NVIDIA AI Blueprint for Video Search and Summarization (VSS) is a reference architecture for applications that use AI to search, analyze, question, and summarize live or recorded video. It is intended for building video analytics agents with vision-language models, language models, retrieval-augmented generation, and NVIDIA NIM microservices. The catalogue entries provide coding-agent rules, skills, commands, agents, instructions, and settings associated with the project.
Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarizationnpx agentmods add skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarizationWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarization)<a href="https://agentmods.dev/skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarization"><img src="https://agentmods.dev/badge/skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarization/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarization"><img src="https://agentmods.dev/badge/skills/nvidia-ai-blueprints/video-search-and-summarization/vss-benchmark-video-summarization.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00044 | $0.05313 |
| Opus 5.5 | $0.00018 | $0.02125 |
| Sonnet 5.5 | $0.00009 | $0.01063 |
| Haiku 4.5 | $0.00004 | $0.00531 |
Grade C, and why
vss-benchmark-video-summarization scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 17d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl -s "http://localhost:8888/videos/" | python3 -m json.tool 2>/dev/null || \ Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
| LVS deployed by you via `vss-build-vision-ai` (lvs profile) with a unique `COMPOSE_PROJECT_NAME` | `curl -sf ${LVS_BACKEND}/v1/ready` returns 200 | How it starts
The opening of the file, as written. The whole thing — 404 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Instructions
Follow the routing table and step-by-step workflow below. Execute each step in order on a first run. For repeat runs, go directly to the Repeat Runs section. Detailed reference material lives in references/ and benchmark scripts live in scripts/.
⚠️ Shared GPU systems: On a multi-user host, multiple LVS instances may be running on different ports. Always benchmark against the instance YOU deployed. Never assume port 38111 — always use the exact endpoint URL returned by
vss-build-vision-aiwhen you deployed LVS. If you did not deploy an LVS instance in this session, deploy one first before running benchmarks.
Purpose
Measure the latency and throughput limits of a deployed LVS (Long Video Summarization) instance, identify GPU and pipeline bottlenecks, and suggest configuration changes that improve performance. Produces XLSX reports and JSON result files per scenario, plus a natural-language analysis with improvement recommendations.
Do NOT use this skill for:
- Deploying LVS — use the
vss-build-vision-aiskill with thelvsprofile. - Production monitoring or alerting — this is an offline benchmarking tool.
- Non-LVS VSS profiles (RTVI, search-only, etc.) — the
/filesAPI and/summarizeendpoint are LVS-specific.
When to Use
- Measure single-file latency or burst throughput of a deployed LVS instance
- Identify whether the VLM or LLM stage is the pipeline bottleneck
- Get concrete config recommendations (chunk size, frame count, resolution, concurrency) after a benchmark run
- Compare performance across hardware profiles or config changes on the same LVS deployment
- Re-run a benchmark scenario after setup is already complete (see Repeat Runs)
Routing
| Situation | Action |
|---|---|
| User has not deployed LVS in this session | Deploy LVS with vss-build-vision-ai (lvs profile) using a unique COMPOSE_PROJECT_NAME — note the exact endpoint URL it returns, then return here |
LVS_BACKEND is not set |
Ask the user for the endpoint URL of the LVS instance they deployed — do not guess or default |
| Results directory already exists with data and user asks to analyze only | Skip to Step 6: Analyze Results |
| First run or new hardware configuration | Run the full workflow Steps 0–7 |
| Repeat run, setup already complete | Use run_benchmark.sh (see Repeat Runs section) |
What ships with it
20 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- .gitignore 188 B
- references/analyzing-results.md 9.9 KB
- references/benchmark-modes.md 5.7 KB
- scripts/base.py 36 KB runs code
- scripts/config.yaml 7.3 KB
- scripts/configmaps/nginx.conf 243 B
- scripts/cpu_monitor.py 9.0 KB runs code
- scripts/defaults.yaml 1.2 KB
- scripts/fetch-videos.sh 4.8 KB runs code
- scripts/file_burst_benchmark.py 35 KB runs code
- scripts/gpu_monitor.py 18 KB runs code
- scripts/latency_tracker.py 7.7 KB runs code
- scripts/media-server.yaml 1.0 KB
- scripts/preflight.sh 9.0 KB runs code
- scripts/requirements.txt 908 B
- scripts/run_benchmark.sh 5.8 KB runs code
- scripts/single_file_benchmark.py 33 KB runs code
- scripts/summarize_results.py 11 KB runs code
- scripts/vss_perf_benchmark.py 20 KB runs code
- skill-card.md 3.3 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 17d ago Changed fb208c0bdb48
- 18d ago Changed · +8 lines 21c8f1834d4e
- 23d ago Changed · +41 lines 4218443d2859
- 27d ago First seen · 355 lines · 44 tokens per session scan C 2159daa0d130
vss-benchmark-video-summarization is a skill published in the GitHub repository NVIDIA-AI-Blueprints/video-search-and-summarization (1,932 stars, last pushed today), licensed Apache-2.0. It adds 44 tokens to every session and 5,313 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-20.
Other skills, from other repositories
agent-platform-rag-engine-management
Manage and query Agent Platform RAG Engine Corpora and retrieve grounded contexts using the Google GenAI SDK. Use when listing RAG corpora or files, inspecting a corpus, retrieving contexts, or generating content grounded in a RAG corpus. Do not use for standard database queries (use SQL/Spanner skills), Google…
graphify
Use for any question about a codebase, its architecture, file relationships, or project content — especially when graphify-out/ exists, where the question should be treated as a graphify query first. Turns any input (code, docs, papers, images, videos) into a persistent knowledge graph with god nodes, community…
google-cloud-solution-rag-enterprise-search-gke-sqldb
Discovers requirements, and generates architectural, design, and deployment guidance for a retrieval-augmented generation (RAG)-capable enterprise search system in Google Cloud. Use when users need a vector-enabled SQL database as the store and index for the embedding vectors, an open model and open-source inferencing…
gpt-image-2
A skill for generating or editing images with GPT Image 2 across local, host-provided, or advisory setups.
similarity-search-patterns
Implement efficient similarity search with vector databases. Use when building semantic search, implementing nearest neighbor queries, or optimizing retrieval performance.
kb-retriever
A retrieval and question-answering assistant for a local folder of documents, including Markdown, text, PDFs, and spreadsheets.