菜单
Build the DS-SOP Docker image and deploy the VSS SOP blueprint end-to-end. Use when asked to deploy SOP, build DS SOP, install SOP, set up the SOP pipeline, verify SOP models, start the SOP blueprint, simulate RTSP for SOP, run the SOP API test, or tear down SOP.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/NVIDIA/sop-monitoring-blueprints --skill vss-sop-deploy命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Autonomous end-to-end orchestrator for SOP fine-tuning. Runs the full Import → Augment → DDM Train → VLM Train → Evaluate → RCA loop. Interprets RCA findings across DDM, VLM and augment axes, applies config fixes autonomously, and iterates until success criteria are met or max_pipeline_iterations reached. Call with a path to an inputs.yaml or with natural language.
Build a custom VSS SOP blueprint from the VSS 3.1 base, then deploy and test in a loop until fully operational. Use when asked to create the SOP blueprint structure, customize VSS compose for SOP, configure SOP services, set up the VSS agent for SOP, or scaffold the SOP app layer on top of met-blueprints 3.1.
Use this skill when building, deploying, evaluating, debugging, or measuring latency for the DeepStream SOP Inference Microservice — a GPU-accelerated FastAPI service that detects whether operators perform assembly-line steps in order via event boundary detection (GEBD) plus VLM classification. Trigger even if the user does not name it: verify operator step sequence, detect missing or out-of-order SOP steps, score factory/work-cell video for procedure compliance, run VLM-based SOP checking on industrial cameras, or call /v1/chat/completions with a file, RTSP, or Basler camera. Also trigger for its internals: SOPVideoProcessor, DeepStream GEBD model (e.g. DDM) via Triton CAPI, nvds_custom_postprocess, Cosmos Reason 1/2 vLLM, SSE streaming, Kafka NvProto/JSON output, Basler/Pylon camera + emulation, Docker compose, chunk-level latency. Do NOT trigger for generic DeepStream pipelines, object detection/tracking, NIM imports, or video summarization.
基于 SOC 职业分类
正在显示 SKILL.md
| name | vss-sop-deploy |
| description | Build the DS-SOP Docker image and deploy the VSS SOP blueprint end-to-end. Use when asked to deploy SOP, build DS SOP, install SOP, set up the SOP pipeline, verify SOP models, start the SOP blueprint, simulate RTSP for SOP, run the SOP API test, or tear down SOP. |
| owner | NVIDIA |
| service | vss-sop |
| version | 1.0.0 |
| license | CC-BY-4.0 AND Apache-2.0 |
| reviewed | "2026-06-23T00:00:00.000Z" |
| metadata | {"openclaw":{"emoji":"🏗️","os":["linux"]},"author":"Quan Vu <qvu@nvidia.com>","tags":["vss","sop","deploy"]} |
Complete end-to-end skill: build the DS-SOP Docker image, verify models, configure the blueprint, deploy, test, and tear down using structured helper scripts.
Use this skill when deploying or tearing down the VSS SOP blueprint, starting RTSP simulations, or validating prerequisites.
Key operations:
/opt.| Item | Value |
|---|---|
| CUDA | 13 |
| Driver | 580 |
| Hardware | H100 / H200 / A100 |
| GPU | >= 80 GB VRAM, >= 1 GPU |
Software: Docker, NVIDIA Container Toolkit 1.18.1, NVIDIA Driver 580, ngc CLI.
# Configure blueprint environment variables
./agentic/vss-sop-skills/vss-sop-deploy/scripts/configure_blueprint.sh --bp-repo . --vlm-mode local
# Start the full deployment stack
./agentic/vss-sop-skills/vss-sop-deploy/scripts/deploy.sh --bp-repo .
# Start simulation stream on rtsp://localhost:8552/sensor_0
./agentic/vss-sop-skills/vss-sop-deploy/scripts/start_rtsp_server.sh --bp-repo .
# Tear down the deployment and clean volumes
./agentic/vss-sop-skills/vss-sop-deploy/scripts/teardown.sh --bp-repo .
Run this phase before every deploy. It checks secret key files, GPU drivers, Docker, NVIDIA Container Toolkit, and NGC configuration, and can automatically install or configure missing prerequisites.
To run the checks and automatically fix/install any missing components (such as NVIDIA Driver 580, CUDA Toolkit 13, Docker, Docker Compose, NVIDIA Container Toolkit, NGC CLI, and NGC CLI configuration):
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/preflight_check.sh --bp-repo . --fix
If you only want to run the checks without modifying your system:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/preflight_check.sh --bp-repo .
If any manual check fails, refer to the printed advice or the following documentation to resolve the issue:
references/ngc.md § Configure NGC API Keyreferences/nvidia_driver.mdreferences/prerequisites.mdVerify that the trained model and config files exist under /opt/models/... and /opt/sop/... directories. For optimal accuracy, you must retrain/fine-tune the models, which can be done using the SOP Training Blueprint. After training, move model and config to /opt/models/... and /opt/sop/... directories. Download the sample video, re-encode it to H.264 at 30 FPS, and prepare cache/datalog directories.
To execute this entire pipeline:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/download_assets.sh --bp-repo .
This automates the following actions:
/opt/models and /opt/sop. If not, prompts the user to retrain/fine-tune the models using the SOP Training Blueprint.ffmpeg (if missing) and re-encodes the sample video to H.264 at 30 FPS.The RTSP stream simulator is started in this phase — ahead of the build, verification, and deployment steps that follow — so a live stream (rtsp://localhost:8552/sensor_0) is available for the end-to-end ingestion testing in later phases.
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/start_rtsp_server.sh --bp-repo .
This installs RTSP server prerequisites and starts rtsp_server.py serving on rtsp://localhost:8552/sensor_0 in the background (which wraps raw MP4 footage and streams it in real time).
To build the ds-sop:1.0.0 Docker image and verify its RTSP components, refer to the detailed reference documentation:
This reference covers:
ds_sop_microservice) — RTSP streaming output is opt-in (ds-sop-skills § 18) and must be requested in the generation prompt ("with rtsp streaming output feature").verify_rtsp_components.py).Configure host IPs, deployment paths, and LLM/VLM endpoints in the .env file. With no extra flags this auto-detects HOST_IP/EXTERNAL_IP and defaults to remote NIM endpoints:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/configure_blueprint.sh --bp-repo .
For local VLM (served by the DS-SOP container on port 8300):
./agentic/vss-sop-skills/vss-sop-deploy/scripts/configure_blueprint.sh --bp-repo . --vlm-mode local
Run configure_blueprint.sh --help for the full set of options (--llm-mode, --vlm-mode, --llm-base-url, --vlm-base-url, --llm-name, --vlm-name).
This script also enforces use_base64: true in config.yml and SOP_MESSAGING_SCHEMA=JSON in the DS-SOP .env — both are required for correct Kibana field mapping and video understanding.
Auto-configuration safety net: If you skip this step,
deploy.shwill detect the unconfigured placeholder and runconfigure_blueprint.shautomatically with default settings before starting services.
Log into NGC registry and start the blueprint Docker compose services:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/deploy.sh --bp-repo .
Note on Volume Permissions & Automatic Index Cleanup:
deploy.shautomatically pre-creates and configures all required bind-mounted data log directories (such askafka,elastic/data,redis/data, etc.) with universal777permissions (usingsudoif needed). This completely avoids container-level "Permission denied" errors.In addition,
deploy.shautomatically cleans up old Elasticsearch (elastic/dataandelastic/logs) and Kafka data from prior runs before launching the containers. This prevents index pattern mapping pollution (such as stale1970-01-01date indices with old protobuf schemas) from interfering with wildcard queries or breaking Kibana visualization fields.
docker ps --format 'table {{.Names}}\t{{.Status}}\t{{.Ports}}'
Deploy is complete when all mdx-* containers show status Up.
| Service | Endpoint |
|---|---|
| VSS-UI | http://<EXTERNAL_IP>:3000 |
| Grafana-UI | http://<EXTERNAL_IP>:35000/ |
| Kibana-UI | http://<EXTERNAL_IP>:5601/app/home#/ |
| VIOS-UI | http://<EXTERNAL_IP>:30888/vst/#/dashboard |
| Phoenix-UI (Telemetry) | http://<EXTERNAL_IP>:6006/projects |
With the RTSP streaming simulator already running, trigger the test client and monitor the integration:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/test_rtsp.sh --bp-repo .
This automates:
mdx-ds-sop-1 logs until the API server is fully up../run_rtsp_test.sh in the background to register streams to VIOS.Self-Healing Integration Loop: To completely avoid the stale RTSP factory bug (where subsequent client runs on
mdx-ds-sop-1allocate random UDP ports but retain the old RTSP factory on port 8554, yielding 0 RTP bytes),run_rtsp_test.shruns a continuous self-healing loop. If the ds-sop client exits or the connection is lost, the script automatically restarts themdx-ds-sop-1container, waits for readiness, resets the VST sensors, and starts a fresh streaming session to resume recording automatically. This ensures continuous livestreaming and recording stability.
To stop the services and clean up all data logs:
cd <bp-repo>
./agentic/vss-sop-skills/vss-sop-deploy/scripts/teardown.sh --bp-repo .
deploy.sh, teardown.sh, test_rtsp.sh, build_ds_sop.sh, and vss_sop_test.py) are fully docker-group aware. If the current shell has inactive docker group membership but the user belongs to the docker group, the scripts will automatically re-execute themselves under the sg docker command wrapper to bypass socket permission errors. If sudo is required, they will prompt/re-run via sudo. You do not need to prefix commands with sg docker manually.unknown or invalid runtime name: nvidia — NVIDIA Container Toolkit not installed or Docker not restarted. See references/prerequisites.md.ngc_api_key.txt and ensure it is exported correctly.mdx-ds-sop-1 not starting — check model paths match between Phase 1 and ds-sop/.env.docker logs mdx-ds-sop-1 for errors or progress.MediaInfo.parse: Unsupported file type error in mdx-ds-sop-1 when downloading video from VST directly — Ensure that use_base64: true is set under video_understanding in deployments/sop/vss-agent/configs/config.yml. This forces the VSS Agent to send base64-encoded video instead of VST links, avoiding VST download parsing errors.SOP_MESSAGING_SCHEMA=JSON and ENABLE_MESSAGING=1 are set in deployments/ds/ds-sop/.env (automatically enforced by configure_blueprint.sh). See vss-sop-build SKILL.md § Error Handling for the full field name reference and validation steps.Failed to load image with src: http://<internal-ip>:30888/... — EXTERNAL_IP in .env is a host-only address. Set it to the public/browser-reachable IP (see Phase 4) and recreate the agent container.Use of this skill is governed by the Creative Commons Attribution 4.0 International License (CC BY 4.0) and the Apache License, Version 2.0.