Xpand：面向阿联酋 PropTech 的决策智能原型，覆盖 39 个区

See what every area in the UAE is missing — and where to build next.

Submission


Project	Xpand — Decision Intelligence for UAE PropTech
Demo	http://localhost:3000 (landing) · http://localhost:3000/dashboard?tour=1 (guided tour)
Track	Decision Intelligence
AI	K2 Think V2 — preconfigured in `.env`
Repo	https://github.com/shivamtiwari3/Expand

npm install && npm run dev
# → open http://localhost:3000

No extra setup — K2 API key is included for the hackathon demo.

What is Xpand?

Xpand is decision intelligence for land, investment and communities — built for the Abu Dhabi AI PropTech Challenge. It joins challenge datasets (parcels, communities, transactions, investors) with real OpenStreetMap amenities and live UAE listings to:

Detect service gaps across 39 districts in 7 emirates
Forecast 12-month demand from price momentum and growth
Turn gaps into RFPs — need, land and matched investors
Explain with AI — K2 Think when configured, fast grounded engine offline

Data is synthetic outside Abu Dhabi core figures — never present as official market data.

Demo flow

Landing · `/`

Marketing site with animated hero, feature grid and scroll sections for Map · Gaps · RFPs · Community · Ask Xpand.

CTA: Start guided tour →

Dashboard · `/dashboard`

Interactive MapLibre map + district sidebar + AI right rail.

9-step onboarding spotlights each panel on first visit (replay anytime via Tour in the header).

3-minute pitch script

Time	Beat
0:00	Planners can't see where services will break next — Xpand scores every district
0:20	Open dashboard → critical district lights up on the map
0:45	Xpand recommends — AI grounded on gaps + forecast + investors
1:30	Open matching RFP → vacant land + interested funds
2:00	Ask K2: "Which areas lack primary care?"
2:30	Advisor wizard — tailored insight for role / budget
2:50	"Today a prototype — with real feeds, the intelligence layer for UAE PropTech."

Full script: docs/demo-script.md

Quick start

git clone https://github.com/shivamtiwari3/Expand.git
cd Expand
npm install
npm run dev

Open http://localhost:3000 — landing page
Open http://localhost:3000/dashboard?tour=1 — guided onboarding

K2 Think is on by default (see .env). Switch to offline mode:

XPAND_AI_PROVIDER=engine

All model calls run server-side via POST /api/run — keys never touch the browser.

Architecture

Landing (/)          Dashboard (/dashboard)
     │                      │
     └──────►  Map + Sidebar + Right rail tabs
                         │
                    POST /api/run
                         │
              LangGraph (lib/ai/graph.ts)
                 gather → synthesize? → finalize
                         │
           ┌─────────────┴─────────────┐
     Xpand engine (default)      K2 Think V2
     lib/engine.ts               lib/ai/model.ts

Tab	Mode	What it does
District	`district`	Auto AI recommendation on district select
RFPs	—	Open proposals from widest gaps
Community	—	Ventures matched to district needs
Ask K2	`qa`	Free-form Q&A across UAE data
Advisor	`insights`	Role → emirate → sectors → budget wizard

Data pipeline: scripts/build_uae.mjs → lib/xpand-data.json + lib/live-snapshot.json

Project structure

app/
  page.tsx              Landing + demo sections
  dashboard/page.tsx    Main interactive app + onboarding tour
  api/run/route.ts      Server-side AI endpoint
components/
  dashboard/            Map, sidebar, panels, OnboardingTour
  marketing/            Hero, header, demo-flow sections
lib/
  engine.ts             Gap scoring, forecast, RFP generation
  ai/                   LangGraph pipeline, K2 client, tools
public/
  readme-banner.svg     Animated README header
docs/
  demo-script.md        3-minute pitch (filled in)

Built with

Next.js 16 · App Router · TypeScript · Tailwind 4
MapLibre GL — UAE district map with severity styling
LangGraph + LangChain — tool-grounded AI pipeline
K2 Think V2 — MBZUAI reasoning model (optional)
Framer Motion — landing animations + onboarding tour