DataStax recently released new tools that promise to help customers build GenAI apps on their data platforms, joining a growing group of database specialists. But for customers using multiple databases, the question remains: which vendor should they choose as the primary foundation to support their GenAI application development?
Based on the open source Cassandra database, DataStax's customer list includes CapitalOne, Condé Nast and Saab. The company acquired Langflow, an open source visual framework for building AI applications, in April and is now previewing Langflow on its Astra Cloud platform, enabling developers to rapidly design and pilot GenAI and search augmentation generation (RAG), a popular approach for augmenting LLM with enterprise data.
DataStax promises a cloud-hosted, drag-and-drop visual interface for LangChain, while its separate product, LangSmith, offers integration with GenAI tools like OpenAI, Hugging Face, document database MongoDB, and vector database Pinecone, as well as RAGStack 1.0, a system designed to streamline RAG implementations at enterprise scale.
talk RegistryJason McClelland, chief marketing officer at DataStax, said organizations are struggling to build AI applications with tools like LangChain, LangSmith and Unstructured.io new to the market and rapidly improving.
“What CTOs are frustrated with is that developers spend all their time building infrastructure instead of building apps,” he says. “Two days later, something works, but then it breaks because one of the underlying tools has changed again. So our premise is to work with these partners to build a version of their thing and then integrate that with ours.”
McClelland claims that the DataStax AI platform provides integrated tools for data preparation, vector padding, data ingest, and data chunking (splitting data into smaller, more manageable chunks).
But DataStax isn't the only database vendor encouraging users to develop AI applications on its platform: NoSQL database vendors Couchbase and MongoDB, like Oracle, also offer AI application development tools.
The problem is that most large enterprises rely on many different databases, some of which have fundamentally different architectures than others, so why build AI on one database and not the others?
IDC research vice president Carl Olofsson said DataStax has developed features that make RAG much faster and easier than usual.
“A natural question is whether this will strengthen or entrench the use of Cassandra, or whether it will become an alternative sales channel for DataStax that is not dependent on users actually adopting Cassandra for data management, since LLM and Vector are typically targeted at unstructured data,” he said.
Matt Aslett, director at technology consulting and research firm ISG, said it was justified for database vendors to add genAI development tools to their platforms because much of a company's data is stored there.
But he cautioned: “Generative AI doesn't change the underlying trends. Companies have always used multiple databases for multiple types of applications and end use cases. This will continue to be the case with AI.”
An initial or pilot project starting with an existing data platform might make sense, especially if the skills are there.
“But if you acknowledge that a company has multiple data platforms and is investing heavily in AI, they likely have a separate team that handles the data science for the AI elements, perhaps AI-specific platforms and tools, in combination with the databases,” he said.
He said organizations looking to tweak existing applications with AI are likely to base them on a database vendor's technology, but those looking to build new applications that leverage GenAI are likely to look for a development platform independent of individual database vendors.®
