Global Explorer: The Autonomous Travel Strategist

This project was developed as part of the AI Agents Technical Course hosted by HIGGS AI Talent Club.

Course Information

• Course: AI Agents Technical Course
• Organization: HIGGS AI Talent Club
• Lecturer: Natia Kukhilava (@Ranginho)
• Video Demo: YouTube

---

Project Overview

Global Explorer is a state-of-the-art autonomous travel planning agent designed to handle complex, multi-turn travel logistics. It leverages the power of LangGraph for state management and Google Gemini 2.5 Flash for high-level reasoning and decision-making.

Technical Architecture

1. Stateful Graph Logic (LangGraph)

The agent is built on a cyclical directed acyclic graph (DAG) that allows it to:

• Reason: Analyze user requirements (departure, dates, budget).
• Act: Execute external tools to gather real-world data.
• Verify: Use a Human-in-the-loop (HITL) interrupt for final itinerary approval.

2. Tool Integration

• DuckDuckGo Search: For real-time flight and hotel pricing.
• Wikipedia: For destination history and cultural context.
• Open-Meteo API: For live weather forecasting.
• Nager.Date API: For local holiday checking to avoid travel conflicts.

3. Safety & Resilience

• Memory: Uses MemorySaver to persist conversation state across threads.
• HITL: Implements a critical approval node to ensure the agent never finalizes a plan without explicit user consent.
• Constraint Enforcement: A strict system prompt ensures the agent collects all mandatory data (budget, dates, location) before planning.

How to Run

1. Open the notebook in Google Colab.
2. Add your GOOGLE_API_KEY to the Colab Secrets (🔑 icon).
3. Run all cells.
4. Interact with the agent in the 'Real-Time Interactive Agent' section.

---

Developed for educational purposes in the pursuit of mastering AI Agent orchestration.