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

DataStax has introduced Astra DB Hybrid Search, which utilizes NVIDIA NeMo Retriever technology to enhance Retrieval Augmented Generation (RAG) systems. This development is reported to improve search relevance by 45%. The hybrid search functionality combines vector and lexical search to provide precise keyword matching and contextual relevance.

Ed Anuff, Chief Product Officer at DataStax, noted that high accuracy in retrieval is crucial for deploying enterprise Artificial Intelligence (AI) effectively. He emphasized that Astra DB Hybrid Search accelerates achieving the necessary accuracy levels. This combination of search methods aims to minimize irrelevant answers that can hinder user experience in Generative AI (GenAI) applications.

The incorporation of NVIDIA NeMo Retriever helps to automatically reorder search results, enhancing the relevance of responses generated by AI. GoDash, a logistics software provider, has indicated that this new hybrid search capability will optimize its operations, allowing for quicker retrieval of shipping details and operational insights.

To facilitate integration, developers can use the Astra DB Python client, which supports schema-less APIs. The hybrid search is available on Astra DB with GPUs, allowing the execution of AI workloads efficiently. This functionality will also be accessible through Langflow, an open-source tool for low-code AI application development.