Mapify AI

A multi-agent city assistant that turns natural-language intent into personalized, explainable local recommendations.

🌐 Website · 🎮 Demo · 🎥 Video

---

Overview

Mapify AI rethinks how people explore cities.

Instead of browsing static map results, users can describe their needs in natural language (e.g., “dinner near ETH tonight under CHF 30”), and the system automatically retrieves, filters, ranks, and explains suitable places.

At its core, Mapify AI is a map-based multi-agent system that integrates heterogeneous signals — including travel time, opening hours, pricing, ratings, and reviews — into a structured decision-making pipeline.

---

System

The system is implemented as a LangGraph-based multi-agent pipeline, where each stage performs a specific reasoning step:

User Query → Intent Parsing → Information Retrieval → Transit Filtering → Score Calculation and Ranking & Review Analysis → Recommendation Synthesis

A key design choice is the separation between:

• evaluation (scoring)
• review understanding (textual signals)

which are later merged by an orchestrator before generating final recommendations.

This modular design enables:

• interpretable intermediate states
• flexible extension of agents
• clearer debugging and evaluation

---

Key Capabilities

• Natural-language query → structured constraints
• Multi-factor ranking (price, travel time, rating)
• User-controlled preference weighting
• Explainable recommendations with reasoning
• Real-time pipeline execution (streaming)

---

Demo

• Website: https://mapify-ai-eight.vercel.app/landing
• Interactive frontend demo available at: https://mapify-ai-eight.vercel.app

---

Architecture

The system consists of:

• a React frontend for interaction and visualization
• a FastAPI + LangGraph backend for multi-agent orchestration
• external services (LLM, Google Maps via Apify, transport APIs)

Workflow as following:

---

Quick Start

Backend

cd backend
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env   # fill in OPENAI_API_KEY and APIFY_API_TOKEN
uvicorn app.main:app --reload

API at http://localhost:8000 · Swagger UI at http://localhost:8000/docs

Frontend

cd frontend
npm install
npm run dev

App at http://localhost:3000

See backend/README.md for full environment variable reference and pipeline details.

---

Limitations

This is a hackathon prototype:

• Merchant-side features are frontend-only
• Some backend modules are partially scaffolded
• External APIs are required for full functionality

---

Future Work

• Full marketplace integration (merchant ↔ user loop)
• Learning-based personalization
• Improved retrieval beyond seed data
• Unified streaming architecture

Contributors

• Deqing Song
• Qing Dai
• Yuqing Huang
• Yuan Yu

---

License

MIT — built at GenAI Zurich Hackathon 2026.