Managing Collections and Storage
Monitor your storage usage and manage your knowledge collections.
Overview
Every document you upload counts against the knowledge storage allowance on your plan. Managing collections is the day-to-day work of keeping the Knowledge Base accurate, lean, and within those limits: pruning outdated files, removing collections that no longer have a purpose, and splitting collections that have grown too broad to query well.
Collections and the documents inside them are workspace-scoped, so changes here affect every workflow in the workspace that queries them. Treat deletes as you would a deploy: confirm what depends on the content before you remove it.
Checking Storage Usage
Open Settings -> Usage to see how much knowledge storage you have consumed against the limit on your current plan. Image-heavy PDFs and large spreadsheets consume far more than plain text, so usage is not always proportional to file count.
Managing Documents
- View documents - Open the Knowledge page and click a collection to see every document in it, along with its status.
- Delete a document - Use the row action on any document to remove it. The embeddings are dropped at the same time, and the document no longer appears in query results.
- Replace a document - Delete the old version, then upload the new one. Spojit does not overwrite in place.
Managing Collections
- Delete a collection - Removes the collection and every document inside it. Any workflow node still pointing at the collection will fail at runtime.
- Split a collection - Create a new, more focused collection and upload the relevant subset there, then delete the originals from the broad collection.
- Audit before deletion - Search your workflows for the collection name to see which Knowledge nodes reference it before you remove it.
Tips
- Schedule a periodic review (quarterly works well) to delete stale documents and reclaim storage.
- Keep one document per logical source. Combining ten policies into one mega-PDF makes query results harder to attribute.
- Prefer narrow, well-named collections over broad ones. They are easier to retire when the underlying topic changes.
Common Pitfalls
- Deleting a collection that production workflows still query. Check workflow references first and update the affected nodes.
- Forgetting to delete the old version of a replaced document. Both versions remain queryable and dilute relevance.
- Hitting the storage limit silently. New uploads can fail or be rejected, so monitor Settings -> Usage before bulk ingestion.
- Assuming deletion is reversible. Deleting a document drops its embeddings; you must re-upload to bring it back.