TravelBud

Inspiration

Planning travel can be overwhelming—finding the best flights, hotels, food, attractions, and local tips often means juggling multiple apps and websites. We wanted to create a smarter, AI-powered assistant that streamlines this process, provides personalized recommendations, and leverages the power of decentralized agents for real-time, trustworthy results.

What it does

TravelBud is an intelligent travel assistant that answers your travel queries in natural language. Whether you need to book a flight, find a hotel, discover local cuisine, or get sightseeing tips, TravelBud connects your request to specialized AI agents. Each agent is designed for a specific domain (flights, hotels, food, attractions, local tips) and uses real-time search via Tavily for up-to-date information. The system is orchestrated using Fetch.ai’s uAgents and deployed on Agentverse for reliability and scalability.

How we built it

We started by designing the core architecture around the Fetch.ai ecosystem, using uAgents for all inter-agent communication. Each specialized agent was initially built as a LangGraph agent, then converted to a uAgent using the uAgents adapter. The user assistant agent acts as the main entry point, classifying incoming queries with ASI LLM and routing them to the relevant specialized agent. All agents are deployed on Agentverse for easy discovery and robust operation. The frontend is built with React framework, communicating with the backend via a REST API.

Challenges we ran into

• Agent orchestration: Ensuring smooth communication and correct routing between multiple agents.

• Protocol compliance: Adhering to Fetch.ai’s chat protocol for message formats and acknowledgements.

• Real-time data: Integrating Tavily search for up-to-date travel info.

• Async coordination: Managing asynchronous responses and timeouts between the user assistant and specialized agents.

• Deployment: Packaging and deploying all agents on Agentverse for public access.

Accomplishments that we're proud of

• Seamless end-to-end travel query handling using decentralized AI agents

• Successfully adapting LangGraph agents to uAgents

• Real-time, accurate travel recommendations powered by Tavily and ASI LLM

• A clean, user-friendly Streamlit interface

What we learned

• Deepened our understanding of agent-based architectures and the Fetch.ai ecosystem

• Practical experience with the ASI LLM for structured query understanding

• Best practices for async programming and RESTful API design in Python

• The importance of modular design for agent scalability and maintainability

What's next for TravelBud

• Expand to support more travel domains (e.g., activities, transportation, budget tracking)

• Enhance personalization using user profiles and preferences

• Integrate more real-time data sources

• Improve UI/UX with richer visualizations and chat features

Built With

• agentverse
• asi1-llm
• fetchai
• langraph
• openai
• python
• react
• tavily
• uagents