Run natural language searches across archived video using NVIDIA's VSS stack with Cosmos Embed1 embeddings. You can ingest files or RTSP streams, then query with plain English like "find all forklifts" or "who entered the restricted area at 2pm." Requires the search profile deployed and Elasticsearch running, plus a three-step chunked upload flow that wires VIOS storage through to the embedding pipeline. Marked alpha and not production ready. The routing is solid but you'll spend time on the prerequisites: make sure your GPU has headroom for the models and your NGC keys are current. Good fit if you're already in the NVIDIA Blueprint ecosystem and need semantic video retrieval without building your own embedding infrastructure.
npx -y skills add nvidia/skills --skill vss-search-archive --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills