Travel Recommendation System Architecture
Welcome to our comprehensive overview of the Travel Recommendation System architecture. This presentation will guide you through the system's layers, components, and deployment options, providing insights into how we leverage modern technologies to deliver personalized travel recommendations.
Our system combines user-friendly interfaces with powerful AI capabilities to process travel preferences and generate tailored itineraries. We'll explore how each component works together to create a seamless experience for travelers seeking destination guidance.

by Petru GIURCA

Tech Stack
The Travel Recommendation System aims to provide comprehensive travel suggestions based on user input and preferences. By combining AI technologies with specialized travel agents, the system processes user requirements and generates personalized itineraries that match budget constraints, destination interests, and accommodation needs.
User & Frontend Layers
User Layer
The foundation of our system begins with the end user who provides essential travel information including:
  • Desired destinations
  • Travel dates
  • Budget constraints
  • Personal interests
Frontend Layer
Built with Streamlit, our frontend provides:
  • Interactive web UI
  • Responsive design
  • Intuitive form inputs
  • Seamless data transfer to core systems
The user interface captures travel preferences and passes them to the core processing layer, ensuring a smooth user experience while collecting all necessary information for personalized recommendations.
Core Processing Layer
Google AI Agentic Framework
Central orchestration component responsible for travel request processing and workflow coordination. Manages communication between specialized agents and ensures cohesive recommendation generation.
LangChain Integration
Facilitates the integration of Large Language Models (LLMs) for natural language understanding and generation. Enables the system to process complex travel queries and generate human-like responses.
Agent Coordination
Orchestrates the various specialized agents, distributing tasks and aggregating results to create comprehensive travel plans tailored to user preferences.
The core layer serves as the brain of our system, processing user inputs and coordinating specialized services to generate cohesive travel recommendations that align with user preferences.
Specialized Travel Agents
Flight Agent
Handles flight queries, performs budget analysis, and matches preferences to find suitable flights.
Hotel Agent
Processes accommodation requests considering type preferences and budget constraints.
Activity Agent
Suggests relevant activities based on interest filtering and weather adaptation.
PDF Generator
Creates formatted travel itineraries for export and sharing.
Each specialized agent focuses on a specific aspect of travel planning, leveraging dedicated algorithms to optimize recommendations within their domain. The agents work in concert, orchestrated by the core layer, to create a comprehensive travel plan.
Containerization & Storage
Docker Container
The entire application is packaged as a Docker container, ensuring consistency across different environments and simplifying deployment processes.
Google Artifact Registry
Serves as the container image repository, storing Docker container images securely and making them available for deployment.
Configuration Files
Includes secret.yaml for sensitive information, configmap.yaml for environment settings, deployment.yaml for pod specifications, and service.yaml for load balancer configuration.
Our containerization approach ensures that the application runs consistently across different environments while maintaining security and configurability through dedicated configuration files.
Deployment Options
Google Kubernetes Engine
Single-node cluster deployment with external access via load balancer
Google Cloud Run
Serverless container platform with auto-scaling capabilities
Security Features
HTTPS endpoints and IAM integration for access control
Performance Optimization
Environment variables for configuration and auto-scaling for traffic management
The system offers flexible deployment options through either Google Kubernetes Engine (GKE) with a single-node cluster or Google Cloud Run as a serverless container platform.
Complete Workflow Summary
User Input
Traveler provides preferences through Streamlit interface
Core Processing
Google AI Agentic Framework orchestrates specialized agents
Data Gathering
Agents query relevant sources for flights, hotels, and activities
Plan Generation
System combines information into comprehensive travel plan
Delivery
Results displayed to user with PDF export option
The complete workflow demonstrates how user inputs flow through the system, triggering specialized agents to gather relevant information that is then combined into a comprehensive travel plan. This end-to-end process leverages key technologies including Streamlit, LangChain, Google AI Agentic Framework, Docker, Python, and Google Cloud services to deliver personalized travel recommendations.