Store structured research rows in Dataset items and create typed Brain schemas for briefs, campaigns, products, approvals, or any repeatable knowledge object.
Where: Flowgen → Brain → item types / New item
Key ideas
- Dataset item: The built-in Dataset type combines a human-readable markdown summary with a native rows field, so research collectors and analysts can pass structured evidence without flattening it into prose.
- Custom schema: Custom item types define named typed fields. Built-in schemas can also be scoped and overridden while retaining their stable preset identity.
- Safe fallback: If a custom type definition is later removed, existing items still resolve through the generic Document body instead of becoming unreadable.
Steps
- Choose Dataset for harvested research, or create a custom item type for a business object you expect to reuse.
- For a custom type, give it a clear singular name and add only stable fields with the correct text, image, design-system, dataset, or other supported type.
- Create representative items and verify that fields remain useful with missing, long, and real values.
- Read one item with Brain Read, search across many with Knowledge Search, or bind an intentional Brain Write target.
- Review item version history before changing a live schema or overwriting a record used by published flows.
Dataset workflow
- A collector or analyst emits a typed dataset payload.
- A Dataset item keeps rows in its data field and an explanation in its summary/body.
- Knowledge Search can retrieve matching record context; dataset-aware analysis nodes can consume the structured rows directly.
Schema design checklist
- Use stable machine keys and user-facing labels.
- Choose typed fields instead of encoding a whole object in one text block.
- Keep the body field when a portable narrative representation is valuable.
- Avoid changing a field type after many items or published flows depend on it.
Tips
- Use custom types for repeated structures, not one-off notes.
- Keep source/provenance and date fields on research items when freshness matters.
Limitations and important notes
- Deleting a type does not migrate its typed field data into a new schema automatically.
- Knowledge Search retrieves the user’s Brain records; it is not a web-search node.