DataStax introduces Astra DB Hybrid Search, boosting AI search relevance by 45%

DataStax has introduced Astra DB Hybrid Search, a new feature designed to improve Retrieval Augmented Generation (RAG) systems by enhancing search relevance by 45%. The offering integrates vector search and lexical search capabilities, which aims to deliver more accurate AI-driven search and recommendation results. The hybrid search leverages NVIDIA NeMo Retriever for text reranking, allowing for increased contextual understanding while ensuring critical keyword matches. This combination is intended to address challenges in Generative AI (GenAI) applications, where poorly ranked results can lead to irrelevant user responses. Ed Anuff, Chief Product Officer at DataStax, emphasized the importance of retrieval accuracy, stating that many customers require over 95% accuracy to effectively implement enterprise Artificial Intelligence (AI) solutions. He noted that Astra DB Hybrid Search accelerates the path to achieving these accuracy levels. Logistics company GoDash plans to utilize Astra DB Hybrid Search to streamline operations and improve insight delivery for shipping clients. According to GoDash's founder and CEO Aditya Swami, the integration of both search methods will allow for real-time retrieval of relevant shipment details and customer feedback, contributing to operational efficiency and customer satisfaction. Developers can implement Astra DB Hybrid Search using the Astra DB Python client and schema-less Application Programming Interface (API). The service is hosted with GPUs to facilitate efficient AI workload operations without the complexity of managing infrastructure. The new hybrid search feature will also be accessible through Langflow, an open-source tool aimed at low-code AI application development, which is designed to simplify the optimization of search relevance for developers.