Database company Couchbase Inc. is enhancing the capabilities of its artificial intelligence agent-building tools, Capella AI Model Services, by integrating them with Nvidia Corp.’s NIM microservices.

The update will make it easier for Couchbase Capella database users to create so-called agentic AI applications that go beyond the capabilities of traditional chatbots by performing tasks on behalf of their users, with minimal supervision required.

Couchbase introduced Capella AI Model Services when it launched its Capella AI Services offering in December, saying they can help simplify the development of AI agents. They also provide tools for creating and embedding managed endpoints within the large language models that power those AI agents, ensuring they can meet organization’s performance, latency and privacy requirements.

Couchbase Capella is the cloud-based version of the open-source Couchbase NoSQL database, and its main advantage compared to traditional databases like Oracle is that it can process both structured and unstructured information, which makes it an ideal choice for AI applications that need to access both types of data. It also acts as a data cache.

High-performance AI agents with NIM

NIM, which stands for “Nvidia Inference Microservices,” packages optimized inference engines, industry-standard application programming interfaces and pretested large language models into software containers for easy deployment. Other packaged tools include Nvidia’s NeMo Guardrails to help enforce policies and safeguards against AI hallucinations. In addition, it also allows organizations to bring their own proprietary data to implement retrieval augmented generation and enhance the knowledge of its LLMs.

By integrating Capella AI Model Services with NIM, Couchbase said customers can bring their AI agents closer to both the underlying data and the graphics processing units that power them. That means they can build higher-throughput AI applications with greater model flexibility.

Couchbase Senior Vice President of Product and Partners Matt McDonough said AI agents demand a unified and highly performant data platform that supports every step of the application development lifecycle.

“We’re giving customers the flexibility to run their preferred AI models in a secure and governed way, while providing better performance for AI workloads and seamless integration of AI with transactional and analytical data,” he explained.

McDonough said this is necessary because enterprises often face challenges in terms of AI agent reliability and compliance. It’s important they get their AI agents right, he said, especially if they’re going to be used in customer-facing applications, as unreliable responses can deal massive damage to brands’ reputations. Moreover, they must also be secure, to avoid data leaks that violate privacy regulations, he added.

Over the last year, Couchbase has enhanced Capella’s capabilities in an effort to position it as the database of choice for AI developers. In February 2024, it introduced vector search and retrieval-augmented generation capabilities, while integrating AI frameworks such as LlamaIndex and LangChain to support the development of generative AI applications. Then in September, the company further enhanced Capella’s AI capabilities with Capella Columnar, to support the development of more advanced generative AI applications that can analyze real-time data to deliver more personalized experiences.

Image: SiliconANGLE/Dreamina