Multiple stages
Nights per stop prevent gaps and overlaps across the journey.
← All projectsCASE 02
Web-App · Live
Travel Planner
AI research becomes a genuinely useful travel product
A complete multi-stop travel app with realistic daily plans, accommodation, a journal and private photo storage.
PORTUGAL · 8 DAYSCoastlines,
Coastlines,
culture & calm.
03 — 11 September
YOUR ITINERARYAlgarve escape
•••03
Arrive in FaroOld town · Dinner
NOW04
Ria FormosaBoat · Island walk
05
Road to LagosCoast · Check-in
1–30days per trip
nflexible stages
privatesigned photo URLs
CHbuilt in Switzerland
The starting point
The real problem.
A useful itinerary must account for dates, stages, transport, budget and personal interests at once. Long research tasks also cannot depend on short web-request timeouts.
The solution
One connected system.
The app models each trip structurally and delegates research to asynchronous workflows. Results are validated, written back and presented in a calm mobile interface.
Product experience
Complex logic.
Clear interaction.
A researched itinerary emerges from destination, dates and interests. The app then continues to support the trip with status, stages, a journal and secure uploads.
PORTUGAL · 8 DAYSCoastlines,
Coastlines,
culture & calm.
03 — 11 September
YOUR ITINERARYAlgarve escape
•••03
Arrive in FaroOld town · Dinner
NOW04
Ria FormosaBoat · Island walk
05
Road to LagosCoast · Check-in
Key capabilities
Asynchronous planning
Long-running research works independently from the browser and returns through a secure callback.
Daily plan & suggestions
Activities, flights and accommodation include sources and cost estimates.
Travel journal
Memories, mood, location and private photos stay connected to the trip.
System architecture
From input to a useful outcome.
Every component has a clear responsibility, keeping data flows, failures and future extensions traceable.
01Trip form→
02App API→
03n8n research→
04Validated callback→
05Supabase & storage→
06Mobile itinerary
Engineering decisions
Deliberate decisions instead of invisible magic.
Good systems define not only what should happen, but also what deliberately remains outside automation.
- Private storage buckets instead of public images
- Magic-byte upload validation
- Rate limits for paid workflows
- Tolerant input handling with strict normalisation