TouaRAG

Enhance Your Travel Experience with Personalized RAG Assistance 🌍✈️

Empowering Personalized Travel Assistance with Retrieval-Augmented Generation

TouaRAG is your comprehensive framework for building and evaluating context-aware travel chatbots. Leverage our modular library touarag—packed with implementations for testing various RAG architectures and evaluating their performance—to deliver tailored travel recommendations. In addition, our solution offers an integrated API backend and UI frontend to effortlessly run a RAG chatbot with profile personalization, model selection, and architecture configuration.

Prerequisites:
This framework is designed for Linux environments (due to bash script usage). Make sure to install the required libraries using requirements.txt before proceeding. Additionally, ensure that PostgreSQL 15 is installed on your system.

Key Features

• Personalized Travel Assistance: Deliver customized travel itineraries based on user profiles.
• RAG Evaluation: Test and evaluate multiple retrieval-augmented generation architectures.
• Integrated Chatbot: Seamlessly run a travel assistant with both API backend and UI frontend.
• Easy Deployment: Manage installation and uninstallation with simple wrapper scripts.
• Modular Design: Swap out models, tweak architectures, and refine evaluation metrics effortlessly.

Installation

Prerequisites

• Operating System: Linux (for bash script compatibility)
• Dependencies: Install required libraries with:

  pip install -r requirements.txt

• Database: PostgreSQL 15 with PGVector plugin must be installed

Install the TouaRAG Library

./install.sh

Uninstall the TouaRAG Library

./uninstall.sh

Secrets Management

To securely store your secrets (e.g., tokens, usernames, passwords), place them in the config.yml file located in the src/api directory.

Quickstart

Launch Your Personalized Travel Chatbot

Get started quickly by running our integrated API backend and UI frontend with a single command:

./start_app.sh

This wrapper script launches the complete TouaRAG travel assistant application, allowing you to:

• Interact with a dynamic travel chatbot
• Personalize user profiles for tailored recommendations
• Choose from various RAG architectures and models

Evaluate a RAG Architecture with TouaRAG

Below is a brief example showcasing how to evaluate a travel query dataset using the touarag library:

query_engine = TransformQueryEngine(...)

    # Create an evaluator
    evaluator = Evaluator(eval_dir=EVAL_DIR,
                        query_engine=query_engine,
                        sampling=False, # Use for testing purposes
                        scenario_label="test_scenario_1"
    )
    evaluator.generate_samples()
    evaluator.save_samples(output_dir="...")
    evaluator.evaluate(output_dir="...")