{"resourceId":"nvidia-vss-blueprint-3-3-adaptive-evs-20260929","versions":[{"version":"external-e75c31f6cfc25ef8290e923cb654a63763961db17ee7bd13dded58edbaaff9c0","resource":{"id":"nvidia-vss-blueprint-3-3-adaptive-evs-20260929","title":"NVIDIA: VSS Blueprint 3.3 adds Build Vision Agent skill and Adaptive EVS, reporting 46% more VLM streams and 80% fewer tokens on hour-long summaries","organization":"NVIDIA","sector":"Vendor (computer vision / video analytics platform)","geography":"Global","publishedAt":"2026-09-29T18:35:09Z","publicationDate":"2026-09-29","eventDate":"2026-09-29","sourceName":"NVIDIA Technical Blog","sourceLabel":"Vendor release / Metropolis VSS Blueprint 3.3","sourceUrl":"https://developer.nvidia.com/blog/lower-the-cost-of-building-and-running-visual-ai-agents-with-nvidia-vss-blueprint-3-3/","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["developers-agents","infrastructure","operating-model","data-security"],"finding":"NVIDIA's Sep 29, 2026 post (modified Oct 1) releases the Metropolis Video Search and Summarization (VSS) Blueprint 3.3. The Build Vision Agent skill (vss-build-vision-ai) composes multi-workflow deployments from a prompt, starting from four validated profiles (base, alerts, lvs, search), adding only needed services and sharing Kafka, Redis and Elasticsearch; it shows an architecture diagram before deployment and defaults to a NemoClaw harness. In a synthetic bottling-line demo on a two-GPU RTX PRO 6000 Blackwell host, a live deployment was previewable in under 30 minutes with a few dollars of coding-agent usage. Adaptive Efficient Video Sampling prunes unchanged patches by cosine similarity and batches around activity; on an RTX PRO 6000 Blackwell with Cosmos 3 Super FP8, NVIDIA reports alert contextualization latency down 17% (1,021 to 844 ms), concurrent real-time VLM streams up 46% (13 to 19), and a 60-minute summary in about half the time with 80% fewer VLM input tokens. NVIDIA says results vary with scene motion, chunk length and similarity threshold, that EVS helps less for long outputs from few frames and runs only inside the RT-VLM container (not against remote endpoints), and that deployments belong on trusted isolated networks with authentication, TLS and rate limiting. Why now: Oct 1 update and livestream, not previously archived; the post names smart-city vehicle detection and collision alerting as target use cases.","sledRelevance":"City traffic, campus safety and public-safety video teams evaluating VLM-based search and alerting can treat the EVS figures as a hypothesis for camera capacity per GPU, provided they benchmark on their own footage and address video privacy governance.","evidence":"Opened NVIDIA's technical blog post (published Sep 29, modified Oct 1, 2026). All performance figures are NVIDIA's own tests on one GPU type and model; the demo footage is synthetic.","architectureImplications":"Adaptive EVS runs inside the local RT-VLM container, so capacity gains require on-premises or edge GPUs rather than remote model endpoints; consolidating shared infrastructure avoids duplicated Kafka and Elasticsearch per workflow.","governanceImplications":"Prompt-generated video-analytics deployments still need human review of the architecture plan, data retention, and camera-use policy before go-live in public spaces.","securityPrivacyImplications":"NVIDIA advises trusted isolated networks with authentication, TLS and rate limiting; video of the public raises surveillance and retention concerns that the blueprint does not resolve.","caveats":"Vendor-reported results on synthetic or NVIDIA-chosen footage; no customer or public-sector outcome; accuracy impact of token pruning is not quantified in the post; tied to NVIDIA Cosmos/Nemotron models and RTX PRO GPUs.","streamIds":["nvidia"],"roles":{"sales":"Position VSS 3.3 as a capacity hypothesis, not a result: NVIDIA reports Adaptive EVS raised concurrent real-time VLM streams 46% (13 to 19) and cut hour-long summary tokens 80% on one RTX PRO 6000, with results varying by scene motion. For city traffic or campus safety video teams, that could mean more cameras per GPU. Ask how much of their footage is static and what retention rules apply. Do not promise these percentages, accuracy parity, or public-sector outcomes, which the post does not show.","engineering":"Fit: local GPU deployments doing alert verification, dense captioning or long-video summarization over mostly static scenes. Integration: Build Vision Agent composes from four validated profiles and shares Kafka, Redis and Elasticsearch; EVS runs only in the RT-VLM container, not against remote endpoints. Prerequisites: RTX PRO-class GPUs, NGC access, representative footage. Constraints: vendor-only benchmarks and synthetic demo video. Proof of value: run EVS on and off over a week of real footage and measure missed events, alert latency and streams per GPU.","delivery":"Delivery work is a bounded video-AI pilot with an EVS on/off comparison on representative cameras. The public-safety or campus-safety technology lead owns operations; privacy and records officers approve retention and camera scope before footage is used. Checkpoints: review of the generated architecture plan, a privacy impact assessment, and an accuracy review. Proposed acceptance: no increase in missed verified events with EVS on, and a measured streams-per-GPU gain. Risks include surveillance concerns, lock-in to NVIDIA models, and over-trusting prompt-built stacks."},"roleDetails":{"sales":{"customerProblem":"Video teams want VLM search and alert verification across many cameras, but GPU cost per stream is high.","stakeholders":["Public-safety technology lead","city traffic engineer","campus safety director","privacy officer"],"discoveryQuestions":["What share of your footage is mostly static?","Local GPUs or remote endpoints?","What retention and camera-use policies apply?"],"valueHypothesis":"If EVS gains hold on local footage, a site may verify alerts on more cameras per GPU; this must be measured.","potentialEngagement":"Two-to-four-week EVS on/off pilot on representative cameras with privacy review.","unsupportedClaims":["46% more streams on any footage","No accuracy loss from pruning","Public-sector deployment outcomes"]},"engineering":{"fit":"Fit for on-premises or edge VLM alert verification and summarization over mostly static scenes.","architectureIntegration":"Build from validated profiles; shared Kafka/Redis/Elasticsearch; EVS inside the RT-VLM container; NemoClaw harness optional.","prerequisites":["RTX PRO-class GPUs","NGC credentials","Representative footage","Isolated network with TLS and auth"],"constraints":["Vendor-only benchmarks","EVS not available for remote endpoints","Less benefit for long outputs","Synthetic demo footage"],"security":"Deploy on trusted isolated networks with authentication, TLS and rate limiting; review generated configs for exposed ports.","proofOfValue":["EVS on/off missed-event comparison","Streams per GPU measurement","Alert latency measurement","Generated build security review"]},"delivery":{"work":"Run a bounded VSS 3.3 pilot comparing EVS on/off on real cameras with privacy controls.","dependencies":["Safety technology team","privacy/records office","network security","GPU infrastructure"],"ownership":"Public-safety or campus-safety technology lead owns operations; privacy office owns policy.","skillsAdoption":"Operators learn to review generated architecture plans and tune pruning thresholds.","governanceCheckpoints":["Architecture plan review","Privacy impact assessment","Accuracy review before expansion"],"acceptanceCriteria":["No increase in missed verified events with EVS on (proposed)","Measured streams-per-GPU gain on site footage (proposed)"],"risks":["Surveillance and retention concerns","Lock-in to NVIDIA models","Over-trusting prompt-built stacks"]}},"retrievedAt":"2026-10-05T03:05:08Z","enrichedAt":"2026-10-05T03:13:11Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Natural-language search and reports could help non-specialist operators review video, but only with training on verifying AI-generated summaries.","procurementImplications":"Price pilots on measured streams per GPU from site footage rather than vendor percentages.","operatingModelImplications":"Prompt-built deployments still need configuration ownership and change control.","sourceVerification":{"openedUrl":"https://developer.nvidia.com/blog/lower-the-cost-of-building-and-running-visual-ai-agents-with-nvidia-vss-blueprint-3-3/","referenceExcerpt":"Results vary with scene motion, chunk length, and similarity threshold; benchmark representative footage before choosing production defaults.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}