Lower the Cost of Building and Running Visual AI Agents with NVIDIA VSS Blueprint 3.3

NVIDIA announced VSS Blueprint 3.3, an update to its Metropolis Blueprint for Video Search and Summarization, aimed at reducing the cost of building and running visual AI agents. The blueprint connects vision-language models such as NVIDIA Cosmos, LLMs such as NVIDIA Nemotron, retrieval-augmented generation, and Model Context Protocol tools to turn live and recorded video into natural-language search, visual Q&A, verified alerts, and automated reporting.
VSS 3.3 introduces two updates. The Build Vision Agent skill (vss-build-vision-ai) composes multi-workflow deployments from a single prompt, starting from one of four validated developer profiles NVIDIA calls the Foundation and adding only the services a requested capability reaches, while converging shared infrastructure such as Kafka, Redis, and Elasticsearch onto single instances. NVIDIA says the skill produced a live, previewable bottling-line deployment with search, alert verification, and shift reporting in under 30 minutes on a two-GPU RTX PRO 6000 Blackwell host, reusing the detector's GPU for FP8 Cosmos 3 Nano.
Adaptive Efficient Video Sampling (EVS) addresses runtime cost by dynamically pruning unchanged visual patches per frame and batching VLM work around moments of activity. On an RTX PRO 6000 Blackwell running Cosmos 3 Super FP8, NVIDIA reports Adaptive EVS cut alert contextualization latency 17%, from 1,021 ms to 844 ms, and increased concurrent real-time VLM streams 46%, from 13 to 19. It also summarized a 60-minute video in about half the time with 80% fewer VLM input tokens. NVIDIA states results vary with scene motion, chunk length, and similarity threshold, and advises benchmarking representative footage before choosing production defaults.
Adaptive EVS runs inside the RT-VLM container rather than against remote endpoints and is optional, enabled through override.env settings. NVIDIA also scheduled a live session on Oct. 1 at 9 a.m. PT demonstrating building a visual AI agent from a single prompt.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Vision-language models have made it possible to build visual AI agents that understand video at production scale. The harder problem is turning that capability...