vss-generate-video-report-rag
Generates VSS video summary reports with LVS HITL and optional Enterprise RAG document grounding. Trigger when the user asks for a frag/RAG-assisted video report, knowledge-enhanced analysis, or Enterprise RAG context in a video summary.
NVIDIA-AI-Blueprints · 已有 AI 审核记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add NVIDIA-AI-Blueprints/video-search-and-summarization --skill vss-generate-video-report-rag技能说明
name: vss-generate-video-report-rag description: Generates VSS video summary reports with LVS HITL and optional Enterprise RAG document grounding. Trigger when the user asks for a frag/RAG-assisted video report, knowledge-enhanced analysis, or Enterprise RAG context in a video summary. license: Apache-2.0 metadata: version: "3.2.0" github-url: "https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarization" tags: "nvidia blueprint operational"
VSS Generate Video Report RAG — Video Analysis with Enterprise RAG
Generate video summary reports using the LVS profile's RAG-enabled agent config. This skill adds Enterprise RAG document grounding and guided human-in-the-loop (HITL) parameter collection on top of the VSS agent.
Always run curl commands yourself; never instruct the user to run them.
Enable Enterprise RAG on the LVS Profile
The repository ships the RAG-enabled LVS agent config at
deploy/docker/developer-profiles/dev-profile-lvs/vss-agent/configs/config_rag.yml.
It is a superset of the default LVS config: regular caption retrieval remains
enabled, and frag_retrieval adds Enterprise RAG document grounding.
Use the normal /vss-deploy-profile workflow for deployment. The source
.env remains read-only; apply non-secret overrides to
deploy/docker/developer-profiles/dev-profile-lvs/generated.env.
generated.env is ignored by the repository, but it is still a plaintext file:
do not commit it, paste it into logs, or store long-lived credentials there.
Prefer a vault, Docker secrets, or ephemeral shell environment variables for
API keys.
Step 1: Configure the generated env file
REPO=${REPO:-$(git rev-parse --show-toplevel)}
cd "$REPO"
cp deploy/docker/developer-profiles/dev-profile-lvs/.env \
deploy/docker/developer-profiles/dev-profile-lvs/generated.env
Set these non-secret values in generated.env:
HOST_IP— host IP (hostname -I | awk '{print $1}')VSS_AGENT_CONFIG_FILE=./deploy/docker/developer-profiles/dev-profile-lvs/vss-agent/configs/config_rag.ymlRAG_SERVER_URL— Enterprise RAG server HTTP endpoint (defaults tohttp://rag-server:8081/v1)KNOWLEDGE_COLLECTION— default Enterprise RAG collection forfrag_retrieval
Keep sensitive values (NGC_CLI_API_KEY, NVIDIA_API_KEY, RAG_API_KEY) out
of generated.env and out of resolved.yml. Do not export them before running
docker compose config > resolved.yml, because Compose expands environment
variables into that file. Use a secret manager, an existing authenticated Docker
session, or a local override file that references an ephemeral shell variable at
up time.
Step 2: Log in to NGC registry
Prefer an existing authenticated Docker session or a secret-managed login. If a login is required, use --password-stdin without printing token values:
read -rsp "NGC API key: " NGC_CLI_API_KEY
printf '%s\n' "$NGC_CLI_API_KEY" | docker login nvcr.io --username '$oauthtoken' --password-stdin
unset NGC_CLI_API_KEY
Step 3: Deploy the LVS profile with the RAG config
Do not export RAG_API_KEY for the dry-run below. If the RAG server requires an
API key, create this untracked local override after resolved.yml is generated:
cat > rag-secret.override.yml <<'EOF'
services:
vss-agent:
environment:
RAG_API_KEY: ${RAG_API_KEY:?Set RAG_API_KEY only for docker compose up}
EOF
REPO=${REPO:-$(git rev-parse --show-toplevel)}
cd "$REPO/deploy/docker"
docker compose --env-file developer-profiles/dev-profile-lvs/generated.env \
config > resolved.yml
uv run "$REPO/skills/vss-deploy-profile/scripts/normalize_resolved_yml.py" \
"$REPO/deploy/docker/resolved.yml"
docker compose --env-file developer-profiles/dev-profile-lvs/generated.env \
-f resolved.yml up -d
When rag-secret.override.yml is needed, use:
read -rsp "RAG API key: " RAG_API_KEY
RAG_API_KEY="$RAG_API_KEY" docker compose \
--env-file developer-profiles/dev-profile-lvs/generated.env \
-f resolved.yml -f rag-secret.override.yml up -d
unset RAG_API_KEY
Step 4: Verify deployment
# Check containers are running
docker ps --format "table {{.Names}}\t{{.Status}}"
# Health check
curl -sf --max-time 5 "http://${HOST_IP}:${VSS_AGENT_PORT:-8000}/health" >/dev/null \
&& echo "VSS LVS RAG agent is running" \
|| echo "VSS LVS RAG agent is NOT reachable"
Tear down
REPO=${REPO:-$(git rev-parse --show-toplevel)}
cd "$REPO/deploy/docker"
docker compose -f resolved.yml down
When to Use
- User wants to generate a video summary or report using the RAG-enabled LVS pipeline
- User asks to analyze a video with Enterprise RAG knowledge context
- User mentions "frag", "enterprise RAG", or "knowledge-enhanced report"
When NOT to Use
- Simple video understanding queries (use
video-understandingskill) - Direct LVS summarization without HITL (use
video-summarizationskill) - Deployment tasks (use
deployskill) - Real-time alerts (use
alertsskill)
Workflow: Generate an LVS Report with Enterprise RAG
Step 1: List available videos
curl -sS -X POST "http://${HOST_IP}:${VSS_AGENT_PORT:-8000}/v1/chat" \
-H "Content-Type: application/json" \
-d '{"messages": [{"role": "user", "content": "What videos are available?"}]}' | \
python3 -c "import json,sys; d=json.load(sys.stdin); print(d['choices'][0]['message']['content'])"
A selected video is required before Step 2. If the user has not already named one, return the short list and stop; resume when the user supplies the video name.
Step 2: Collect parameters from the user
Required user-provided parameters:
- Scenario — scenario label for the video. Example: "warehouse monitoring", "traffic monitoring", "retail store activity"
- Events — comma-separated event names to detect. Example: "accident, forklift stuck, workers not wearing PPE, person entering restricted area"
- Objects of Interest — focus objec