
Delete the old source
In the library's Sources sub-tab, find the outdated source and delete it.
NoteIf that source is also used in other libraries, deleting it offers two choices. Remove from this library takes it out here while it stays available in the others — the right pick when you only want to refresh this library. Delete sends it to the Recycle Bin and removes it from every library at once (restorable). A non-shared source just deletes to the Recycle Bin.
Upload the new version
Add the new file the same way you originally added it. It becomes a fresh source and starts processing.
NoteDuring the handover you'll see two rows — the old source on its way out and the new one processing. That's normal, and it resolves itself once both finish.
Wait for it to reach Ready
Watch the new source until it reaches Ready — the Activity sub-tab tracks its progress (hit Refresh to update the feed). From its next reply onwards, the Clone retrieves from the new version.
Keeping website sources fresh
A website source is a snapshot, not a live mirror. The crawl captures the site at that moment, and your Clone reads from the captured copy — update the live site five minutes later and the Clone won't notice. You have three ways to bring a snapshot up to date:
- Scheduled re-sync — set a cadence on the Synchronization sub-tab and the platform re-fetches the site automatically. The next section walks through it.
- Manual re-sync — trigger a one-off re-sync from the same sub-tab when you've just published a change you want reflected straight away.
- Replace the source — delete the website source and re-add it. A fresh crawl from scratch is sometimes cleaner when the site's structure has changed significantly.
While a re-sync crawls, the source row shows a live status such as Syncing: 22 found, 140 reviewed — the reviewed count keeps climbing even when found plateaus, so a rising reviewed number means the sync is still working.
Pages that are no longer reachable from the start URL are moved to the Recycle Bin on the next sync — restorable as one set — and the run reports them on the Activity sub-tab as Removed (N) — no longer on the site, with a Recoverable from the Recycle Bin hint. Content that has moved behind a login simply stops appearing. If a specific page must survive regardless, ingest it separately as a file source — export it to PDF and upload that.
Scheduling automatic syncs
The Synchronization sub-tab on the Knowledge page controls re-fetching for every syncable collection in your account. It has two halves: Global defaults on top, Per-collection overrides below.
Open the Synchronization sub-tab
From the left sidebar, open Knowledge and switch from Sources to Synchronization.
Switch on auto-sync
Enable the Auto-sync enabled master toggle. When it's off, no collection auto-syncs regardless of its own setting — it's the kill-switch.
Pick a default cadence
Set how often syncable collections re-fetch — every N day(s), week(s) or month(s). This default applies to every collection that doesn't override it.
Override per collection where needed
Each row in Per-collection overrides has its own Auto-sync toggle and a Cadence that reads INHERITED until you set a custom value with the edit pencil, plus Last sync and Next sync timestamps.
TipTo pause a single collection without touching the rest, switch its Auto-sync off.
Global defaults also include Limits — cost guards that stop an automatic sync from running up your bill. Anything that exceeds them is skipped and surfaced in the Activity log, so you can see what was filtered out:
| Limit | Default | What it protects against |
|---|---|---|
| Skip videos longer than [N] minutes | 60 minutes | Transcription costs on auto-discovered videos |
| Skip audio longer than [N] minutes | 90 minutes | The same, with extra headroom for podcasts |
| Skip files larger than [N] MB | 500 MB | Accidental huge-file drops in synced folders |
| Max new items per sync | 50 items | A backlog discovery loading everything at once |
| Pause auto-sync if balance is below [USD N] | USD 10 | Sync costs draining a low balance — resumes once you top up |
Confirming changes in Activity
The Activity sub-tab is the audit trail — "Background jobs across all your libraries — uploads, crawls, syncs, retries", as the page itself puts it. Each row is one processing event, with Collection, Phase, Status and When columns; the phases you'll see include Discovery, Loading, Sync and Upload to AI.
| Filter | Options |
|---|---|
| Status | All · In progress · Completed · Failed · Cancelled |
| Type | All · Discovery · Loading · Sync |
| Range | Last 7 days by default; other ranges in the dropdown |
| Collection | Scope the feed to a single collection |
Two buttons round it out: Refresh reloads the feed while you wait on an in-progress job, and Open details drawer gives a closer view of a selected entry. Items skipped by the sync limits also surface here, so a quieter-than-expected sync run has an explanation.
When a Clone seems to be missing a fact you know you added, look here first — confirm the job completed before debugging anything else.
If you try to leave the Knowledge page while a review, upload, source-edit, or connector-setup popup is open, you'll be asked to confirm — your staged files and reviewed selections aren't lost unless you choose to leave.
Frequently asked
That's not documented yet — whether one file from a previously imported archive can be re-uploaded on its own is still being confirmed. The safe path: upload the updated file as a separate source, and if the old version still exists as its own row from the import, delete that row.
They're skipped during that auto-sync run rather than processed, and surfaced in the Activity log so you can see what was filtered out. If something you wanted was skipped, raise the relevant limit on the Synchronization sub-tab so the next sync can take it.
The deletion may not have fully propagated through the index yet. Wait a few minutes and re-test; if the old content persists past roughly fifteen minutes, the source may be stuck — contact support with the source ID.
Open Activity first — a failed run shows its reason there. If the run completed, confirm the Clone is attached to this library, then check its match-strictness setting: Strict or Exact can miss new passages even when they're indexed.
Last updated July 6, 2026 · Knowledge Library
