Verba, developed by Weaviate, is an open-source Retrieval-Augmented Generation (RAG) application designed to provide personalized, AI-driven answers using your own data. Its modular architecture allows for easy customization and integration with various AI models and data sources.

Key Features: • Data Ingestion: Verba supports importing various data formats, including text files, PDFs, and GitHub repositories. It utilizes tools like UnstructuredIO for seamless data ingestion. • Data Chunking: The application offers multiple chunking strategies, such as token-based and sentence-based chunking, to segment documents into manageable pieces for efficient processing. • Embedding and Retrieval: Verba integrates with embedding models from providers like OpenAI, Cohere, and Hugging Face. It uses Weaviate’s vector database for storing and retrieving data based on semantic relevance.  • Answer Generation: The platform supports various language models, including GPT-4o to generate accurate and contextually relevant answers. • User Interface: Verba features a user-friendly web interface that allows for easy data upload, management, and interaction. Users can view sources directly in the UI, with highlighted chunks for transparency.