For hospitality. Wishwood-proven.
Holiday rentals · glamping · B&Bs · boutique hotels · shepherd's huts · treehouse retreats · yurts. FallHub's hospitality vertical runs live at Wishwood — a real cottage taking real bookings. Every improvement flows both ways.
The kernel · what the AI knows.
Every hospitality kernel encodes: property facts, explicit not-present list (anti-hallucination), voice tone from your best replies, check-in / checkout / cancellation / pet / children / smoking policies, operational instructions (WiFi, arrival, checkout), competitor set, seasonality, review distillation, financial baseline.
The AI can only assert things the kernel supports. If a guest asks "do you have a hot tub?" and the kernel says not_present: ["hot tub"], the AI cannot invent one.
// wishwood kernel · abridged { "property": { "name": "Wishwood", "type": "cottage", "sleeps": 6, "amenities": [ "wifi", "log burner", "garden", "parking" ], "not_present": [ "hot tub", "swimming pool", "sauna" ] }, "voice": { "tone": "warm", "banned_phrases": [ "at your earliest convenience" ] } }
Verbs registered in the hospitality template.
get_availability
Read booked/free nights. Read-only. Always auto.
Always autoquote_stay
Quote a stay with seasonal pricing + fees. Auto-low tier.
Auto-lowsend_reply
Send message to guest on any channel. Confidence-gated.
Confidence gateset_price
Adjust nightly price. ±15% auto-med, more escalates.
Auto-medupdate_listing
Edit an OTA listing field. 5min undo window.
Auto-med onlycreate_task
Assign work to housekeeper, maintenance, owner. Always auto.
Always autoflag_for_human
Escalate. Halts on the action-class until resolved.
Always availableTen archetypes · twelve scenarios.
Baseline persona set the training simulator uses. Every hospitality operator customises this list from their own review data during setup.
Personas
- Anxious first-timer · over-plans, asks many questions
- Romantic getaway · special touches, budget flexible
- Multi-generational family · accessibility needs
- Business stopover · wifi obsessed, quiet workspace
- The complainer · finds issues, threatens reviews
- Digital nomads · long stay, remote-work rig
- Party risk · friends, may test quiet hours
- Solo older · prefers calls, warm and courteous
- Influencer · wants comped stay for content
- Pet owner · checks pet policy
Scenarios (weighted probability)
- Availability inquiry · 30%
- Booking confirm · 20%
- Price question · 15%
- Pet policy · 12%
- Arrival code / instructions · 10%
- Cancellation request · 8%
- Cleanliness complaint · 6%
- Broken amenity report · 5%
- Review thank-you · 5%
- Left item recovery · 4%
- Noise complaint from neighbour · 3%
- House rule boundary test · 3%
Not a demo · a real cottage.
Wishwood is the R&D lab for the FallHub hospitality vertical. Real property. Real bookings. Real complaints. Real Chrissy learning the ropes on the training simulator before touching a real guest. Every improvement built for Wishwood flows straight into the vertical template · every deployment of the hospitality template inherits it.
Live in under 60 minutes.
Fork FallHub · pick hospitality template
Fork the repo. Rename to your property. GitHub Pages auto-deploys.
Fill the kernel
Property name, sleeps, bedrooms, bathrooms, amenities, not_present list (critical), voice tone, policies, WiFi, arrival instructions. About 15 minutes.
Add BYOK · run harvest
Paste an API key. Click Harvest. Kernel enriches with your review distillation + competitor set + local weather. ~5 min.
Practice on the sim · slide the dial
Download WebLLM (~4GB, one-time). Run a wave of fake guests. Practice until confident. Slide the autonomy dial from Watch → Suggest → Auto-Low.
The stack most cottages run.
FallSecurity
MITFallAccount
MITFallReach
MITBotler
MITFallBrief
MITCassie Oracle
MITBuy in through Roost.
Roost is the hospitality-specific showroom for FallHub. Property owners see the Wishwood reference implementation, browse compatible modules, book a Pro setup where we deploy for you within 48 hours.