Backpacking Assistant Agent

A unified, multi-agent travel companion designed to address the fragmentation and manual overhead inherent in planning complex, multi-destination trips.

Overview

The process of planning a complex backpacking trip is often fragmented and labor-intensive, forcing travelers to rely on a disjointed ecosystem of tools. This project centralizes trip planning and management into a single, "AI-native" platform.

At its core is a modular AI architecture where a cohort of specialized agents automates key planning domains. These agents leverage large language models like Google’s Gemini 2.5 Flash for creative and structured data generation and Perplexity AI for real-time, citation-backed web research.

The system is architected to provide holistic and personalized support, maintaining a persistent, context-aware understanding of user preferences, past trips, and chat history.

Key Features

• Specialized AI Agents: A cohort of agents handles specific domains:
  • Itinerary Generation: Automates route planning and activity suggestions.
  • Visa & Vaccination: conducts research on entry and health requirements.
  • Accommodation: Provides personalized stay suggestions.
• Advanced AI Integration:
  • Gemini 2.5 Flash: Used for creative content and structured data generation.
  • Perplexity AI: Enables real-time, citation-backed web research.
• Unified Experience: Eliminates context-switching by combining research, itinerary management, and budget tracking in one modern web application.
• Scalable Architecture: Built with LangGraph for complex orchestration, currently optimized for performance via direct agent invocation.

Tech Stack

• Monorepo: Turborepo
• Frontend: Next.js 15 with React
• Backend: FastAPI (Python 3.11+)
• Database: PostgreSQL 16 (via Supabase)
• AI Framework: LangGraph
• LLMs: Google Gemini 2.5 Flash, Perplexity AI
• ORM: Drizzle ORM (TypeScript) / SQLModel (Python)
• Real-time: Supabase Realtime / Yjs
• Auth: NextAuth.js
• Deployment: Vercel (frontend), Railway (backend)
• Containerization: Docker

Getting Started

1. Clone the repository:

   git clone https://github.com/your-username/backpacking-assistant-agent.git
   cd backpacking-assistant-agent

2. Install dependencies:

   npm install

3. Set up environment variables:

   • Create a .env file in apps/web by copying the example: cp apps/web/.env.example apps/web/.env
   • Fill in the required values in apps/web/.env.

4. Run the development servers:

   docker-compose up --build

The frontend will be available at http://localhost:3000 and the backend at http://localhost:8000.

Citation

Kohki Hatori. 2025. Backpacking Assistant Agent. 1, 1 (December 2025), 8 pages. https://doi.org/10.1145/nnnnnnn.nnnnnnn