skills.rest
NVIDIA
NVIDIAOFFICIAL
Explore all skills by this author →
WebsiteGitHub

rtvi-cv-scaffold-vss-service

Scaffold a DeepStream YOLO26 microservice that publishes detection metadata to VSS via Kafka mdx-raw.

3.2k|370|Updated Feb 25, 2026
Search Tags:#perception#Kafka#microservice#yolo26#vss#DeepStream#Scaffold
One-click install
npx skills add https://github.com/NVIDIA/skills --skill rtvi-cv-scaffold-vss-service
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: rtvi-cv-scaffold-vss-service
Source: https://github.com/NVIDIA/skills/tree/main/skills/rtvi-cv-scaffold-vss-service
Command: npx skills add https://github.com/NVIDIA/skills --skill rtvi-cv-scaffold-vss-service

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Building a custom perception microservice that integrates with NVIDIA VSS (Video Search and Summarization) requires getting many integration details exactly right: the Kafka mdx-raw topic, protobuf payload type 2, nvmsgconv/nvmsgbroker wiring, host networking, and VSS Search/Alerts profile flags. This Skill generates a runnable, contract-compliant microservice repo instead of leaving engineers to assemble it by hand.

Core Features & Use Cases

  • Scaffold generation: Produces a complete repo with Dockerfile, DeepStream pipeline configs (YOLO26 pgie, NvDCF tracker, msgconv/msgbroker), Compose file with VSS profile gates, Python adapter, unit tests, and a Kafka smoke-test tool.
  • Fixed VSS integration contract: Encodes the non-negotiable elements (mdx-raw topic, msg-conv-payload-type=2, libnvds_kafka_proto.so, host networking) so downstream Search, Alerts, and Behavior Analytics consume the output without changes.
  • Segmentation extension guidance: Documents the msgconv wrapper pattern for adding frame-mask payloads while preserving the base object contract.
  • Use Case: A customer wants to replace VSS's default RT-DETR perception with their own YOLO26 detector on their hardware. The Skill scaffolds the service, runs host-only unit tests, and explains exactly which customer assets (ONNX, labels, parser .so) are required before live deployment.

Quick Start

Ask your agent to scaffold a custom RTVI CV microservice that runs YOLO26 in DeepStream and publishes object metadata to Kafka so VSS Search and Alerts profiles can consume it.

Frequently Asked Questions about rtvi-cv-scaffold-vss-service

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build a custom perception microservice for NVIDIA VSS?▼

Run the scaffold script with a service name, output directory, class count, and target VSS profiles. It generates a DeepStream pipeline with YOLO26 inference, NvDCF tracking, and Kafka publication to mdx-raw, plus Compose files gated by the bp_developer_search_2d and bp_developer_alerts_2d_cv profiles.

How do I integrate a custom DeepStream detector with VSS Search and Alerts?▼

Publish protobuf payload type 2 to the mdx-raw Kafka topic using nvmsgconv and nvmsgbroker with libnvds_kafka_proto.so, over host networking. The generated service starts as a separate Compose application after the VSS kafka-topic-init-container completes.

Can the live VSS integration run without a trained YOLO26 ONNX model?▼

No. Scaffolding and host-only unit tests work with placeholders, but the live DeepStream app requires the customer-supplied ONNX export, a labels file matching num-detected-classes, and a compiled parser library exposing NvDsInferParseYolo26.

Why must the service publish to the mdx-raw Kafka topic?▼

VSS downstream consumers including vss-search-analytics, vss-behavior-analytics, and vss-video-analytics-api are pinned to mdx-raw. Publishing to any other topic silently produces an empty search index and no behavior analytics output.

When should I use rtvi-cv-customize-model instead of this scaffold?▼

Use rtvi-cv-customize-model when you only want to swap the detector inside the stock vss-rt-cv perception container. Use this scaffold when you need a new standalone microservice repo with its own pipeline, tests, and deployment lifecycle.

Does the Kafka smoke test validate the protobuf payload schema?▼

No. The kafka_smoketest.py tool only confirms a non-empty message arrived on mdx-raw within the timeout. To validate the nv.Frame schema, decode the payload with the nv.Frame proto and assert sensorId, objects, bbox, and tracking IDs explicitly.

Related Skills

Explore More Software Engineering Skills→
yo-stevenyo-steven

deepstream-import-vision-model

Convert HuggingFace or NVIDIA NGC vision models into DeepStream pipelines with benchmarking.

Community
Advanced
rblake2320rblake2320

vss-manage-alerts

Route VSS camera alerts to Slack and dashboard backends.

Community
Advanced
yo-stevenyo-steven

alerts

Execute mode-specific VSS alert workflows for CV verification and VLM real-time monitoring.

Community
Advanced
NVIDIANVIDIA
49

vss-sop-build

Build and validate a customized VSS SOP monitoring blueprint from VSS 3.1.

Official
Advanced