Training, refresh and failures
Read processing states, retry failed knowledge and understand auto-sync.
Read the processing state
| State | Meaning | Next action |
|---|---|---|
| Queued | Accepted for processing; the answer index is not ready yet. | Wait and refresh the knowledge list. |
| Processing | Content is being extracted and indexed. | Allow processing to finish before testing new content. |
| Ready | The document has been indexed. | Test a specific fact and inspect the cited source. |
| Failed | Extraction or indexing could not finish. | Read the displayed error, correct the source and retry. |
Manual refresh and auto-sync
Use the document refresh/sync action when a public website page changes and you need a new crawl. Growth website auto-sync is monthly; Scale is weekly. Contract settings can differ. A scheduled crawl is not a promise that content published tonight will be indexed tomorrow.
- A login-protected page, blocked crawler or inaccessible URL may not provide readable content.
- Files and imported Notion/Drive text are snapshots; replace/re-import changed content.
- Do not repeatedly queue the same source while its existing job is running.
- Quota admission can prevent additional knowledge from being queued; check Usage and the selected bot.
Verify an update
- Open the exact source and confirm the new fact is present.
- Refresh or re-import the correct knowledge document.
- Wait for Ready and check the last synced/processing information.
- Ask a narrow question in a fresh Playground conversation.
- If the answer is stale, check for another document containing the old fact and correct that source too.