The Module Fit lane (Arrival board)

The Module Fit lane on the Arrival board lists the Turtini modules we'd turn on for you at Create, based on the vertical the classifier inferred. Each row is a checkbox you can flip. The header reads "N of M on" so you always know how many will be activated.

How the suggestions are computed:

Each vertical (TemplateCategory) maps to a small handful of module keys:
• **Restaurant** → websites, restaurant, workforce, accounting
• **Lodging** → websites, lodging, workforce, accounting
• **Outdoor** → websites, events, workforce, accounting
• **Retail** → websites, retail, workforce, accounting
• **Services** → websites, requests, accounting
• **Federal** → websites, federalGov, contracts, accounting
• **Nonprofit** → websites, events, accounting
• **Events** → websites, events, venue
• **Portfolio** → websites
• **Landing** → websites
• **Other** → websites

The map lives in src/lib/arrival.ts and stays in sync with the keyword classifier's vertical detection. Wally's LLM classifier overrides the vertical (and therefore the module set) when its second-opinion arrives, but you can always uncheck anything.

Why these defaults:

Each vertical's defaults are the smallest set we're confident will be useful day one — not everything that COULD apply. A restaurant gets the restaurant module, a workforce module (assuming staff), and accounting (assuming books). A nonprofit gets events and accounting because they're the highest-leverage starting points; payroll only matters once they have employees and is one click away from /modules.

Editing after Create:

The Setup Cockpit at /welcome has a "Your modules" section showing every active module + a "Recommended for you" section with Wally's LLM-driven suggestions for inactive modules that fit your business description. You can turn anything on/off from Settings → Modules at any time.

Sticky toggle behavior:

If you uncheck a module and keep typing in the describe textarea WITHOUT changing the vertical, your choice sticks. If the classifier flips to a different vertical (you typed "wine bar" → it inferred Restaurant; you added "with rooms upstairs" → it inferred Lodging), the suggestion set re-seeds. This is intentional: a vertical flip is a structural change, your per-vertical choices are a refinement.