NVIDIA AI Blueprint for video search and summarization (VSS) provides GPU-accelerated reference architectures for real-time video analytics, Q&A, and summarization workflows. It includes multiple microservices, agent workflows, and deployment options.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (2 so far).
What it is
NVIDIA AI Blueprint for Video Search and Summarization (VSS) is a GPU-accelerated reference architecture for building vision agents and AI-powered video analytics applications. The repository implements the blueprint and powers the NVIDIA build experience for natural-language video agents—search, summarization, visual Q&A, and related workflows—backed by generative AI, VLMs and LLMs, and NVIDIA NIM microservices as configured in the stacks below.
How it works
The blueprint combines real-time video intelligence, downstream analytics, and agent/offline processing. It integrates three core areas: real-time feature extraction and embeddings, downstream analytics for enriched metadata and verified alerts, and an agent layer that uses the Model Context Protocol to access data and orchestrate tools such as VLMs, embeddings, long video summarization, and clip retrieval. It relies on NIM microservices (Cosmos3 Nano Reasoner, NVIDIA Nemotron-Nano-9B-v2), VLMs, and LLMs to perform search, Q&A, and summarization tasks.
Getting started
Quickstart Guide
Follow the steps from the documentation and notebook in the deploy/docker/scripts directory to deploy the blueprint using Brev Launchable. The launchable notebook is deploy_vss_launchable.ipynb and is tailored for AWS CSP.
Docker Compose Deployment
See prerequisites and system requirements listed in the README, including OS versions, NVIDIA drivers, Docker tools, and NGC CLI versions. A Docker upper bound note is included: Docker Engine 29.5.0+ may fail pulling NGC-hosted images; use Docker Engine 28.3.3 or another supported version below 29.5.0.
Prerequisites
Obtain API keys via NVIDIA API catalog or NG C and ensure access to NVIDIA NIM for local hosting. API keys steps are provided in the linked docs.
License
License information is referenced but not listed in the repository metadata.
Recent releases
- v3.2.1 VSS 3.2.1 (2026-07-23): Key Features include updating the default VLM model to Cosmos Reason 3 Nano and updating the Real-Time Video Language Model (RT-VLM) microservice and agent workflows; Smart City Industry profile updated.
- v3.2.0 VSS 3.2.0 (2026-06-16): Source code released for all microservices and agent workflows; Agent Skills (Early Access) available for agent-driven deployment and operations.
- v3.1.0 VSS 3.1.0 (2026-03-18)
- v2.4.1 VSS 2.4.1 (2026-01-28): v2.4.1
- v2.4.0 VSS 2.4.0 (2025-09-24)
Traction
- 1773 stars, 367 forks, 124 open issues; 19 stars gained in the last 7 days.
Behind the repo
N/A
Caveats
- License: none listed
- Created: 2024-10-22
- Last push: 2026-08-04
- Requires NVIDIA API keys and access to NVIDIA NIM via NVIDIA AI Enterprise license; prerequisites reference external docs