Training & sync
Knowledge limits, how retraining works, the difference between stored and re-fetched sources, and sync status.
Once a source is added, FluentChat indexes ("trains on") its content so the agent can search it. This page covers the limits and how to keep knowledge fresh.
Knowledge allowance
Each agent has its own knowledge allowance, measured as extracted text:
Prop
Type
A usage bar on the Data sources page shows how much you've used, and warns you as you approach the limit. Remember that files count their extracted text, not the original file size — see Files.
Retraining
Click Retrain on a source to re-process it. What that does depends on the source type:
Prop
Type
Retraining is rate-limited (repeated clicks within a few seconds collapse into one job, and there's a per-workspace ceiling per minute). If you hit the limit, wait a moment and try again.
Sync status
Sources show where they are:
- Pending — queued to index.
- Syncing — being processed.
- Ready — indexed and searchable.
- Failed — something went wrong; the error is shown so you can fix and retry.
Plan changes
If your plan changes so a source's feature is no longer included (for example web crawl), your existing content isn't deleted — it stays searchable — but re-syncing that source is paused until the feature is available again.