On June 12th, Artium Engineer Justin Beall hosted a virtual meet up where he shared his hands-on experience using OpenAI Assistants. During the session he walked through code examples and demonstrated how to process files, create an assistant using the OpenAI API, and run chat sessions.

One of the projects he demoed was the Amazon Treasures Chat. This project came to light after having built an AI chatbot for a large event ticketing client using natural language search as a mechanism for product discovery. Justin wanted to find a data set that he could use to create something similar without any proprietary IP attached. So he had a conversation with APEX (Artium Product Explorer) and hashed out the details.

Amazon Treasures Chat

Anyone who has used Amazon knows that it can sometimes be hard to find exactly what you're looking for due to their massive catalog. Amazon Treasures Chat aims to address this issue by providing a conversational interface for easy product discovery and tracking. It also ensures that only relevant results from the curated catalog are returned.

Technology

The core technology stack includes:

• OpenAI API Assistants: For creating conversational AI that can interact with users, providing information, tracking products, and personal recommendations.

• Vector Store: A newly released feature by OpenAI for simplifying the retrieval of information, essential for the proposed simplification of the RAG architecture.

• Python and OpenAI’s Python SDK: For backend development, including data loading, normalization, and interaction with OpenAI’s services.

• Amazon Products Dataset 2023: As the primary dataset, containing detailed information on 1.4M products for users to explore. Cloud Services: For hosting, storage, and scalable computation needs, potentially using AWS given the Amazon-centric nature of the product.

Key Features

• Conversational UI for intuitive product exploration.

• Data normalization and preparation for the Amazon Products Dataset 2023.

• Integration with OpenAI API Assistants and Vector Store for efficient search and retrieval.

• Real-time tracking and update functionalities for user’s favorite items.

• Personalized product recommendations based on user behavior.

• Seamless backend management tools for developers to create, test, and manage the AI Assistant.

• Robust security and privacy measures to protect user data.

• Scalable architecture to accommodate expanding product catalogs and user base.

• Multi-platform support, ensuring access through various operating systems and devices.

Check out the full product details on GitHub as well as a recording of the entire virtual meet up.