JFrog AI Catalog now integrates NVIDIA NIM (Inference Microservices) models as governed artifacts within the JFrog Software Supply Chain Platform. Previously, developers and coding agents pulled NIM models directly from NVIDIA's NGC registry, bypassing organizational security policies, RBAC, and audit trails. With this integration, NIM models flow through JFrog Artifactory alongside Docker images and packages, subject to the same Curation allow/block policies, version tracking, and audit logging. The setup involves a three-step process: authenticate via Docker login through Artifactory, discover models in a unified catalog alongside Hugging Face models, and deploy via generated Docker commands. The goal is to eliminate the governance gap where AI models could enter production environments without policy checks or traceability.
