Introduction & Problem Statement

This project addresses the limitation of traditional flight search tools that prioritize only cost or speed. It offers travelers a way to discover meaningful journeys by factoring in enriching stopovers, turning layovers into mini-adventures personalized to the user’s interests.

Specifically, it helps plan trips from Melbourne (MEL) to Bengaluru (BLR) with either direct or one-stop flights, and for stopovers like Singapore, provides curated city guides including food, hotels, and attractions.

System Architecture & How It Works

The system is a multi-agent AI travel planner built using CrewAI to orchestrate modular agents:

• Flight Planner Agent: Searches for direct and one-stop flights via the Amadeus API.
• Route Evaluator Agent: Scores flight options combining cost and stopover experience aligned to user interests.
• Local Guide Agent: Generates 1-2 day city guides for stopover cities using Serper web search and fallback scraping techniques.
• User Interface: A Streamlit app lets users input trip details and interests, then displays enriched journey options.

The agents collaborate seamlessly, passing context and results to deliver both cost-effective and experience-rich travel plans.

Production-Readiness Enhancements

Testing

• Basic unit tests exist to verify agent outputs and core functions.
• Tests cover flight search responses, route scoring, and itinerary generation.
• Tests can be run using pytest.

Safety & Guardrails

• User inputs are validated in the UI to prevent invalid queries.
• Fallback logic ensures city guides are always generated even if web scraping or API calls fail.
• API keys and sensitive credentials are stored securely in environment variables (.env), with no hardcoded secrets.

User Interface

• The Streamlit UI provides an interactive form to enter trip parameters.
• Results page shows direct and stopover flight options side-by-side.
• Stopover city guides detail food, hotel, and attraction recommendations.
• Users can download their full travel plan.

Operational Resilience

• Agents have error handling and retry logic for API failures.
• Progress feedback in UI improves user experience.
• Modular codebase structured for maintainability.
• Deployment tested on Streamlit Cloud.

Documentation

• Comprehensive README with setup and usage instructions.
• Clear explanations of architecture and design choices.
• Environment variable requirements documented.

Failure Handling, Monitoring, and Deployment

• API request failures trigger retries with exponential backoff.
• Fallback city guides ensure consistent output despite external service failures.
• Logs essential steps for debugging (local logs currently; can be extended).
• Deployed via Streamlit Cloud for easy access and testing.

Summary

This project delivers a production-grade multi-agent AI travel planner that balances cost and meaningful experiences. It demonstrates:

• Effective multi-agent orchestration with CrewAI.
• Integration of real-time flight data and AI-generated city guides.
• A user-friendly UI that supports exploration and decision-making.
• Basic testing and safety features ensuring stable operation.

This system is ready to be extended further but currently provides a fully functional, deployable, and testable travel planning solution.