'/%3e%3cpath%20id='cross_2'%20data-name='cross%202'%20class='cls-2'%20d='M20.612,30.43h0a5.361,5.361,0,1,0-10.723,0H0A33.709,33.709,0,0,1,30.569,0V9.75a5.361,5.361,0,1,1,0,10.723V30.43Z'%20transform='translate(67.193%200.007)%20rotate(90)'/%3e%3cpath%20id='cross_3'%20data-name='cross%203'%20class='cls-3'%20d='M20.612,30.43h0a5.361,5.361,0,1,0-10.723,0H0A33.709,33.709,0,0,1,30.569,0V9.75a5.361,5.361,0,1,1,0,10.723V30.43Z'%20transform='translate(67.186%2067.041)%20rotate(180)'/%3e%3cpath%20id='cross_4'%20data-name='cross%204'%20class='cls-4'%20d='M20.6,30.5h0a5.361,5.361,0,1,0-10.723,0H0A33.71,33.71,0,0,1,30.56,0V9.818a5.361,5.361,0,1,1,0,10.723V30.5Z'%20transform='translate(0%2067.041)%20rotate(-90)'/%3e%3c/svg%3e)
Operating Knowledge Storage
Knowledge indexing stores searchable vectors in the selected vector store. These vectors are derived data: changing or resetting the store does not change the source files in the Workspace.
Before you begin
- Required role: Workspace administrator
- Check the current driver and connection state under
Knowledge Services>Knowledge Storage. - Schedule migrations and resets for a suitable time because vector results may be incomplete while reindexing runs.
Choose a vector store
| Driver | Storage location | Collection preparation | Typical use |
|---|---|---|---|
| Automatic (SQLite) | SpaceBuilder internal storage | None | Default embedded storage |
| Local (SQLite) | local_vectors.db under a configured path | None | Smaller installations that need a specific local path |
| Qdrant | External Qdrant service | Created on first indexing | Separate service or larger vector sets |
| Milvus | External Milvus service | Create it before connecting | Existing Milvus deployments and large clusters |
Vector count means document chunks, not documents. A long document produces multiple vectors. Configure an embedding model even when the storage connection itself is healthy.
Operate Local (SQLite)
Local storage keeps vectors in local_vectors.db. Relative paths are resolved from the SpaceBuilder configuration directory. The status page also shows the SQLite database, WAL, SHM, and reusable free space.
Change the store or embedding model
Saving any of the following changes starts a full reindex from the source content:
- driver, endpoint, collection, or local data path
- embedding model or vector dimension
The migration proceeds as follows:
- Save the new configuration.
- The indexing scheduler detects the change and queues a full reindex.
- SpaceBuilder embeds every source document into the new store.
- After indexing and search validation finish, it removes the old vectors from their original store.
Keep the previous external store reachable until cleanup finishes. When changing vector dimensions, select a new compatible collection; an existing Qdrant or Milvus collection cannot accept vectors with a different dimension.
During migration, vector search may be paused briefly or return only content already written to the new store. Keyword search remains available.
Read operating state
Knowledge Storage shows connection state, driver, entity count, collection count, and collection compatibility.
| State | Meaning and action |
|---|---|
| Compatible dimension and distance | The collection is ready. |
| Not created; created on first indexing | A Qdrant collection does not exist yet. This is expected before the first write. |
| Collection missing | Create the Milvus collection before indexing. |
| Check collection settings | Dimension, distance metric, or vector layout does not match the SetFN configuration. |
Use Ingestion & Indexing to review pending updates, pending embeddings, embedding totals, and pending vector deletions. Cleanup retries use increasing delays up to five minutes when an old store is unavailable.
Reindex or reset
Both administrator actions preserve source files.
| Action | Result | When to use it |
|---|---|---|
| Reindex all data | Reads, chunks, and embeds every source again | Extraction rules changed or search data needs rebuilding |
| Reset vectors and reindex | Deletes this SpaceBuilder installation's vectors from the current store, then queues a full reindex | The derived store is inconsistent or must be rebuilt from zero |
Completion is shown on Ingestion & Indexing. A reset does not delete vectors owned by another installation that shares the collection.
Share an external service safely
Every SpaceBuilder installation writes an isolation key and limits search, statistics, and reset operations to its own data. Separate collections are still recommended because they simplify capacity management, backups, and deletion. The settings page warns when another local SpaceBuilder project uses the same collection.
Back up and restore knowledge state
Stop the SpaceBuilder server before running these commands. The backup target must be a new path. Restore is limited to a backup from the same installation and home directory.
~/SetFnSpaceBuilder admin knowledge-backup <backup-directory>
~/SetFnSpaceBuilder admin knowledge-restore <backup-directory>
The backup includes team and personal indexing records. A local local_vectors.db is included only when it is under the SpaceBuilder home or configuration directory. Qdrant and Milvus vectors are excluded because they can be rebuilt from source files.
After restoring, start the server and run Reindex all data in Knowledge Storage.
Troubleshoot storage
| Symptom | What to check |
|---|---|
| Disconnected | Endpoint, port, firewall, service status, and credentials |
| Dimension mismatch | Configure the embedding dimension used by the collection, or use a new collection |
| Unnamed dense vector error | Use a single unnamed dense vector in Qdrant instead of a named or hybrid layout |
| Entity count does not increase | Indexing enabled, allowed schedule, required models, and pending errors |
| Cleanup retries continue | Restore connectivity to the previous store so queued deletion can finish |