[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71QB71YVCKW3YBJ23THHB2Z":3,"skill-same:01M71QB71YVCKW3YBJ23THHB2Z":123},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":29,"publishedAt":50,"linkOnly":51,"markdownContent":52,"markdownBody":53,"frontMatter":54,"refName":59,"commitSha":60,"skillDirectory":61,"forksCount":14,"repositoryHtmlUrl":62,"repositoryHomepageUrl":63,"ownerHtmlUrl":64,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":65,"install":66,"resources":79},"01M71QB71YVCKW3YBJ23THHB2Z","vss-deploy-dense-captioning",null,"Use this skill when deploying standalone RT-VLM dense captioning or calling its REST API (uploads, captions, streams, chat-completions, Kafka). Not for VSS profile deploy or video-search ingestion.","pattanation\u002FNorvik","pattanation","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F235419268?v=4","01M71M5THHYYB15ME8R2E4FCZG","Patta","01M71M5TJM1NDM992M8PDKYR2K",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002FSKILL.md",[18,23,26],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署",{"categoryNo":27,"categoryName":28,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体",[30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49],"captioning","deploy","vss deploy dense captioning","chat","uploads","api","when","Blueprint","use","calling","streams","search","deployment","for","NVIDIA","Captions","kafka","standalone","ingestion","Video","2026-11-10 13:02:02",false,"---\nname: vss-deploy-dense-captioning\ndescription: Use this skill when deploying standalone RT-VLM dense captioning or calling its REST API (uploads, captions, streams, chat-completions, Kafka). Not for VSS profile deploy or video-search ingestion.\nlicense: Apache-2.0\nmetadata:\n  version: \"3.2.0\"\n  github-url: \"https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-Blueprints\u002Fvideo-search-and-summarization\"\n  tags: \"nvidia blueprint operational deployment\"\n---\n## Purpose\n\nStand up the RT-VLM dense-captioning microservice on its own and exercise every endpoint it exposes (file upload, generate_captions, stream add\u002Fdelete, chat-completions, Kafka topics).\n\n## Prerequisites\n\nFor standalone RT-VLM deployment:\n- Docker, Docker Compose, NVIDIA Container Toolkit, and a visible GPU.\n- NGC registry credentials in `$NGC_CLI_API_KEY` for `docker login nvcr.io`,\n  image pulls, and local NGC model\u002Fartifact downloads.\n- `curl`, `jq`, and any writable working directory for the standalone compose copy.\n\nFor API calls against an existing service:\n- Running RT-VLM service reachable at `$BASE_URL`.\n- Bearer token in `$RTVI_VLM_API_KEY` or `$NGC_CLI_API_KEY`, depending on how the\n  service was configured.\n\nFor full VSS profile deployment:\n- Use `..\u002Fvss-deploy-profile\u002FSKILL.md`; this skill does not deploy full VSS profiles.\n\n## Instructions\n\nFollow the routing tables and step-by-step workflows below. Each section that ends in *workflow*, *quick start*, or *flow* is intended to be executed top-to-bottom. Detailed reference material lives in `references\u002F`; execute the documented workflows directly unless a future revision names a concrete helper.\n\n## Examples\n\nWorked end-to-end examples are kept under `evals\u002F` (each `*.json` manifest contains a runnable scenario) and inline in the per-workflow `curl` blocks below. Run a Tier-3 evaluation with `nv-base validate \u003Cthis-skill-dir> --agent-eval` to replay them.\n\n## Limitations\n\n- Requires either a standalone RT-VLM service deployed via this skill or an\n  existing RT-VLM service reachable from the caller.\n- NGC-hosted models and NIMs may be subject to rate-limits, GPU memory requirements, and license restrictions.\n- Concurrency, GPU memory, and storage limits depend on the host hardware and the profile's compose file.\n- Keep `NGC_CLI_API_KEY`, `RTVI_VLM_API_KEY`, and `.env` files out of git and out of logs; do not echo credential values or include them in final responses.\n- Docker group access and `sudo` are effectively root-level privileges. Use the non-interactive `sudo -n` guard in the deploy reference and stop for host-owner action when passwordless sudo is unavailable.\n\n## Troubleshooting\n\n- **Error**: REST call returns connection refused. **Cause**: target microservice not running. **Solution**: probe `\u002Fdocs` or `\u002Fhealth`; redeploy via `vss-deploy-profile` or the matching `vss-deploy-*` skill.\n- **Error**: HTTP 401\u002F403 from NGC pulls. **Cause**: missing\u002Fexpired `NGC_CLI_API_KEY`. **Solution**: `docker login nvcr.io` and re-export the key before retrying.\n- **Error**: container OOM or model fails to load. **Cause**: insufficient GPU memory for the selected profile. **Solution**: switch to a smaller variant or free GPUs via `docker compose down`.\n\n# Deploy and Use RT-VLM Dense Captioning (VSS 3.2)\n\nRT-VLM is NVIDIA's real-time vision-language microservice: decode video (file or\nRTSP), segment it into chunks, run a VLM (`cosmos-reason1`, `cosmos-reason2`, or any\nOpenAI-compatible model), stream dense captions back over SSE\u002FHTTP, and publish\ncaptions, incident alerts, and errors to Kafka. Use this skill to deploy the\nstandalone RT-VLM service when a full VSS profile is not already running, then call\nits `\u002Fv1\u002F...` API for caption generation, file upload, live-stream management, health\nchecks, NIM-compatible chat completions, or Prometheus metrics. API reference:\n\u003Chttps:\u002F\u002Fdocs.nvidia.com\u002Fvss\u002Flatest\u002Freal-time-vlm-api.html>.\n\n## Deployment Routing\n\nIf the user asks to deploy a full VSS profile, use\n[`..\u002Fvss-deploy-profile\u002FSKILL.md`](..\u002Fvss-deploy-profile\u002FSKILL.md). That skill\nowns profile routing, `generated.env`, `resolved.yml`, multi-service sizing, and\nfull-stack deploy\u002Fteardown.\n\nIf the user asks for standalone RT-VLM dense captioning, or no VSS profile is\nalready running, use the standalone RT-VLM flow in\n[`references\u002Fdeploy-rt-vlm-service.md`](references\u002Fdeploy-rt-vlm-service.md)\nbefore calling the API. This follows the same compose-centric pattern as\n`vss-deploy-profile`: gather context, run preflights, work from a local copy,\ndry-run with `docker compose config`, review, deploy, then wait for health.\n\n## Standalone Deployment Flow\n\nAlways follow this sequence. Never skip the dry-run.\n\n```bash\n# 1. Copy deploy\u002Fdocker\u002Fservices\u002Frtvi\u002Frtvi-vlm\u002Frtvi-vlm-docker-compose.yml\n#    into any writable standalone working directory.\n# 2. Derive RTVI_VLM_IMAGE_TAG from that compose copy.\n# 3. Strip the standalone-only dangling depends_on block from the copy.\n# 4. Create a gitignored .env with the required RT-VLM values.\n# 5. Prepare host bind paths such as $VSS_DATA_DIR\u002Fdata_log\u002Fvst\u002Fclip_storage.\n#    Use `sudo -n` for ownership fixes; if passwordless sudo is unavailable,\n#    stop and ask the host owner to run the printed command manually.\n# 6. docker compose --env-file .env -f rtvi-vlm-docker-compose.yml config --quiet\n# 7. docker pull the exact RT-VLM image tag.\n# 8. docker compose ... up -d rtvi-vlm, wait for ready, then smoke test.\n```\n\nRun preflights before any pull or `up`; stop and fix failures here before\ndebugging RT-VLM itself:\n\n```bash\nnvidia-smi --query-gpu=index,name --format=csv,noheader\nnvidia-container-cli info\ndocker compose version\ndocker run --rm --gpus all nvidia\u002Fcuda:12.4.0-base-ubuntu22.04 nvidia-smi\n```\n\nFor standalone single-file deployments, do not run the raw\n`deploy\u002Fdocker\u002Fservices\u002Frtvi\u002Frtvi-vlm\u002Frtvi-vlm-docker-compose.yml` directly: it\ncontains `depends_on` references to sibling VLM\u002FNIM services that are only\ndefined in the full VSS\u002Fmet-blueprints compose project. The standalone reference\nshows how to copy the compose file, derive the current image tag from it, strip\nthe `depends_on` block, and validate the result before `up`.\n\nFor agent-driven validation, never let `sudo` prompt interactively. Before any\nprivileged ownership or Docker operation, use the non-interactive guard in\n[`references\u002Fdeploy-rt-vlm-service.md`](references\u002Fdeploy-rt-vlm-service.md):\nprefer plain `docker`; otherwise use `sudo -n docker`; if `sudo -n` fails, stop\nwith the exact manual command for the host owner instead of retrying with\ninteractive sudo or weakening permissions.\n\nIf `docker pull` fails with a containerd snapshotter\u002Funpack error on Docker 28+,\napply the `\u002Fetc\u002Fdocker\u002Fdaemon.json` `containerd-snapshotter=false` fix in the\nstandalone reference before retrying.\n\nMinimum standalone `.env` values:\n\n| Host env var | Required when | Purpose |\n|---|---|---|\n| `NGC_CLI_API_KEY` | Standalone deploy path | NGC registry image pull and NGC model\u002Fartifact download |\n| `RTVI_VLM_API_KEY` or `NGC_CLI_API_KEY` | Authenticated API calls | RT-VLM bearer auth after the service is running |\n| `RTVI_VLM_PORT` | Always | Host API port mapped to container `8000` |\n| `HOST_IP` | Always | Kafka bootstrap host (`${HOST_IP}:9092`) |\n| `VSS_DATA_DIR` | Always | Required clip-storage bind mount |\n| `RTVI_VLM_MODEL_TO_USE` | Always for standalone | Backend selector; use `cosmos-reason2` for the default local model or `openai-compat` for a remote\u002Fsibling endpoint |\n| `RTVI_VLM_MODEL_PATH` | Local self-hosted model | Source-backed Cosmos Reason 2 path: `ngc:nim\u002Fnvidia\u002Fcosmos-reason2-8b:hf-1208` |\n| `RTVI_VLM_ENDPOINT` | `RTVI_VLM_MODEL_TO_USE=openai-compat` | Remote\u002Fsibling OpenAI-compatible VLM endpoint |\n| `VLM_NAME` | `RTVI_VLM_MODEL_TO_USE=openai-compat` | Model\u002Fdeployment name exposed by that endpoint |\n\n## Setup\n\n```bash\nexport BASE_URL=\"http:\u002F\u002Flocalhost:${RTVI_VLM_PORT:-8018}\"  # host-side RT-VLM port\nexport API_KEY=\"${NGC_CLI_API_KEY:-${RTVI_VLM_API_KEY:-}}\" # bearer token used by host-side curl commands\n: \"${API_KEY:?Set NGC_CLI_API_KEY or RTVI_VLM_API_KEY before calling authenticated endpoints}\"\n```\n\nEvery request below uses `Authorization: Bearer $API_KEY`. Health endpoints\n(`\u002Fv1\u002Fhealth\u002F*`, `\u002Fv1\u002Fready`, `\u002Fv1\u002Flive`, `\u002Fv1\u002Fstartup`) typically work without auth.\n\n**Smoke test before use:**\n```bash\ncurl -fsS \"$BASE_URL\u002Fv1\u002Fhealth\u002Fready\"\nMODEL_ID=\"$(curl -fsS \"$BASE_URL\u002Fv1\u002Fmodels\" -H \"Authorization: Bearer $API_KEY\" | jq -r '.data[0].id \u002F\u002F .id')\"\ncurl -fsS \"$BASE_URL\u002Fopenapi.json\" | jq -r '.paths | keys[]' | sort\n```\n\n## RTSP Sample Stream Guard\n\nWhen a task or eval names `RTSP_SAMPLE_URL`, treat that exact environment\nvariable as a required input. Verify it is set and non-empty before probing or\nregistering any stream; if it is missing, stop with a clear failure message. Do\nnot derive a substitute from NvStreamer, VIOS, sample-data bundles, or any other\nfallback, because that validates a different stream than the caller requested.\n\n```bash\n: \"${RTSP_SAMPLE_URL:?Set RTSP_SAMPLE_URL to a reachable RTSP sample stream before RTSP validation}\"\ncase \"$RTSP_SAMPLE_URL\" in\n  rtsp:\u002F\u002F*) ;;\n  *) echo \"RTSP_SAMPLE_URL must be an rtsp:\u002F\u002F URL, got: $RTSP_SAMPLE_URL\" >&2; exit 1 ;;\nesac\n\nif command -v ffprobe >\u002Fdev\u002Fnull 2>&1; then\n  ffprobe -v error -rtsp_transport tcp \\\n    -select_streams v:0 -show_entries stream=codec_type \\\n    -of csv=p=0 \"$RTSP_SAMPLE_URL\" | grep -qx video\nelif command -v gst-discoverer-1.0 >\u002Fdev\u002Fnull 2>&1; then\n  gst-discoverer-1.0 \"$RTSP_SAMPLE_URL\" | grep -qi 'video'\nelse\n  echo \"Install ffprobe or gst-discoverer-1.0 before RTSP validation.\" >&2\n  exit 1\nfi\n```\n\n## Quick Start — dense captions from a local video\n\n```bash\n# 1. Upload the video, capture its file id\nFILE_ID=$(curl -fsS -X POST \"$BASE_URL\u002Fv1\u002Ffiles\" \\\n  -H \"Authorization: Bearer $API_KEY\" \\\n  -F \"file=@\u002Fpath\u002Fto\u002Fwarehouse.mp4\" \\\n  -F \"purpose=vision\" \\\n  -F \"media_type=video\" | jq -r '.id')\n\n# 2. Generate captions + alerts (SSE stream of chunked responses)\ncurl -N -X POST \"$BASE_URL\u002Fv1\u002Fgenerate_captions\" \\\n  -H \"Authorization: Bearer $API_KEY\" \\\n  -H \"Content-Type: application\u002Fjson\" \\\n  -d \"{\n    \\\"id\\\": \\\"$FILE_ID\\\",\n    \\\"prompt\\\": \\\"Write a concise dense caption for each 10-second segment of this warehouse video.\\\",\n    \\\"model\\\": \\\"$MODEL_ID\\\",\n    \\\"chunk_duration\\\": 10,\n    \\\"stream\\\": true\n  }\"\n```\n\n## API Surface\n\nUse the live OpenAPI as the source of truth before calling optional endpoints:\n\n```bash\ncurl -fsS \"$BASE_URL\u002Fopenapi.json\" | jq -r '.paths | keys[]' | sort\n```\n\nCore paths for VSS 3.2 are:\n\n- `POST \u002Fv1\u002Ffiles` for multipart media upload; pass the returned file `id` into\n  caption generation and delete the file when finished.\n- `POST \u002Fv1\u002Fgenerate_captions` for file or stream captioning. Use the exact\n  model id returned by `GET \u002Fv1\u002Fmodels`; aliases such as `cosmos-reason2` are\n  backend selectors, not request model ids.\n- `POST \u002Fv1\u002Fstreams\u002Fadd`, `GET \u002Fv1\u002Fstreams\u002Fget-stream-info`, and\n  `DELETE \u002Fv1\u002Fstreams\u002Fdelete\u002F{stream_id}` for RTSP lifecycle. Parse stream ids\n  from `results[0].id`.\n- `POST \u002Fv1\u002Fchat\u002Fcompletions` for OpenAI-compatible text and multimodal calls.\n  Current 26.05 builds return HTTP 400 for text-only `\u002Fv1\u002Fcompletions`; treat\n  that as expected when validating legacy behavior.\n- `GET \u002Fv1\u002Fhealth\u002Fready`, `\u002Fv1\u002Fmodels`, `\u002Fv1\u002Fassets\u002Fstats`, and `\u002Fv1\u002Fmetrics`\n  for service probes. Do not assume `\u002Fv1\u002Flicense` exists unless OpenAPI lists it.\n\nDetailed endpoint schemas, response shapes, CV-style singular stream endpoints,\nand 26.05 compatibility notes live in\n[`references\u002Fapi-surface-26.05.md`](references\u002Fapi-surface-26.05.md).\n\n## Common Workflows\n\n- Stored file captioning: upload with `POST \u002Fv1\u002Ffiles`, call\n  `\u002Fv1\u002Fgenerate_captions` with the returned file id, use `stream=true` for SSE,\n  then delete the file to release storage.\n- RTSP live captioning: when the caller provides `RTSP_SAMPLE_URL`, use that\n  exact URL and run the **RTSP Sample Stream Guard** before registration. Do not\n  derive a replacement stream from NvStreamer or VIOS when `RTSP_SAMPLE_URL` is\n  empty; fail fast instead. Require an actual video stream\u002Fcaps entry before\n  registration; add the stream, caption it, then unregister it.\n- Alert prompts: include a deterministic `Anomaly Detected: Yes\u002FNo` line.\n  Kafka publication is server-side config, additive to HTTP responses, and\n  documented in [`references\u002Fkafka-workflows.md`](references\u002Fkafka-workflows.md).\n- Kafka validation: trust the live `vss-rtvi-vlm` environment for topic names.\n  In a full VSS alerts real-time profile, use the existing VSS Kafka container\n  `mdx-kafka` for CLI checks and final incident-consumer commands. For\n  standalone validation, use a broker that advertises `${HOST_IP}:9092`; never\n  stop or replace a pre-existing broker without user confirmation.\n\n## Error Reference\n\nCommon causes: 400 for invalid request shape or model id, 401\u002F403 for missing\nor wrong bearer token, 404 for deleted files\u002Fstreams or unsupported endpoints,\n413 for oversized uploads, 422 for schema validation, 429 for too much\nconcurrency, 500 for inference\u002Fruntime failures, and 503 while startup is still\nin progress. Inspect `docker logs vss-rtvi-vlm` for service-side failures.\n","## Purpose\n\nStand up the RT-VLM dense-captioning microservice on its own and exercise every endpoint it exposes (file upload, generate_captions, stream add\u002Fdelete, chat-completions, Kafka topics).\n\n## Prerequisites\n\nFor standalone RT-VLM deployment:\n- Docker, Docker Compose, NVIDIA Container Toolkit, and a visible GPU.\n- NGC registry credentials in `$NGC_CLI_API_KEY` for `docker login nvcr.io`,\n  image pulls, and local NGC model\u002Fartifact downloads.\n- `curl`, `jq`, and any writable working directory for the standalone compose copy.\n\nFor API calls against an existing service:\n- Running RT-VLM service reachable at `$BASE_URL`.\n- Bearer token in `$RTVI_VLM_API_KEY` or `$NGC_CLI_API_KEY`, depending on how the\n  service was configured.\n\nFor full VSS profile deployment:\n- Use `..\u002Fvss-deploy-profile\u002FSKILL.md`; this skill does not deploy full VSS profiles.\n\n## Instructions\n\nFollow the routing tables and step-by-step workflows below. Each section that ends in *workflow*, *quick start*, or *flow* is intended to be executed top-to-bottom. Detailed reference material lives in `references\u002F`; execute the documented workflows directly unless a future revision names a concrete helper.\n\n## Examples\n\nWorked end-to-end examples are kept under `evals\u002F` (each `*.json` manifest contains a runnable scenario) and inline in the per-workflow `curl` blocks below. Run a Tier-3 evaluation with `nv-base validate \u003Cthis-skill-dir> --agent-eval` to replay them.\n\n## Limitations\n\n- Requires either a standalone RT-VLM service deployed via this skill or an\n  existing RT-VLM service reachable from the caller.\n- NGC-hosted models and NIMs may be subject to rate-limits, GPU memory requirements, and license restrictions.\n- Concurrency, GPU memory, and storage limits depend on the host hardware and the profile's compose file.\n- Keep `NGC_CLI_API_KEY`, `RTVI_VLM_API_KEY`, and `.env` files out of git and out of logs; do not echo credential values or include them in final responses.\n- Docker group access and `sudo` are effectively root-level privileges. Use the non-interactive `sudo -n` guard in the deploy reference and stop for host-owner action when passwordless sudo is unavailable.\n\n## Troubleshooting\n\n- **Error**: REST call returns connection refused. **Cause**: target microservice not running. **Solution**: probe `\u002Fdocs` or `\u002Fhealth`; redeploy via `vss-deploy-profile` or the matching `vss-deploy-*` skill.\n- **Error**: HTTP 401\u002F403 from NGC pulls. **Cause**: missing\u002Fexpired `NGC_CLI_API_KEY`. **Solution**: `docker login nvcr.io` and re-export the key before retrying.\n- **Error**: container OOM or model fails to load. **Cause**: insufficient GPU memory for the selected profile. **Solution**: switch to a smaller variant or free GPUs via `docker compose down`.\n\n# Deploy and Use RT-VLM Dense Captioning (VSS 3.2)\n\nRT-VLM is NVIDIA's real-time vision-language microservice: decode video (file or\nRTSP), segment it into chunks, run a VLM (`cosmos-reason1`, `cosmos-reason2`, or any\nOpenAI-compatible model), stream dense captions back over SSE\u002FHTTP, and publish\ncaptions, incident alerts, and errors to Kafka. Use this skill to deploy the\nstandalone RT-VLM service when a full VSS profile is not already running, then call\nits `\u002Fv1\u002F...` API for caption generation, file upload, live-stream management, health\nchecks, NIM-compatible chat completions, or Prometheus metrics. API reference:\n\u003Chttps:\u002F\u002Fdocs.nvidia.com\u002Fvss\u002Flatest\u002Freal-time-vlm-api.html>.\n\n## Deployment Routing\n\nIf the user asks to deploy a full VSS profile, use\n[`..\u002Fvss-deploy-profile\u002FSKILL.md`](..\u002Fvss-deploy-profile\u002FSKILL.md). That skill\nowns profile routing, `generated.env`, `resolved.yml`, multi-service sizing, and\nfull-stack deploy\u002Fteardown.\n\nIf the user asks for standalone RT-VLM dense captioning, or no VSS profile is\nalready running, use the standalone RT-VLM flow in\n[`references\u002Fdeploy-rt-vlm-service.md`](references\u002Fdeploy-rt-vlm-service.md)\nbefore calling the API. This follows the same compose-centric pattern as\n`vss-deploy-profile`: gather context, run preflights, work from a local copy,\ndry-run with `docker compose config`, review, deploy, then wait for health.\n\n## Standalone Deployment Flow\n\nAlways follow this sequence. Never skip the dry-run.\n\n```bash\n# 1. Copy deploy\u002Fdocker\u002Fservices\u002Frtvi\u002Frtvi-vlm\u002Frtvi-vlm-docker-compose.yml\n#    into any writable standalone working directory.\n# 2. Derive RTVI_VLM_IMAGE_TAG from that compose copy.\n# 3. Strip the standalone-only dangling depends_on block from the copy.\n# 4. Create a gitignored .env with the required RT-VLM values.\n# 5. Prepare host bind paths such as $VSS_DATA_DIR\u002Fdata_log\u002Fvst\u002Fclip_storage.\n#    Use `sudo -n` for ownership fixes; if passwordless sudo is unavailable,\n#    stop and ask the host owner to run the printed command manually.\n# 6. docker compose --env-file .env -f rtvi-vlm-docker-compose.yml config --quiet\n# 7. docker pull the exact RT-VLM image tag.\n# 8. docker compose ... up -d rtvi-vlm, wait for ready, then smoke test.\n```\n\nRun preflights before any pull or `up`; stop and fix failures here before\ndebugging RT-VLM itself:\n\n```bash\nnvidia-smi --query-gpu=index,name --format=csv,noheader\nnvidia-container-cli info\ndocker compose version\ndocker run --rm --gpus all nvidia\u002Fcuda:12.4.0-base-ubuntu22.04 nvidia-smi\n```\n\nFor standalone single-file deployments, do not run the raw\n`deploy\u002Fdocker\u002Fservices\u002Frtvi\u002Frtvi-vlm\u002Frtvi-vlm-docker-compose.yml` directly: it\ncontains `depends_on` references to sibling VLM\u002FNIM services that are only\ndefined in the full VSS\u002Fmet-blueprints compose project. The standalone reference\nshows how to copy the compose file, derive the current image tag from it, strip\nthe `depends_on` block, and validate the result before `up`.\n\nFor agent-driven validation, never let `sudo` prompt interactively. Before any\nprivileged ownership or Docker operation, use the non-interactive guard in\n[`references\u002Fdeploy-rt-vlm-service.md`](references\u002Fdeploy-rt-vlm-service.md):\nprefer plain `docker`; otherwise use `sudo -n docker`; if `sudo -n` fails, stop\nwith the exact manual command for the host owner instead of retrying with\ninteractive sudo or weakening permissions.\n\nIf `docker pull` fails with a containerd snapshotter\u002Funpack error on Docker 28+,\napply the `\u002Fetc\u002Fdocker\u002Fdaemon.json` `containerd-snapshotter=false` fix in the\nstandalone reference before retrying.\n\nMinimum standalone `.env` values:\n\n| Host env var | Required when | Purpose |\n|---|---|---|\n| `NGC_CLI_API_KEY` | Standalone deploy path | NGC registry image pull and NGC model\u002Fartifact download |\n| `RTVI_VLM_API_KEY` or `NGC_CLI_API_KEY` | Authenticated API calls | RT-VLM bearer auth after the service is running |\n| `RTVI_VLM_PORT` | Always | Host API port mapped to container `8000` |\n| `HOST_IP` | Always | Kafka bootstrap host (`${HOST_IP}:9092`) |\n| `VSS_DATA_DIR` | Always | Required clip-storage bind mount |\n| `RTVI_VLM_MODEL_TO_USE` | Always for standalone | Backend selector; use `cosmos-reason2` for the default local model or `openai-compat` for a remote\u002Fsibling endpoint |\n| `RTVI_VLM_MODEL_PATH` | Local self-hosted model | Source-backed Cosmos Reason 2 path: `ngc:nim\u002Fnvidia\u002Fcosmos-reason2-8b:hf-1208` |\n| `RTVI_VLM_ENDPOINT` | `RTVI_VLM_MODEL_TO_USE=openai-compat` | Remote\u002Fsibling OpenAI-compatible VLM endpoint |\n| `VLM_NAME` | `RTVI_VLM_MODEL_TO_USE=openai-compat` | Model\u002Fdeployment name exposed by that endpoint |\n\n## Setup\n\n```bash\nexport BASE_URL=\"http:\u002F\u002Flocalhost:${RTVI_VLM_PORT:-8018}\"  # host-side RT-VLM port\nexport API_KEY=\"${NGC_CLI_API_KEY:-${RTVI_VLM_API_KEY:-}}\" # bearer token used by host-side curl commands\n: \"${API_KEY:?Set NGC_CLI_API_KEY or RTVI_VLM_API_KEY before calling authenticated endpoints}\"\n```\n\nEvery request below uses `Authorization: Bearer $API_KEY`. Health endpoints\n(`\u002Fv1\u002Fhealth\u002F*`, `\u002Fv1\u002Fready`, `\u002Fv1\u002Flive`, `\u002Fv1\u002Fstartup`) typically work without auth.\n\n**Smoke test before use:**\n```bash\ncurl -fsS \"$BASE_URL\u002Fv1\u002Fhealth\u002Fready\"\nMODEL_ID=\"$(curl -fsS \"$BASE_URL\u002Fv1\u002Fmodels\" -H \"Authorization: Bearer $API_KEY\" | jq -r '.data[0].id \u002F\u002F .id')\"\ncurl -fsS \"$BASE_URL\u002Fopenapi.json\" | jq -r '.paths | keys[]' | sort\n```\n\n## RTSP Sample Stream Guard\n\nWhen a task or eval names `RTSP_SAMPLE_URL`, treat that exact environment\nvariable as a required input. Verify it is set and non-empty before probing or\nregistering any stream; if it is missing, stop with a clear failure message. Do\nnot derive a substitute from NvStreamer, VIOS, sample-data bundles, or any other\nfallback, because that validates a different stream than the caller requested.\n\n```bash\n: \"${RTSP_SAMPLE_URL:?Set RTSP_SAMPLE_URL to a reachable RTSP sample stream before RTSP validation}\"\ncase \"$RTSP_SAMPLE_URL\" in\n  rtsp:\u002F\u002F*) ;;\n  *) echo \"RTSP_SAMPLE_URL must be an rtsp:\u002F\u002F URL, got: $RTSP_SAMPLE_URL\" >&2; exit 1 ;;\nesac\n\nif command -v ffprobe >\u002Fdev\u002Fnull 2>&1; then\n  ffprobe -v error -rtsp_transport tcp \\\n    -select_streams v:0 -show_entries stream=codec_type \\\n    -of csv=p=0 \"$RTSP_SAMPLE_URL\" | grep -qx video\nelif command -v gst-discoverer-1.0 >\u002Fdev\u002Fnull 2>&1; then\n  gst-discoverer-1.0 \"$RTSP_SAMPLE_URL\" | grep -qi 'video'\nelse\n  echo \"Install ffprobe or gst-discoverer-1.0 before RTSP validation.\" >&2\n  exit 1\nfi\n```\n\n## Quick Start — dense captions from a local video\n\n```bash\n# 1. Upload the video, capture its file id\nFILE_ID=$(curl -fsS -X POST \"$BASE_URL\u002Fv1\u002Ffiles\" \\\n  -H \"Authorization: Bearer $API_KEY\" \\\n  -F \"file=@\u002Fpath\u002Fto\u002Fwarehouse.mp4\" \\\n  -F \"purpose=vision\" \\\n  -F \"media_type=video\" | jq -r '.id')\n\n# 2. Generate captions + alerts (SSE stream of chunked responses)\ncurl -N -X POST \"$BASE_URL\u002Fv1\u002Fgenerate_captions\" \\\n  -H \"Authorization: Bearer $API_KEY\" \\\n  -H \"Content-Type: application\u002Fjson\" \\\n  -d \"{\n    \\\"id\\\": \\\"$FILE_ID\\\",\n    \\\"prompt\\\": \\\"Write a concise dense caption for each 10-second segment of this warehouse video.\\\",\n    \\\"model\\\": \\\"$MODEL_ID\\\",\n    \\\"chunk_duration\\\": 10,\n    \\\"stream\\\": true\n  }\"\n```\n\n## API Surface\n\nUse the live OpenAPI as the source of truth before calling optional endpoints:\n\n```bash\ncurl -fsS \"$BASE_URL\u002Fopenapi.json\" | jq -r '.paths | keys[]' | sort\n```\n\nCore paths for VSS 3.2 are:\n\n- `POST \u002Fv1\u002Ffiles` for multipart media upload; pass the returned file `id` into\n  caption generation and delete the file when finished.\n- `POST \u002Fv1\u002Fgenerate_captions` for file or stream captioning. Use the exact\n  model id returned by `GET \u002Fv1\u002Fmodels`; aliases such as `cosmos-reason2` are\n  backend selectors, not request model ids.\n- `POST \u002Fv1\u002Fstreams\u002Fadd`, `GET \u002Fv1\u002Fstreams\u002Fget-stream-info`, and\n  `DELETE \u002Fv1\u002Fstreams\u002Fdelete\u002F{stream_id}` for RTSP lifecycle. Parse stream ids\n  from `results[0].id`.\n- `POST \u002Fv1\u002Fchat\u002Fcompletions` for OpenAI-compatible text and multimodal calls.\n  Current 26.05 builds return HTTP 400 for text-only `\u002Fv1\u002Fcompletions`; treat\n  that as expected when validating legacy behavior.\n- `GET \u002Fv1\u002Fhealth\u002Fready`, `\u002Fv1\u002Fmodels`, `\u002Fv1\u002Fassets\u002Fstats`, and `\u002Fv1\u002Fmetrics`\n  for service probes. Do not assume `\u002Fv1\u002Flicense` exists unless OpenAPI lists it.\n\nDetailed endpoint schemas, response shapes, CV-style singular stream endpoints,\nand 26.05 compatibility notes live in\n[`references\u002Fapi-surface-26.05.md`](references\u002Fapi-surface-26.05.md).\n\n## Common Workflows\n\n- Stored file captioning: upload with `POST \u002Fv1\u002Ffiles`, call\n  `\u002Fv1\u002Fgenerate_captions` with the returned file id, use `stream=true` for SSE,\n  then delete the file to release storage.\n- RTSP live captioning: when the caller provides `RTSP_SAMPLE_URL`, use that\n  exact URL and run the **RTSP Sample Stream Guard** before registration. Do not\n  derive a replacement stream from NvStreamer or VIOS when `RTSP_SAMPLE_URL` is\n  empty; fail fast instead. Require an actual video stream\u002Fcaps entry before\n  registration; add the stream, caption it, then unregister it.\n- Alert prompts: include a deterministic `Anomaly Detected: Yes\u002FNo` line.\n  Kafka publication is server-side config, additive to HTTP responses, and\n  documented in [`references\u002Fkafka-workflows.md`](references\u002Fkafka-workflows.md).\n- Kafka validation: trust the live `vss-rtvi-vlm` environment for topic names.\n  In a full VSS alerts real-time profile, use the existing VSS Kafka container\n  `mdx-kafka` for CLI checks and final incident-consumer commands. For\n  standalone validation, use a broker that advertises `${HOST_IP}:9092`; never\n  stop or replace a pre-existing broker without user confirmation.\n\n## Error Reference\n\nCommon causes: 400 for invalid request shape or model id, 401\u002F403 for missing\nor wrong bearer token, 404 for deleted files\u002Fstreams or unsupported endpoints,\n413 for oversized uploads, 422 for schema validation, 429 for too much\nconcurrency, 500 for inference\u002Fruntime failures, and 503 while startup is still\nin progress. Inspect `docker logs vss-rtvi-vlm` for service-side failures.\n",{"name":5,"license":15,"metadata":55,"description":7},{"tags":56,"version":57,"github-url":58},"nvidia blueprint operational deployment","3.2.0","https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-Blueprints\u002Fvideo-search-and-summarization","claude\u002Finstall-gsd-build-foQdV","d86cd127f527124b5fd0fd0bc679a4d79b7d01d7",".agents\u002Fskills\u002Fvss-deploy-dense-captioning","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik","https:\u002F\u002Fnorvik-orpin.vercel.app","https:\u002F\u002Fgithub.com\u002Fpattanation","2026-07-17 11:17:04",{"source":8,"skillName":5,"commands":67,"prompt":77,"zipAvailable":78},[68,71,74],{"tool":69,"command":70},"npx","npx skills add pattanation\u002FNorvik --skill vss-deploy-dense-captioning",{"tool":72,"command":73},"bunx","bunx skills add pattanation\u002FNorvik --skill vss-deploy-dense-captioning",{"tool":75,"command":76},"pnpm","pnpm dlx skills add pattanation\u002FNorvik --skill vss-deploy-dense-captioning","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002FSKILL.md",true,[80,86,91,95,99,103,107,111,115,119],{"path":81,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":84,"sourceHtmlUrl":85},"BENCHMARK.md","REFERENCE","参考文档",3898,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002FBENCHMARK.md",{"path":87,"resourceType":88,"resourceTypeDesc":89,"sizeBytes":14,"sourceHtmlUrl":90},"evals\u002F.gitkeep","OTHER","其他","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fevals\u002F.gitkeep",{"path":92,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":93,"sourceHtmlUrl":94},"evals\u002Falerts_profile_api.json",6033,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fevals\u002Falerts_profile_api.json",{"path":96,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":97,"sourceHtmlUrl":98},"evals\u002Fevals.json",2304,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fevals\u002Fevals.json",{"path":100,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":101,"sourceHtmlUrl":102},"evals\u002Fstandalone_api.json",7632,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fevals\u002Fstandalone_api.json",{"path":104,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":105,"sourceHtmlUrl":106},"references\u002Fapi-surface-26.05.md",5626,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Freferences\u002Fapi-surface-26.05.md",{"path":108,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":109,"sourceHtmlUrl":110},"references\u002Fdeploy-rt-vlm-service.md",36327,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Freferences\u002Fdeploy-rt-vlm-service.md",{"path":112,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":113,"sourceHtmlUrl":114},"references\u002Fkafka-workflows.md",15004,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Freferences\u002Fkafka-workflows.md",{"path":116,"resourceType":82,"resourceTypeDesc":83,"sizeBytes":117,"sourceHtmlUrl":118},"skill-card.md",4070,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fskill-card.md",{"path":120,"resourceType":88,"resourceTypeDesc":89,"sizeBytes":121,"sourceHtmlUrl":122},"skill.oms.sig",6045,"https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvss-deploy-dense-captioning\u002Fskill.oms.sig",[124,154,179,203,225,240],{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":128,"sourceHtmlUrl":129,"categories":130,"keywords":137,"publishedAt":50},"01M71QDV4DM8T19HEHWATEF1K4","webapp-testing","Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fwebapp-testing\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fwebapp-testing\u002FSKILL.md",[131,134],{"categoryNo":132,"categoryName":133,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发",[138,139,140,43,141,142,143,144,145,146,147,148,149,150,151,152,153],"webapp","testing","webapp testing","applications","local","frontend","debugging","toolkit","logs","browser","with","playwright","web","behavior","ui","screenshots",{"skillNo":155,"name":156,"nameZh":6,"description":157,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":158,"sourceHtmlUrl":159,"categories":160,"keywords":162,"publishedAt":50},"01M71QDH2GGMXGA9105C9W1N7B","web-artifacts-builder","Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn\u002Fui). Use for complex artifacts requiring state management, routing, or shadcn\u002Fui components - not for simple single-file HTML\u002FJSX artifacts.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fweb-artifacts-builder\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fweb-artifacts-builder\u002FSKILL.md",[161],{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},[163,164,165,150,166,167,168,43,169,170,143,171,38,172,173,174,175,176,177,178],"builder","artifacts builder","artifacts","web artifacts builder","single-file","css","Tailwind","ui components","single","file","components","react","multi","state","routing","state management",{"skillNo":180,"name":181,"nameZh":6,"description":182,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":183,"licenseFileUrl":184,"sourceHtmlUrl":185,"categories":186,"keywords":188,"publishedAt":50},"01M71QD9T2XC3EWMQNN9ACNWQ1","vueuse-functions","Apply VueUse composables where appropriate to build concise, maintainable Vue.js \u002F Nuxt features.","MIT","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvueuse-functions\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvueuse-functions\u002FSKILL.md",[187],{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},[189,190,191,192,193,194,195,196,197,198,199,200,201,202],"functions","vueuse","vueuse functions","project","nuxt","apply","github","concise","build","https","features","vue","js","composables",{"skillNo":204,"name":205,"nameZh":6,"description":206,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":183,"licenseFileUrl":207,"sourceHtmlUrl":208,"categories":209,"keywords":212,"publishedAt":50},"01M71QD51HWSGYMCMB7RW7PGK8","vue-testing-best-practices","Use for Vue.js testing. Covers Vitest, Vue Test Utils, component testing, mocking, testing patterns, and Playwright for E2E testing.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-testing-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-testing-best-practices\u002FSKILL.md",[210,211],{"categoryNo":132,"categoryName":133,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},[213,139,214,200,215,149,216,217,218,195,219,43,220,221,222,38,223,201,224],"vue testing best practices","testing best practices","best practices","test","testing patterns","component","vitest","mocking","e2e","utils","patterns","e2e testing",{"skillNo":226,"name":227,"nameZh":6,"description":228,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":183,"licenseFileUrl":229,"sourceHtmlUrl":230,"categories":231,"keywords":233,"publishedAt":50},"01M71QD358H2DJ9THJRBP77T3T","vue-router-best-practices","Vue Router 4 patterns, navigation guards, route params, and route-component lifecycle interactions.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-router-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-router-best-practices\u002FSKILL.md",[232],{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},[234,200,235,215,236,218,237,195,223,238,239],"router","vue router best practices","route","guards","navigation","lifecycle",{"skillNo":241,"name":242,"nameZh":6,"description":243,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":183,"licenseFileUrl":244,"sourceHtmlUrl":245,"categories":246,"keywords":248,"publishedAt":50},"01M71QCYPNWDE2MQZ1TF3ED5S4","vue-best-practices","MUST be used for Vue.js tasks. Strongly recommends Composition API with `\u003Cscript setup>` and TypeScript as the standard approach. Covers Vue 3, SSR, Volar, vue-tsc. Load for any Vue, .vue files, Vue Router, Pinia, or Vite with Vue work. ALWAYS use Composition API unless the project explicitly requires Options API.","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-best-practices\u002FLICENSE.md","https:\u002F\u002Fgithub.com\u002Fpattanation\u002FNorvik\u002Fblob\u002Fclaude\u002Finstall-gsd-build-foQdV\u002F.agents\u002Fskills\u002Fvue-best-practices\u002FSKILL.md",[247],{"categoryNo":135,"categoryName":136,"taxonomyType":21,"taxonomyTypeDesc":22},[249,200,215,250,192,251,252,35,234,253,254,255,256,257,258,43,259,38,260,261],"vue best practices","Composition","files","tasks","pinia","work","options","ssr","typescript","as","vite","load","standard"]