Couchbase enhanced its Capella iQ platform by integrating Amazon Bedrock, allowing it to support multiple foundation models and improve operational resilience. The company aimed to create a flexible, scalable, and resilient inference architecture to meet growing enterprise demand for AI-powered developer assistance. This integration enables Capella iQ to handle complex tasks such as generating SQL++ queries, recommending indexes, and managing multi-turn conversations without requiring code changes or downtime.
The architecture spans two AWS Regions, us-east-1 and us-west-2, ensuring high availability and supporting Cross-Region Inference (CRIS). It uses an Amazon Elastic Kubernetes Service (Amazon EKS) cluster to run Capella iQ microservices, including the cp-api pod, which orchestrates inference calls and manages session context. The system routes inference traffic through a private VPC interface endpoint, keeping data within AWS infrastructure and meeting enterprise security and data residency requirements.
Couchbase evaluated multiple models on Amazon Bedrock, ultimately selecting Anthropic’s Claude Sonnet 4.5 for its 76 percent accuracy in internal evaluations. This model supports Capella iQ’s diverse workflows, including structured code generation and multi-turn reasoning. The company also developed a model evaluation framework to quickly qualify and adopt newer models as they become available on Amazon Bedrock.
Source: awsml
