Three canvas nodes work with records: Record (read) binds a record and emits its markdown; Write Record archives-then-overwrites a bound (or newly created) record with incoming content; Markdown Doc is an editable doc you can save to a record and export to PDF.
Where: flowGen → left rail → Add → Words & data — Add "Record (read)", "Write Record", or "Markdown Doc"
Key ideas
- Record (read): A bound reader. Pick a kind filter and a record on the card; it caches the markdown and emits it on "Record (markdown)". The refresh / "Read record" action re-reads the latest content.
- Write Record: Takes content on its "Content" input and, on run, archives the target record's previous content then overwrites it — or creates a new record (pick a kind, optional title, and org-scope) when nothing is bound. It passes the content through on its output.
- Markdown Doc: An editable markdown document on the canvas with inline and full-screen split-pane editing. Incoming content is staged and only applied when you click "Load from input", so your edits are never silently overwritten. Save persists to a bound generic record; "Export PDF" prints the rendered markdown.
Steps
- Add → Words & data → "Brain Read" (or a kind preset such as "Persona · Read"), then bind a record to emit its markdown.
- To update a record from a flow, add "Write Record", wire content into its Content input, bind a target record (or set a Create kind + title), and run — it archives then writes.
- To draft a document, add "Markdown Doc", type in the editor (or "Load from input" to pull staged content), then Save to a record or Export PDF.
- Choose or create a named record from the card, confirm its field rail, then connect record_in to an exposed mini-app question if the person running the flow should choose the record.
Typed entities and selecting a record on the card
Brain Read now has record_in to choose which record it binds, and the picker is available on the card with inline New actions. Brand includes its design system and separate web/video CSS, HTML, and motion. Product, Character, and Place add typed visual, voice, 3D, HDRI, and audio fields instead of hiding everything in a prose body.