Skip to content
FluentBot Beta
This site is under construction. Content, screenshots, and workflows may change at any time.

Retrain & Update

Knowledge sources don’t stay current on their own. As your docs, help center, or tickets change, the bot needs to re-fetch and re-index that content. FluentBot supports three update paths:

  • Scheduled refresh — runs automatically per source on a cadence you pick.
  • Manual retry — re-fetch a failed source from the source dialog.
  • Manual replacement — replace uploaded files or pasted text when the source content changes.

Scheduled refresh

Open any Website Links source from Knowledge Base. The dialog has a Scheduled Refresh dropdown:

Source detail dialog with Scheduled Refresh dropdown open showing Never, Daily, Weekly, Monthly options
OptionBehavior
NeverNo automatic refresh. Source stays as-is until you retry manually.
DailyRe-fetch once per day.
WeeklyRe-fetch once per week.
MonthlyRe-fetch once per month.

Scheduled refresh runs automatically based on the cadence you choose.

What gets updated

When a refresh runs, FluentBot checks the source again and updates the indexed content when changes are found.

Manual retry

Click any source row in All Sources to open its dialog. If the source’s current status is Failed or Partially Indexed, a Retry button appears alongside the failure reason.

Source detail dialog showing Failed status with the Retry button and reason text

Clicking Retry starts the source again and resets it to the Queued state. The same retry path applies whether the source is a single URL, a sitemap, or a Crawl.

Inside the Documents list, each indexed document has its own actions menu (three dots). The only per-document action available there is Delete document — there’s no per-document retry; if a single document needs re-fetching, retry the whole source.

Source detail dialog for an Indexed source — Documents list with one row and per-document actions menu

Source statuses

A source moves through these states. Status badges in the All Sources table use the same labels:

StatusMeaning
QueuedWaiting to start processing.
ScrapingFetching content from the source.
IndexingSaving extracted documents and making them searchable.
IndexedReady to answer questions.
Partially IndexedSome items succeeded, some failed. Open the source and Retry to re-attempt the failed items.
FailedEvery item failed, or the source itself is unreachable. Open the source and Retry.
DeletingRemoval in progress.
DeletedRemoved from the bot.

Removing a source

Delete Source at the bottom-left of the source dialog removes the source completely:

  • Deletes every document under the source.
  • Removes the source from future answers.
  • Deletes uploaded files for File sources.

Deletion is permanent — there’s no undo. Re-adding the same URL afterwards is a fresh source with a new ID.

Choosing a refresh cadence

Pick the cadence based on how often the underlying content actually changes:

  • Marketing site, product brochures: Monthly is usually enough.
  • Help center, public docs: Weekly fits most teams.
  • Release notes, changelog, ticket-based knowledge: Daily.
  • Static reference (PDFs that never change): Files have no refresh — replace them manually after each revision.

More frequent refresh is best for content that changes often. For large sites that rarely change, weekly or monthly refresh is usually enough.

Troubleshooting

  • “Source keeps failing” — open the source, read the failure reason, fix the underlying issue (auth, blocked URL pattern, dead domain, JS-rendered content with no fallback), then click Retry.
  • “New content isn’t showing up in answers” — confirm the source is Indexed, then test the question in Playground.
  • Indexed but Documents = 0 — the page returned no extractable text. Try the source as a File upload if the underlying content is a PDF.
  • “Can’t find a Scheduled Refresh setting” — the source is a Text or File source. These don’t auto-refresh; replace the source manually if the content changes.

What’s next