The bottom of the Agent's Knowledge Sources page contains Existing sources.
This list is where you review each Source's processing state, sync information, page progress, errors, and maintenance actions.
Use Search sources… to locate a Source when the Agent has many entries.
The Sources list is paginated so large Knowledge collections can be reviewed without loading every Source at once.
The deployed interface recognizes these primary Source statuses:
Pending means the Source has been created or requeued but processing has not yet completed.
Crawling indicates that website content is currently being fetched or processed through the crawl pipeline.
Allow background processing to complete before concluding that the Source has failed.
Indexed indicates successful Source processing. For multi-page Sources, also review the number of pages indexed compared with the total expected page count.
Failed indicates that the Source could not complete processing. The Sources list can display a human-readable error next to the Source.
Use that message to decide whether to correct the Source or try Reindex. If the message indicates an external connection problem, use your normal support process rather than changing Workspace connection settings.
Each Source can display:
Select the Preview icon for a Source to open What the AI sees for this source.
The preview reports the Source status, page indexing progress, and any error.
For extracted Documents, the preview can display:
Select View chunks within the Preview to inspect the actual extracted text.
Each shown Chunk includes its order and token count.
This is one of the most useful checks for determining whether the Agent received meaningful content.
If the Source has no extracted Documents yet, the interface states:
No content extracted yet. Try Reindex, or check the error message above.
The Source Preview is useful for checking one ingestion Source. The separate What the AI knows page provides an Agent-wide view of indexed Documents and Chunks across all Sources.