VIOS Video I/O and Storage Manager
Manage VIOS sensors, streams, recordings, snapshots, clips, and storage through its REST API.
The document discloses API boundaries, credential variables, token non-disclosure, health checks, stop-or-ask behavior, and distinct cleanup paths for file and RTSP sensors. However, deployment involves Docker commands and a Docker socket; confirmation boundaries, privilege isolation, rollback, and sensitive-data-flow details are incomplete, so points are deducted.
Core workflows include health probes, identifier and timeline resolution, known-error handling, and direct binary endpoints. However, endpoint resolution, deployment modes, and multiple reference contracts create substantial dependency complexity, with no executable CI or actual key-path test results shown; the static cap keeps this below 10.
The intended use cases, triggers, and exclusions for VLM, search, summarization, and analytics are clearly stated. Input/output contracts, non-fit environments, Chinese-language interaction, and network reachability are less explicit, and the skill depends on a local or pre-deployed VSS environment, so points are deducted.
The skill provides purpose, prerequisites, layered references, version 3.2.0, an Apache-2.0 license, examples, limitations, and troubleshooting. However, BENCHMARK identifies missing metadata.author, and there is no clear maintenance owner, changelog, stable-parameter policy, or update path; deeply nested references also increase usage cost.
It directly covers sensor, stream, upload, timeline, snapshot, and clip operations, and the examples and evaluation specifications describe useful API workflows. Static evidence does not verify representative usable outputs, while deployment prerequisites and operational cost are significant, so the score is conservatively limited.
Detailed REST contracts, error codes, evaluation checks, and a benchmark report provide some auditability. However, there are no CI execution records, real captured outputs, or independent corroboration; the benchmark claims cannot be independently reproduced from the supplied static files, so points are deducted.
- Deployment can execute Docker and host-level operations and, in some modes, mounts the Docker socket; authorization, least privilege, and recoverable rollback should be established first.
- Uploads, sensor registration, and deletion create external state changes; the document lacks a uniform per-operation confirmation and cleanup-confirmation policy.
- The skill depends on local VSS/VIOS, Docker, curl, jq, NGC credentials, and GPU/storage configuration; mainland-China reachability and Chinese-language interaction are not demonstrated.
- The benchmark report and checklists are not independent execution evidence for this static review.
What does this skill do, and when should you use it?
This skill handles VIOS and NvStreamer operations for NVIDIA Video Search and Summarization deployments. It manages cameras and sensors, RTSP streams, recording timelines, uploads, snapshots, and clip extraction. It assumes that VIOS is already deployed and reachable, and uses curl-based REST API calls. It does not perform VLM inference, semantic archive search, or video summarization.
Probes VIOS health; lists, inspects, adds, and deletes sensors and streams; checks stream status and recording timelines; uploads and deletes video files; retrieves live or historical snapshots; downloads clips; calls NvStreamer when a synthetic RTSP feed is explicitly required; and resolves sensorId and streamId from sensor names or IP addresses. Responses may be JSON, booleans, null, or locally saved binary files.
- A VSS operator needs to list active cameras and inspect their stream status.
- An integration tester needs to upload an MP4 file and obtain its generated VIOS stream ID.
- A video analyst needs to download a short clip from a recorded timeline.
- An operator needs a snapshot from a live stream or historical replay.
- A test engineer needs to serve a local sample video as a synthetic RTSP camera through NvStreamer.
What are this skill's strengths and limitations?
- Covers the main VIOS surfaces for sensors, streams, recordings, storage, snapshots, and clip extraction.
- Ships with reference contracts for VIOS, NvStreamer, integration, and deployment.
- Clearly separates direct VIOS file uploads from NvStreamer-based synthetic RTSP workflows.
- Documents health checks, identifier resolution, authentication retry, timeline handling, and common failure modes.
- Requires a deployed and reachable VIOS/VST backend; it is not a complete deployment tool.
- Requires shell access, networking, curl, jq, and Docker; NGC credentials may be needed for image pulls.
- Uploads can fail when libav is missing, and deployment can be affected by leftover containers or volume drift.
- Some /url responses have a duplicated http:// defect in version 3.2.0; the source provides no standalone test suite or platform support matrix.
How do you install this skill?
Install the specific skill with the NVIDIA skills CLI: npx skills add nvidia/skills --skill vss-manage-video-io-storage --yes. You may target a client explicitly, for example: npx skills add nvidia/skills --skill vss-manage-video-io-storage --agent codex. The README does not document additional client-specific installation steps for this skill.
How do you use this skill?
After the skill is loaded, make a concrete request such as “List the active VIOS sensors and show their stream status” or “Upload this sample video to VIOS and return the generated stream id.” A reachable VST endpoint and running VIOS deployment are required. The Agent first performs the sensor/version health check, then calls the VIOS REST API. Plain MP4 uploads use the direct VIOS file-upload endpoint; NvStreamer is reserved for explicit synthetic RTSP, NvStreamer, or generated-RTSP-URL requests. When time ranges are omitted, the Agent must read the stream timeline before selecting timestamps.
How does this skill compare with similar options?
Unlike vss-ask-video, vss-search-archive, and vss-summarize-video, this skill focuses on VIOS video input/output and storage rather than VLM questions, semantic search, or narrative summaries. Compared with using NvStreamer alone, it also manages VIOS sensors, recordings, and storage; NvStreamer is used only for explicitly requested synthetic RTSP workflows.