Universal Reservation Platform • Version 3.0 | Prepared by Javad Ghased

AI-Powered Universal Reservation Platform

A conversational booking platform that replaces rigid search forms with natural language. Plugs into any reservation backend via a sub-second Jev + 9router AI pipeline.

< 500ms

Jev Pre-Classification

Intercepts greetings & missing info without touching the LLM.

60–70% of traffic handled free

60–70%

AI Cost Optimization

9router routes each query to the most cost-effective capable model.

Drastic monthly savings

100%

Domain-Agnostic Backend

Accommodations, flights, car rentals, venues, or activities.

Universal schema integration

The Proposition

Summary

Eliminating the friction between customer intent and confirmed reservations through conversational AI.

JetTrip is a conversational booking platform that replaces rigid search forms with natural language. A customer simply types «یک رزرو برای دو نفر آخر هفته با دسترسی به ویلچر» and instantly sees matching availability on an interactive map with cards, detail panels, and a one-tap booking flow.

While originally designed for travel, the platform is entirely domain-agnostic and applicable to any backend service for reserving anything — accommodations, flights, rental cars, venues, restaurant tables, or activities.

What makes JetTrip fundamentally different from traditional booking platforms is its AI-first architecture:

1. Jev — The Fast Intelligence Layer

A lightweight intent classifier powered by the TypeSafe AI SDK that intercepts simple requests (greetings, clarifications) in under 500ms without touching an LLM.

2. LLM via 9router — Intelligent Routing

Complex queries (multi-constraint search, spatial proximity, sorting) are routed to the optimal large language model with tool calling, saving 60–70% in API costs.

نمونه ورودی کاربر: «یه جای ارزون برای هفته آینده نزدیک مرکز شهر»

طبیعی و محاوره‌ای

The Problem

Why current form-based booking engines leak revenue and frustrate modern mobile users.

67%

Search Abandonment

Users abandon searches when forced through more than 3 filter steps. Dropdowns, date pickers, and checkbox matrices bleed conversions.

24/7

Support Overhead

Call centers and live chat agents handle repetitive queries ("سلام، برای آخر هفته جای خالی دارید؟") that should be answered automatically in <500ms.

70%+

Mobile Frustration

Over 70% of bookings originate on smartphones, where multi-column filter forms and tiny dropdowns are nearly unusable.

Commoditization

When multiple agencies sell similar inventory, customer loyalty is won by who offers the fastest, most effortless booking experience.

How It Works

Every query moves through a high-throughput sequence with deterministic verification and sub-second feedback.

User Prompt

Natural Persian or English speech/chat input.

«رزرو برای فردا»

Latency: 0ms

Jev Fast Layer

TypeSafe AI classifies intent & checks parameters.

< 500ms instant

Bypasses LLM

9router & LLM

Selects optimal model. Strictly calls typed tools.

searchAvailability()

Zero Hallucination

Reservation API

Proprietary DB with real-time pricing & PostGIS query.

Live Inventory

Direct DB Sync

Map & Drawer UI

Pins, cards, and pre-filled checkout link.

One-Tap Reserve

Mobile Responsive

Backend Communication

How AI Talks To Your System

The LLM never generates reservation data directly. It only produces structured JSON payloads validated against a strict schema.

{
  "action": "searchAvailability",
  "parameters": {
    "location": "tehran",
    "capacity": 2,
    "startDate": "2026-10-15",
    "duration": 2,
    "tags": ["quiet", "accessible"],
    "priceTo": 5000000
  }
}

What It Looks Like

Every touchpoint engineered for desktop dual-drawer efficiency and mobile bottom-sheet ergonomics.

SCREENSHOT 01

Main Chat & Map First-View

The user opens JetTrip to an immersive map interface with conversational natural language input at the bottom.

Fullscreen Mapbox / Neshan

SCREENSHOT 02

Search Results & Spatial Pin Clustering

Interactive pins populate dynamically on the map with price indicators, auto-panning to encapsulate all matching properties.

Geo-Spatial Clustering

SCREENSHOT 03

Dual-Drawer Ranked Results List

Cards stack vertically in a side drawer showing photos, ratings, amenities, and nightly prices. Selecting a card pans the map to its pin.

Bidirectional Map-List Sync

SCREENSHOT 04

Multi-Turn Conversational Refinement

Follow-up prompts like "ارزون‌ترش رو نشون بده" re-sort and filter existing results instantly without reloading or starting from scratch.

Persistent Conversation Memory

When It Ships

Delivering Plan

Structured in 5 distinct milestones with bi-weekly demo releases on staging.

Phase 01

Foundation & API Integration

Weeks 1–3 (3 Weeks)

Project setup, schema design, universal search endpoint with dynamic filtering, PostGIS geospatial query layer for proximity search.

Phase 02

AI Integration (Jev + LLM)

Weeks 4–6 (3 Weeks)

Jev intent classification (greeting, missing_info). 9router setup. Function calling schema (searchAvailability, filter, sort). Multi-turn context memory.

Phase 03

Chat & Map UI Implementation

Weeks 7–9 (3 Weeks)

Message bubbles with RTL Persian support. Phased loading indicator. Map pin rendering and dual-drawer results list + detail view.

Phase 04

Booking Flow & Mobile Polish

Weeks 10–11 (2 Weeks)

Direct reserve button checkout handoff. Pre-fill booking dates from chat context. Responsive bottom-sheet mobile layout. Search caching.

Phase 05

Testing, Analytics & Launch

Weeks 12–14 (3 Weeks)

End-to-end Playwright tests, conversion analytics dashboard, white-label styling configuration, staging UAT, and production launch.

The Next Step

Going Multi-Source

While this Platform Edition connects directly to your agency's proprietary inventory, JetTrip can also be deployed as a meta-search aggregator.

The Aggregator Edition queries Jabama, Jajiga, Eghamat24, and Shab.com simultaneously, executes parallel fan-out requests with a 3-second timeout, deduplicates listings within a 50m radius, and shows cross-provider price comparison tags.

Read Full Aggregator Proposal (v3.1) 16-Week Multi-Provider Plan

Why Choose JetTrip

Capability Dimension	Traditional Booking Portals	JetTrip Universal Platform
Search Modality	Rigid dropdowns and calendar date pickers	✓ 100% Natural language Persian & English
Search Abandonment	67% drop off after 3+ filter steps	✓ Conversational frictionless prompt matches intent
AI Response Speed	3 to 5 second wait for simple greetings	✓ Sub-500ms immediate response via Jev layer
AI Operating Cost	High monthly frontier model API bills	✓ 60-70% lower costs through Jev + 9router
Hallucination Risk	AI invents non-existent properties	✓ Strict schema tool calling — 100% real database results
Mobile Experience	Cramped multi-filter forms on 6" screens	✓ Conversational chat + full screen map & bottom sheets
White-Label Ready	Rigid brand templates	✓ Fully themeable colors, prompts, and domain per agency