...
Elevator pitch :
Are you tired of sifting through endless documents to find the information you need? Our Hyperledger Labs project is here to revolutionize your search process. We’re developing an open-source conversational AI tool designed to streamline your workflow by delivering context-specific answers swiftly and accurately.
Our prototype features a chatbot that leverages a RESTful API and requires GPU resources for optimal performance. Built with Python, this project serves as a versatile framework that other communities and organizations can adapt to their needs. By employing Retrieval Augmented Generation (RAG), our system enhances answer quality by integrating knowledge from external databases, like vector databases.
The application’s architecture supports two key workflows: data ingestion and chat interaction. During ingestion, the system loads context documents and constructs a vector database from resources such as online guides and wiki pages. Post-ingestion, the chatbot is primed to respond to user queries efficiently.
Currently utilizing the open-source HuggingFace Zephyr-7b-beta model, we plan to explore other models to further improve performance. Users can interact with the system via HTTP requests, with future plans to offer UI samples for enhanced accessibility. Join us every Monday for a public meeting to learn more about our progress and contributions to the Hyperledger Labs community.