Welcome to the Disney Chatbot! Interact with live data for Disney and other global theme park attractions! 🌍
This project showcases a Retrieval-Augmented Generation (RAG) chatbot designed to generate human-like responses based on real-time data from worldwide theme parks. The chatbot uses:
- Langchain and Qdrant for vector-based data retrieval and generation.
- OpenAI GPT-3.5-turbo for natural language generation.
- DeepEval Evaluation to assess response quality, relevance, and factuality.
Started by connecting to the ThemeParks API to gather data on various theme parks and their attraction wait times. Each park has a unique ID for accessing detailed attraction data, which is exported to an Excel file for easier visualization.
The retrieved data is stored in the Qdrant vector database. Any pre-existing data is cleared and replaced with the latest data for testing purposes. In a full implementation, functions would be created to detect and handle data updates automatically. Qdrant was chosen for its open-source model and active community, making it ideal for this project.
Using Langchain, I seamlessly connect OpenAI’s API with Qdrant, enabling complex functions and classes. This setup allows for smoother data retrieval and more effective RAG-based conversations.
I refined responses by embedding contextual information around user queries. This process enhances answer quality, creating conversational, informative responses while reducing potential AI hallucinations.
In the final stage, DeepEval is used to measure response accuracy, relevancy, and detect any factual inaccuracies. Responses are iteratively adjusted to sound more human-like, prioritizing clarity and conversational tone.
This chatbot architecture provides a foundation for future enhancements, such as:
- Predictive routing and user journey optimization.
- Advanced wait time forecasting based on live data.
🎉 Thank you for exploring this project! Feel free to dive into the code and reach out with any questions.
Icons sourced from Font Awesome and Material Icons