WTF – Where’s The Food 🍔📸

Track: Mobile App

---

Table of Contents

• Overview
• How to Set-Up
• Core Features
  • 📷 Food Image → Restaurant Search
  • 🗺️ Contextual Planning
  • 🧠 Multi-Agent Dining Evaluation System
  • 📞 Action Layer
  • 🛡️ Safety & Relevance Guardrails
• System Architecture
• Backend Pipelines
  • 🔹 Pipeline 1 – Image to Yelp Discovery
  • 🔹 Pipeline 2 – Multi-Agent Verdict System
• Compliance with Hackathon Rules
• Live App Builds
• App Screenshots
• Demonstration Video

---

Overview

WTF (Where’s The Food) is a mobile-first application that helps users identify where they can find a dish they see online or in real life.
Users upload a screenshot or food photo, choose location, date, and time, and our system uses computer vision + LLM reasoning + the Yelp AI API to discover matching restaurants, rank options, and provide an agent-driven dining verdict — including whether it’s better to dine in or order delivery at that moment.

The app is designed for social-media-driven discovery:

Saw food on Instagram or TikTok and want to know where to get it? Screenshot → upload → decide.

---

How to Set-Up

1. Clone the Repository

Open your terminal and run the following command to clone the project into your desired folder:

git clone https://github.com/GIND123/Yelp-AI.git

2. Navigate to the Project Directory

Move into the mobile application folder:

cd Yelp-AI/yelp-mobile

3. Install Dependencies

Install the necessary packages using npm:

npm install

4. Run the Application

Start the development server. This command will generate a QR code in your terminal:

npm start

5. Launch on Your Device (Android)

To view the app on your physical device:

1. Download and install the Expo Go app from the Google Play Store on your Android device.
2. Crucial: Ensure your computer and your phone are connected to the same Wi-Fi network.
3. Open Expo Go on your phone.
4. Use the "Scan QR Code" feature within Expo Go to scan the QR code generated in your terminal.

Core Features

📷 Food Image → Restaurant Search (Primary Yelp AI Workflow)

• Upload an image or provide a caption.
• Our AI generates a precise Yelp AI query sentence including:
  • Dish type inferred from the image
  • User intent (dietary preferences or style)
  • Location, date, and time
• Query is sent directly to Yelp AI Chat API to retrieve candidates.
• Results are ranked by rating and review count from Yelp’s data.

🗺️ Contextual Planning

Users specify:

• Location
• Date
• Time

This enables:

• Checking availability patterns
• Prioritizing places likely open and ready to serve
• Identifying ideal options for dine-in vs pick-up windows

---

🧠 Multi-Agent Dining Evaluation System

Each selected restaurant is analyzed through a 3-agent debate system:

✅ Optimistic Agent

Summarizes:

• Strengths
• Food quality highlights
• Good service patterns
• Convenience and value

❌ Critical Agent

Identifies:

• Recurring drawbacks
• Reliability issues
• Crowding, cleanliness, or service risks

⚖️ Judge Agent (Final Verdict)

Produces a single neutral recommendation paragraph:

• Balanced overall assessment
• Ideal visitor type or time window
• Cautions if relevant

The verdict answers:

Is this the right place for me right now? Order in or dine out?

---

📞 Action Layer

Each recommendation includes instant actions:

• 📞 Call Now – opens native phone dialer
• 🗓️ Book on Yelp – deep links to Yelp’s reservation/booking page
• 📍 View Location – quick navigation support

---

🛡️ Safety & Relevance Guardrails

A built-in moderation layer ensures:

• Only food- or dining-related searches proceed.
• Irrelevant or unsafe queries are blocked or redirected.
• Image uploads unrelated to dining discovery are automatically rejected.

This keeps the system aligned strictly with its intended use case.

---

System Architecture

---

Backend Pipelines

🔹 Pipeline 1 – Image to Yelp Discovery

• Accepts uploaded images or text captions.
• Uses Gemini multimodal generation to produce a single precise Yelp query sentence.
• Queries Yelp AI Chat Endpoint to retrieve businesses.
• Normalizes:
  • Ratings
  • Review counts
  • Photos
  • Hours
  • Booking availability
• Sorts by rating + popularity.

Implemented in: Pipeline1Backend.py :contentReference[oaicite:1]{index=1}

---

🔹 Pipeline 2 – Multi-Agent Verdict System

For a selected Yelp business:

1. Fetch business details and real reviews using Yelp Fusion API.
2. If reviews are unavailable, fallback to Yelp AI summary extraction.
3. Run the Optimist, Critic, and Judge agents using Gemini.
4. Produce the final actionable verdict.

Implemented in: Pipeline2Backend.py :contentReference[oaicite:2]{index=2}

---

Compliance with Hackathon Rules

✅ Primary Data Source: Yelp AI API
✅ No Third-Party Location Data Mix
✅ Original Work Created During Submission Period
✅ Fully Functional End-to-End Flow
✅ Public Repository with Setup Instructions
✅ Hosted Build for Testing
✅ 3-Minute Demo Video Included

---

Live App Builds

• Android APK:
  👉 https://your-hosted-apk-link.com

---

App Screenshots

---

Demonstration Video

🎥 Watch full demo (≈3 minutes):
👉 https://youtube.com/your-demo-video

The video covers:

• Image upload
• Time/location selection
• AI processing
• Yelp AI discovery flow
• Restaurant ranking
• Multi-agent verdict
• Booking/calling actions

---

Includes:

• Mobile frontend
• Backend pipelines
• Environment configuration
• API setup instructions

---