Peakride: Offline-First Fleet Operations for Mountain Towns

Mussoorie, perched at 2,000 meters in Uttarakhand, sees massive influxes of weekend tourists escaping the heat of Delhi and Dehradun. Local scooty and motorcycle rental operators run brisk, high-stakes businesses, but are hamstrung by five pervasive friction points:

Instead of building a typical thin client tethered to a remote backend API, we built an offline-first distributed system consisting of two unified interfaces:

"Built for a hill town where the network drops constantly. The operator can take a booking with zero connectivity — it queues in IndexedDB and reconciles on reconnect, including conflict handling when two devices booked the same scooty offline."

When building offline collaborative applications, engineers frequently reach for CRDTs (Conflict-free Replicated Data Types) or Operational Transformation. However, in rental operations, a physical vehicle is an exclusive, indivisible resource. You cannot merge two bookings for the same motorcycle. Merging edits or averaging time intervals is nonsensical when only one physical handlebar exists.

Instead, we designed an authoritative pure domain rule engine (R1–R10) executed with idempotent operation queues. The supreme rule of the physical world governs resolution: Physical vehicle handover trumps a paper booking.

In our client-side demo, overlap prevention is enforced inside single Dexie read-write transactions. When deploying to a production backend, this same mathematical guarantee maps directly to PostgreSQL GiST exclusion constraints:

ALTER TABLE bookings
ADD CONSTRAINT no_overlap
EXCLUDE USING gist (
  vehicle_id WITH =,
  tstzrange(start_at, end_at + interval '30 minutes') WITH &&
)
WHERE (status IN ('held', 'confirmed', 'on-road'));

This eliminates race conditions at the database engine level, ensuring zero double-booking is physically possible even under extreme concurrent requests.

To eradicate customer disputes over scratches, Peakride uses client-side cryptographic hashing:

• Photo Capture Hashing: At pickup, 6 guided photographs (front, left, right, rear, odometer, fuel) are captured and immediately digested into SHA-256 hashes in memory.
• Seal Chaining: The inspection payload (odometer, fuel bars, damage zones, and photo hashes) is sealed into a root hash. Upon return, the return inspection chains to the pickup seal hash.
• Customer Verification Report: When a tourist opens /report, the browser recomputes the cryptographic hashes of every photo and reading from storage. If anyone modified a number or photo, the badge instantly switches from Verified ✓ to Tampered ✗.

We architected Peakride to respect the Digital Personal Data Protection Act, 2023 principles:

This portfolio showcase proves the complete user experience, data schema, and business domain logic client-side. To transition from this zero-database demo to multi-location production, our agency modularised clean interface adapters: