Remove knowledge
Delete a source from this chatbot and get the space back. There is no undo, and cleanup is not instant.
Browse topics
What it is
Removing a source deletes it from this chatbot: the stored file, the extracted text, and the search index built from it. Once removal finishes, the chatbot can no longer answer from that material and the space it used comes back to your allowance.
Because a source belongs to exactly one chatbot, there is no "detach from this chatbot but keep it elsewhere". Removal is removal.
When you would use it
- The content is out of date and a wrong answer is worse than no answer.
- You need space under your knowledge cap.
- A crawl collected pages you never wanted — tag archives, pagination, an old blog.
- A document imported badly and you are replacing it with a better version.
- You uploaded to the wrong chatbot.
For the fourth case there is often a better tool: Replace file on the existing row swaps the document while keeping the title, the collection, and the row's place in the table. Use replace when it is the same document in a newer version, and remove when the material should simply be gone.
Where to find it
Open Dashboard → Chatbots → your chatbot → Knowledge.
Before you start
- Removal is not reversible. There is no undo and no recycle bin. If you might want the material back, keep a copy of the original file or text somewhere first.
- Check you are on the right chatbot. Two chatbots can hold sources with identical titles.
- If several sources cover the topic, decide which one is authoritative before you delete the others, so you do not accidentally remove the only good copy.
Steps
- Open Dashboard → Chatbots → your chatbot → Knowledge.
- Select the row, or tick several rows to remove them together.
- Choose Remove.
- Read the confirmation dialog. Where it shows an impact summary — how much space you get back, which chatbot is affected — read it rather than clicking through.
- Type the confirmation the dialog asks for and confirm.
- Wait. The row stays visible with a removing badge until cleanup finishes.
- Once the row disappears, refresh the usage meter to see the space returned.
- Ask Test Chatbot the question that used to hit that source, and confirm the chatbot no longer answers from it.
What you will see
Removal is a confirm dialog, not a one-click hide. After you confirm, the row does not vanish immediately — it stays with a removing badge while the search index is cleaned up.
That visible intermediate state is deliberate. Deleting from the index is background work, and hiding the row instantly would let you believe the material was gone while the chatbot could still find it. The badge tells you the truth: not gone yet.
If cleanup fails, the row stays on the list with an error you can retry, rather than disappearing and leaving orphaned material behind.
Why it is not instant
Extracted text lives in two places: the record you see on the table, and the search index the chatbot actually queries. Both have to be cleared, and clearing the index happens in the background alongside everything else in the queue.
Two consequences:
- Do not trust Test Chatbot until the badge clears. Between confirming and completion, the chatbot may still find the passage.
- Do not delete and re-add in a loop. Each cycle queues more work and pushes you further back. If a removal looks slow, wait, then retry once.
Removing a whole crawl
Deleting a crawl session from the crawl history is not the same as removing the pages it produced. The history entry is the record of the run; the pages are ordinary knowledge sources and stay until you remove them from the Knowledge table.
To undo a crawl completely: cancel or delete the session in the history drawer so it does not continue or rescan, then filter the Knowledge table and remove the rows it created. See Recrawl or stop a crawl.
Limits and plan notes
- You need a role that permits managing knowledge. A Viewer can see sources but cannot remove them. See Roles and permissions.
- Removing knowledge never requires you to pause or deactivate the chatbot, and the chatbot stays online throughout.
- If you remove every source, the chatbot stays active but has nothing to answer from. Visitors then see the not-ready experience described in The hosted page says the chatbot is not ready.
- Space returns to your allowance only when removal completes, not when you confirm.
- Conversations that already quoted the removed source keep their transcripts. Removing knowledge does not rewrite history.
Common problems
The row came back after I refreshed.
Removal was still running, or it failed. Look for a removing or failed badge and retry once.
I removed a file and the chatbot still cites it.
Either cleanup has not finished, or the same content also exists in another source — a pasted copy, or a crawled page carrying the same text. Search the table for the topic and check what else covers it.
Removal failed.
Retry from the row. A repeated failure on one specific source is worth reporting; note the title and roughly when it started.
The meter did not go down.
Wait for every removing badge to clear, then refresh. Multi-row removals report space back as each one completes.
I deleted the wrong source.
There is no undo. Add the material again from your original copy. If it was pasted text you no longer have, you will need to rewrite it — which is the argument for keeping originals somewhere outside Agentency.
I want to remove everything and start over.
Select all rows and remove them in one operation, wait for the badges to clear, then rebuild. Do not delete the chatbot itself unless you also want to lose its settings, conversations, and hosted link.
Common questions
Can I undo a removal?
No. There is no recycle bin. Keep a copy of the original file or text outside Agentency if there is any chance you will want the material back.
Should I remove or replace an outdated document?
Replace when it is a newer version of the same document — it keeps the title and collection. Remove when the material should simply be gone.
Why does the row stay visible after I confirm?
Cleanup of the search index runs in the background. The removing badge is honest about that: until it clears, the chatbot may still find the passage.
Does deleting a crawl from history remove the pages it collected?
No. The history entry is the record of the run. The pages are ordinary sources and stay on the Knowledge table until you remove them there.
Was this article helpful?
Related articles
Knowledge limits
Your cap counts extracted text, not file size. Free 400 KB through Agency 60 MB, per chatbot.
Dataset status and training
What each badge on the Knowledge table means, what the failure messages are telling you, and when to just wait.
Recrawl or stop a crawl
Cancel a run in flight, switch off a sitemap schedule, or refresh content — without ending up with two of every page.
Collections and ownership
Collections are folders for one chatbot's sources. Ownership is the rule that a source belongs to exactly one chatbot.
Ready to try it on your own content?
Create a free workspace, add a document, and ask the questions your team is tired of answering.