• 

  adk web

• 

  app diagram

ADK MultiAgent Tourist Guide - Project Showcase

Inspiration

Travel planning is notoriously fragmented and time-consuming. We were inspired by the frustration of having to visit multiple websites and apps to gather weather information, find attractions, discover restaurants, and plan walking routes. We wanted to create a unified AI-powered solution that could handle all aspects of travel planning through natural conversation, leveraging Google's cutting-edge Agent Development Kit (ADK) to demonstrate the power of multi-agent AI systems.

What it does

Our Travel Assistant Agent is a comprehensive AI-powered travel companion that consolidates all travel planning into one intelligent conversation. It features:

• Multi-Agent Architecture: An orchestrator agent that intelligently routes requests to 8 specialized sub-agents
• Real-time Weather: Current conditions and forecasts for any destination
• Smart Attractions: Top sights and hidden gems with detailed descriptions
• Restaurant Recommendations: Best dining spots with cuisine types and price ranges
• Interactive Walking Routes: Google Maps links with step-by-step directions
• Photo Analysis: Upload travel photos for instant insights and recommendations
• Travel Blog Generation: AI-written blog posts about destinations
• Visual Search: Find images of landmarks and attractions
• Conversation Memory: Maintains context across interactions

The system provides 90% time savings in travel research while delivering personalized, real-time recommendations.

How we built it

Backend Architecture:

• FastAPI with Google ADK for the core AI framework
• Multi-Agent System: Orchestrator agent + 8 specialized sub-agents
• Real-time APIs: OpenWeather for weather data, Google Maps for navigation
• Session Management: Maintains conversation context across interactions
• Multimodal Support: Text and image processing capabilities

Frontend:

• Streamlit with beautiful, responsive UI
• Real-time Chat Interface: Natural conversation flow
• Photo Upload: Drag-and-drop image analysis
• Interactive Elements: Buttons, maps, and rich content display

Deployment:

• Railway for backend hosting with automatic scaling
• Streamlit Cloud for frontend deployment
• Docker containerization for consistent environments
• Comprehensive Testing: Unit tests, integration tests, and end-to-end validation

Tech Stack:

• Python 3.10+, FastAPI, Google ADK, Streamlit
• Docker, Railway, Streamlit Cloud
• Google AI API, OpenWeather API
• Comprehensive logging and monitoring

Challenges we ran into

Multi-Agent Orchestration: The biggest challenge was designing an effective orchestrator agent that could intelligently route requests to the appropriate specialized agents. We had to carefully craft prompts and implement context-aware decision-making logic.

API Integration Complexity: Integrating multiple external APIs (Google AI, OpenWeather, Google Maps) while handling rate limits, error responses, and data formatting required extensive testing and error handling.

Deployment Issues: Railway deployment initially failed due to builder configuration conflicts. We had to switch from Nixpacks to Dockerfile builder and resolve routing issues where the /api prefix was being stripped.

Photo Processing: Implementing multimodal capabilities for photo analysis required careful payload structuring to ensure images were properly transmitted to the AI model.

Agent Reliability: Some agents (like walking routes) initially failed on vague queries. We had to expand the knowledge base and improve error handling to make them more robust.

Accomplishments that we're proud of

Production-Ready Multi-Agent System: Successfully built and deployed a sophisticated 8-agent AI system that demonstrates the power of Google ADK in real-world applications.

Comprehensive Travel Solution: Created a unified platform that handles weather, attractions, restaurants, navigation, photo analysis, and content generation - all through natural conversation.

Robust Deployment: Achieved successful deployment on both Railway (backend) and Streamlit Cloud (frontend) with proper error handling, monitoring, and scaling.

Real-time Integration: Successfully integrated live weather data, Google Maps navigation, and interactive map links that work seamlessly in the chat interface.

User Experience: Built an intuitive chat interface that feels natural and provides immediate value, with features like photo upload and interactive map links.

Extensible Architecture: Created a modular system where new agents can be easily added and existing ones can be enhanced without affecting the overall system.

What we learned

Multi-Agent AI Design: Deep understanding of how to design, implement, and orchestrate multiple specialized AI agents using Google ADK, including prompt engineering and context management.

Production Deployment: Learned the intricacies of deploying AI applications to cloud platforms, including environment configuration, API key management, and scaling considerations.

API Integration Best Practices: Gained expertise in integrating multiple external APIs, handling rate limits, error responses, and data transformation.

User Experience for AI: Discovered how to design intuitive interfaces for AI-powered applications, balancing functionality with usability.

Real-time System Design: Learned how to build systems that provide immediate, accurate responses while maintaining conversation context and handling concurrent users.

Testing AI Systems: Developed strategies for testing AI applications, including unit tests for individual agents and integration tests for the complete system.

What's next for ADK MultiAgent Tourist Guide

Enhanced Personalization: Implement user profiles and preference learning to provide more personalized recommendations based on travel history and preferences.

Voice Integration: Add voice input/output capabilities for hands-free travel assistance, especially useful while exploring destinations.

Real-time Translation: Integrate language translation services to help travelers communicate in foreign countries.

Booking Integration: Partner with travel booking APIs to allow users to book flights, hotels, and activities directly through the chat interface.

Offline Capabilities: Implement offline mode with cached data for areas with poor internet connectivity.

Social Features: Add the ability to share travel plans with friends and family, and collaborate on group trip planning.

Advanced Analytics: Implement usage analytics to understand user behavior and continuously improve the recommendation engine.

Mobile App: Develop native mobile applications for iOS and Android to provide a more seamless mobile experience.

Expanded Agent Capabilities: Add new specialized agents for transportation planning, cultural insights, safety information, and local events.

Enterprise Features: Develop enterprise versions for travel agencies, hotels, and tourism boards to enhance their customer service capabilities.

Built With

• docker
• fastapi
• google-adk
• google-ai-api
• openweather-api
• python-3.10+
• railway
• streamlit
• streamlit-cloud