Restaurant Finder with AG-UI and CrewAI

October 3, 2025 ยท View on GitHub

This repository contains a complete solution for finding restaurant recommendations using AI agents. It demonstrates the integration of CrewAI with the AG-UI protocol to create interactive, human-in-the-loop AI applications.

Project Structure

This project consists of two main components:

  1. Backend Agent (ag-ui-restaurant-agent/): A Python-based AI system using CrewAI for orchestrating multi-agent restaurant search and recommendation workflow.

  2. Frontend Application (ag-ui-restaurant-frontend/): A Next.js web application that provides a user interface for interacting with the restaurant finder agent.

Features

  • Restaurant Discovery: Find restaurants in any location worldwide
  • Real-time Updates: Follow the agent's progress with live state updates
  • Structured Results: Receive restaurant information in a consistent, well-organized format
  • Human-In-The-Loop: Provide feedback on recommendations to get refined suggestions
  • Streaming Responses: Experience real-time responses as the agent works
  • AG-UI Protocol: Standardized communication between frontend and backend

Backend Agent

The backend agent uses CrewAI to coordinate specialized AI agents:

  • Restaurant Research Specialist: Searches for and gathers details about restaurants
  • Recommendation Specialist: Creates personalized recommendations based on research

For full details and setup instructions, see the agent README.

Frontend Application

The frontend application provides an intuitive user interface for:

  • Entering location queries
  • Viewing restaurant recommendations
  • Providing feedback to refine recommendations
  • Following the agent's progress in real-time

For frontend setup instructions, see the frontend README.

Getting Started

To run the complete application:

  1. Set Up the Backend Agent:

    cd ag-ui-restaurant-agent
poetry install
poetry run uvicorn src.restaurant_finder_agent.api:app

  2. Set Up the Frontend Application:

    cd ag-ui-restaurant-frontend
npm install
npm run dev

  3. Open your browser and navigate to http://localhost:3000

Architecture Overview

The application follows a client-server architecture where:

  • The backend agent provides a FastAPI server implementing the AG-UI protocol
  • The frontend consumes this API using CopilotKit's React components
  • Real-time communication happens via Server-Sent Events (SSE)
  • State management ensures consistent UX with optimistic updates

Demo

To test the system quickly, you can run the example client:

cd ag-ui-restaurant-agent
poetry run uvicorn src.restaurant_finder_agent.api:app

Development

Each component can be developed independently:

  • The backend agent can be tested in standalone mode using python -m restaurant_finder_agent.main
  • The frontend can be developed with mock data in development mode

License

This project is intended as a demonstration of AI agent technology and CrewAI integration with AG-UI protocol.

Acknowledgments

  • CrewAI - Multi-agent orchestration framework
  • CopilotKit - Frontend toolkit for AI agents