[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7A93KSW98SX7RRD8JAD52DT":3,"skill-same:01M7A93KSW98SX7RRD8JAD52DT":81},{"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":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":37,"commitSha":38,"skillDirectory":39,"forksCount":40,"repositoryHtmlUrl":41,"repositoryHomepageUrl":42,"ownerHtmlUrl":43,"ownerBlogUrl":44,"ownerDescription":45,"updatedAt":46,"install":47,"resources":60},"01M7A93KSW98SX7RRD8JAD52DT","amc-setup-calibration-stack",null,"Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.","themusebook\u002Fskills","themusebook","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F349927247?v=4","01M7A5C1RV5S3V5FF6ZACVDNRN","musebook","01M7A5C1SFC3KRZAXNN60XJ7KM",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: \"amc-setup-calibration-stack\"\ndescription: \"Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.\"\nmetadata:\n  tags: [amc, deepstream, docker, calibration, setup, ngc]\nowner: \"NVIDIA CORPORATION\"\nservice: \"auto-magic-calib\"\nversion: \"1.0.0\"\nreviewed: \"2026-04-28\"\nlicense: \"Apache-2.0\"\ndata_classification: public\n---\n\n# Skill: Launch AutoMagicCalib Release Containers\n\n## When to Use This Skill\n\nActivate this skill when the user wants to bring up the AutoMagicCalib stack. Typical prompts:\n\n- \"deploy auto calibration\"\n- \"launch auto calibration\" \u002F \"launch AMC\"\n- \"start MS+UI\" \u002F \"bring up the microservice and UI\"\n- \"set up auto-magic-calib\"\n\nPrerequisite: NGC API key (the skill prompts for it). For first-time setup on a fresh machine, this also resolves or clones the `auto-magic-calib` repo after explicit user confirmation.\n\n## Overview\nLaunch the AutoMagicCalib microservice (MS) and UI from pre-built release images via Docker Compose. Use this for production deployments or when you want to run the full stack (backend + web UI) without building images locally.\n\n## Prerequisites\n- Docker and Docker Compose installed\n- NVIDIA Docker Runtime configured (for GPU support)\n- `auto-magic-calib` repo on disk — Step 0b auto-resolves an existing checkout via `git rev-parse --show-toplevel`, or offers to clone `https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib` into `~\u002Fauto-magic-calib` if none is found\n- NGC account with access to NVIDIA container registry\n- **Docker must be runnable without `sudo`** — verify with `docker ps`. If it fails, ask the user to follow the post-install steps: https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002Flinux-postinstall\u002F\n\n> **If `docker ps` fails with \"permission denied\"**, ask the user to run:\n> ```bash\n> sudo usermod -aG docker $USER\n> newgrp docker   # applies group change in current shell without logout\n> ```\n> Then verify with `docker ps` before continuing.\n\n## Instructions\n\n### Step 0: Verify Docker Runs Without sudo\n\n```bash\ndocker ps\n```\n\n- If it succeeds → continue.\n- If it fails with \"permission denied\" → the user is not in the `docker` group. Ask the user to run:\n  ```bash\n  sudo usermod -aG docker $USER && newgrp docker\n  ```\n  Then ask the user to confirm `docker ps` works before continuing.\n\n> **Agent note**: If `docker ps` cannot be run from within the agent sandbox, ask the user to confirm it works (e.g. \"Can you confirm `docker ps` runs without sudo?\") before proceeding.\n\n### Step 0b: Resolve Repo Checkout\n\nThe skill needs `compose\u002F`, `assets\u002Fsdg_08_2_sample_data_010926.zip`, and a `models\u002F` mount point — all of which live in the `auto-magic-calib` repo. If you're already inside a checkout, the skill uses it. Otherwise it asks for explicit user confirmation using the host's question mechanism; if none is available, ask in chat and wait before cloning into `~\u002Fauto-magic-calib`.\n\n```bash\nREPO_URL=\"https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib.git\"\nDEFAULT_CLONE_DIR=\"$HOME\u002Fauto-magic-calib\"\n\n# 1. Already inside a usable checkout?\nREPO_ROOT=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)\"\nif [ -n \"$REPO_ROOT\" ] && [ -f \"$REPO_ROOT\u002Fcompose\u002Fcompose.yml\" ] && [ -d \"$REPO_ROOT\u002Fassets\" ]; then\n  echo \"✓ Using existing checkout: $REPO_ROOT\"\n\n# 2. Default clone dir already populated from a prior run?\nelif [ -d \"$DEFAULT_CLONE_DIR\u002F.git\" ] && [ -f \"$DEFAULT_CLONE_DIR\u002Fcompose\u002Fcompose.yml\" ]; then\n  REPO_ROOT=\"$DEFAULT_CLONE_DIR\"\n  echo \"✓ Found existing clone at $REPO_ROOT\"\n\n# 3. Nothing on disk — STOP and ask the user for confirmation using the\n#    host's question mechanism; if none is available, ask in chat and wait.\n#    Do NOT clone silently from this block.\nelse\n  echo \"No auto-magic-calib checkout found. Ask the user for confirmation:\"\n  echo \"  Clone $REPO_URL into $DEFAULT_CLONE_DIR? [y\u002FN]\"\n  echo \"On 'y' — run: git clone \\\"$REPO_URL\\\" \\\"$DEFAULT_CLONE_DIR\\\"\"\n  exit 1\nfi\n\ncd \"$REPO_ROOT\"\nexport REPO_ROOT\necho \"REPO_ROOT=$REPO_ROOT\"\n```\n\n> **Agent note**: do **not** clone silently. Ask the user first using the host's question mechanism; if none is available, ask in chat and wait. Example: \"auto-magic-calib repo not found. Clone `https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib` into `~\u002Fauto-magic-calib`? (or provide your own path)\". Honour an alternate path if the user offers one. The clone is a few hundred MB (compose files + sample dataset + assets).\n\n### Step 0c: Install Python venv (New Systems Only)\n\nOn a fresh system, `pip` and `python3-venv` may not be available. Install them first:\n\n```bash\n# Create a venv for HuggingFace CLI (project-local preferred)\nREPO_DIR=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)\"\nHF_VENV=\"${REPO_DIR}\u002Fvenv\"\npython3 -m venv \"$HF_VENV\" 2>\u002Fdev\u002Fnull || {\n  echo \"ERROR: python3-venv not available.\" >&2\n  echo \"Install it manually: sudo apt install -y python3-venv python3-pip\" >&2\n  exit 1\n}\n\n# Install HuggingFace hub (needed for VGGT download)\n\"$HF_VENV\u002Fbin\u002Fpip\" install --upgrade pip huggingface_hub\n```\n\n> **Note**: Skip this step if a venv with `hf` already exists (check `venv\u002Fbin\u002Fhf` in the repo root or `~\u002Fvenv\u002Famc\u002Fbin\u002Fhf`).\n\n### Step 1: Login to NGC\n\nAsk the user for their NGC API key using the host's question mechanism; if none is available, ask in chat and wait. Then run:\n\n```bash\necho \"\u003CNGC_API_KEY>\" | docker login nvcr.io --username '$oauthtoken' --password-stdin\necho \"✓ NGC authentication complete\"\n```\n\n### Step 2: Download VGGT Model (If Not Already Present)\n\n```bash\nexport REPO_ROOT=$(git rev-parse --show-toplevel)\ncd \"$REPO_ROOT\"\n\nif [ -f \"models\u002Fvggt\u002Fvggt_1B_commercial.pt\" ]; then\n  echo \"✓ VGGT model already present\"\nelse\n  echo \"✗ VGGT model not found\"\n  echo \"Options:\"\n  echo \"  1. Continue without VGGT (AMC only - sufficient for most use cases)\"\n  echo \"  2. Download VGGT model (~4.7GB, requires HuggingFace account)\"\nfi\n```\n\n**To download VGGT** (ask the user for a HuggingFace token using the host's question mechanism; if none is available, ask in chat and wait):\n\n**Step 2a: Accept Model License** (required, one-time):\n1. Visit: https:\u002F\u002Fhuggingface.co\u002Ffacebook\u002FVGGT-1B-Commercial\n2. Log in to your HuggingFace account\n3. Click \"Agree and access repository\" to accept license terms\n\n**Step 2b: Get HuggingFace Token**:\n1. Visit: https:\u002F\u002Fhuggingface.co\u002Fsettings\u002Ftokens\n2. Create new token with \"Read\" access (starts with `hf_...`)\n3. Ask the user for the token using the host's question mechanism; if none is available, ask in chat and wait. Do NOT ask them to run a command\n\n**Step 2c: Download** (pass token via `HF_TOKEN` env var — keeps the secret out of `hf`'s argv \u002F `ps aux`; no interactive login needed):\n```bash\nREPO_DIR=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)\"\ncd \"$REPO_DIR\"\n\n# Find the HuggingFace CLI binary (named 'hf', not 'huggingface-cli')\nHF_BIN=\"$(find \"$REPO_DIR\u002Fvenv\" ~\u002Fvenv\u002Famc -name hf -type f 2>\u002Fdev\u002Fnull | head -1)\"\n{ [ -z \"$HF_BIN\" ] || [ ! -x \"$HF_BIN\" ]; } && { echo \"ERROR: hf binary not found or not executable; install the hf CLI (Step 0c) or set HF_BIN\" >&2; exit 1; }\n\n# Do NOT use --token on the command line (leaks via ps\u002Fargv). The HF CLI\n# reads HF_TOKEN from the environment automatically.\nHF_TOKEN=\"\u003CHF_TOKEN>\" \"$HF_BIN\" download facebook\u002FVGGT-1B-Commercial \\\n  --local-dir models\u002Fvggt\u002F\n\n# Verify\nls -lh models\u002Fvggt\u002Fvggt_1B_commercial.pt\n# Should show ~4.7GB file\n```\n\n> **Important**: Download BEFORE setting `chown 1000:1000` on the models directory — the current user needs write access during download. Set permissions in Step 4 after download completes.\n\n### Step 3: Configure .env Variables\n\nThe `.env` file at `compose\u002F.env` controls ports and paths. Update it before launching:\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Find available backend port (8000-8009)\nfor port in {8000..8009}; do\n  if ! lsof -Pi :$port -sTCP:LISTEN -t >\u002Fdev\u002Fnull 2>&1; then\n    MS_PORT=$port\n    echo \"Using backend port: $MS_PORT\"\n    break\n  fi\ndone\n[ -z \"$MS_PORT\" ] && { echo \"ERROR: no free backend port in 8000-8009; free one or widen the range.\" >&2; exit 1; }\n\n# Find available UI port (5000-5009)\nfor port in {5000..5009}; do\n  if ! lsof -Pi :$port -sTCP:LISTEN -t >\u002Fdev\u002Fnull 2>&1; then\n    UI_PORT=$port\n    echo \"Using UI port: $UI_PORT\"\n    break\n  fi\ndone\n[ -z \"$UI_PORT\" ] && { echo \"ERROR: no free UI port in 5000-5009; free one or widen the range.\" >&2; exit 1; }\n\n# Get host IP\nHOST_IP=$(hostname -I | awk '{print $1}')\necho \"Host IP: $HOST_IP\"\n\n# Update .env — preserve any keys the user has already set; back up first.\n# .env may contain credentials, so restrict permissions on both the backup and\n# the live file (chmod 600). Add `compose\u002F.env.bak.*` to .gitignore.\nENV_FILE=\".env\"\nif [ -f \"$ENV_FILE\" ]; then\n  BACKUP=\"${ENV_FILE}.bak.$(date +%s)\"\n  cp \"$ENV_FILE\" \"$BACKUP\"\n  chmod 600 \"$BACKUP\"\nfi\ntouch \"$ENV_FILE\"\nchmod 600 \"$ENV_FILE\"\nset_env_key() {\n  local k=\"$1\" v=\"$2\"\n  if grep -qE \"^${k}=\" \"$ENV_FILE\"; then\n    sed -i \"s|^${k}=.*|${k}=${v}|\" \"$ENV_FILE\"\n  else\n    echo \"${k}=${v}\" >> \"$ENV_FILE\"\n  fi\n}\nset_env_key AUTO_MAGIC_CALIB_MS_PORT \"${MS_PORT}\"\nset_env_key AUTO_MAGIC_CALIB_UI_PORT \"${UI_PORT}\"\nset_env_key PROJECT_DIR \"..\u002F..\u002Fprojects\"\nset_env_key MODEL_DIR \"..\u002F..\u002Fmodels\"\nset_env_key HOST_IP \"${HOST_IP}\"\n\n# Keep timestamped .env backups out of git.\nGITIGNORE=\"$REPO_ROOT\u002F.gitignore\"\ntouch \"$GITIGNORE\"\ngrep -qxF \"compose\u002F.env.bak.*\" \"$GITIGNORE\" || echo \"compose\u002F.env.bak.*\" >> \"$GITIGNORE\"\n\necho \"✓ .env updated\"\ncat .env\n```\n\n**Important**: `HOST_IP` must be the machine's network IP (not `localhost`) so the UI container can reach the backend from a browser.\n\nOptional: set `VGGT_MODEL_PATH` only if the VGGT model is mounted at a non-default container path; default is `\u002Ftmp\u002Fvggt_model\u002Fvggt_1B_commercial.pt` inside the MS container.\n\n### Step 4: Set Directory Permissions\n\nThe containers run as UID\u002FGID 1000. The `projects` and `models` directories must be owned by this UID for containers to read\u002Fwrite properly:\n\n```bash\ncd \"$REPO_ROOT\"\n\n# Create projects directory if it doesn't exist\nmkdir -p projects\n\n# Set ownership (required for containers to write calibration outputs).\n# Do this AFTER VGGT download is complete (current user needs write access during download).\n# Get explicit user confirmation before running sudo chown — it recursively changes\n# ownership of $REPO_ROOT\u002Fprojects and $REPO_ROOT\u002Fmodels to UID\u002FGID 1000.\n[ -d projects ] && [ -d models ] || {\n  echo \"ERROR: expected projects\u002F and models\u002F under $REPO_ROOT\" >&2; exit 1;\n}\necho \"About to chown -R 1000:1000 on:\"\necho \"  $REPO_ROOT\u002Fprojects\"\necho \"  $REPO_ROOT\u002Fmodels\"\necho \"(required because containers run as UID 1000). Confirm before proceeding.\"\nsudo chown 1000:1000 -R projects\nsudo chown 1000:1000 -R models\n\necho \"✓ Permissions set\"\n```\n\n### Step 5: Launch Services\n\nBefore pulling, fail fast if the NGC key authenticated in Step 1 but cannot actually access a release image — otherwise `docker compose up` aborts partway with a 401\u002F403 after some work is already done.\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Fail-fast image-access check: confirm the NGC key can reach every release\n# image BEFORE pulling. `docker manifest inspect` checks registry access without\n# downloading layers, and the image list is read from the resolved compose so it\n# tracks the release tag automatically.\nIMAGES=$(docker compose config --images | sort -u)\n[ -z \"$IMAGES\" ] && { echo \"ERROR: no images resolved from compose — check compose\u002F.env and the chosen profile.\" >&2; exit 1; }\nfor img in $IMAGES; do\n  echo \"Checking access: $img\"\n  if ! docker manifest inspect \"$img\" >\u002Fdev\u002Fnull 2>&1; then\n    echo \"NGC login succeeded, but this key cannot access the required image:\" >&2\n    echo \"  $img\" >&2\n    echo \"Provide an NGC key with access to this image's namespace, then re-run Step 1 (login) and retry.\" >&2\n    exit 1\n  fi\ndone\n\n# Start all services (images pulled automatically on first run)\ndocker compose up -d\n\n# Check containers are running\ndocker compose ps\n```\n\n**Expected output**:\n```\nNAME                    IMAGE                                                              STATUS\nauto-magic-calib-ms-1   nvcr.io\u002Fnvidia\u002Fauto-magic-calib:2.0.0           Up (healthy)\nauto-magic-calib-ui-1   nvcr.io\u002Fnvidia\u002Fauto-magic-calib-ui:2.0.0        Up\n```\n\n### Step 6: Verify Services Are Running\n\n```bash\n# Read ports from .env\nMS_PORT=$(grep AUTO_MAGIC_CALIB_MS_PORT $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\nUI_PORT=$(grep AUTO_MAGIC_CALIB_UI_PORT $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\nHOST_IP=$(grep HOST_IP $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\n\n# Wait for microservice readiness. Cold image pulls or first startup can need\n# extra time after `docker compose up -d` returns.\nREADY_URL=\"http:\u002F\u002Flocalhost:${MS_PORT}\u002Fv1\u002Fready\"\necho \"Waiting for microservice readiness at ${READY_URL} ...\"\nready_response=\"\"\nfor attempt in $(seq 1 24); do\n  if ready_response=$(curl -fsS --max-time 5 \"${READY_URL}\" 2>\u002Fdev\u002Fnull) && \\\n     echo \"${ready_response}\" | grep -q '\"code\"[[:space:]]*:[[:space:]]*0'; then\n    echo \"Microservice ready: ${ready_response}\"\n    break\n  fi\n  if [ \"${attempt}\" -lt 24 ]; then\n    printf \"  [%02d\u002F24] Microservice not ready yet; retrying in 5s...\\n\" \"${attempt}\"\n    sleep 5\n  fi\ndone\n\nif ! echo \"${ready_response}\" | grep -q '\"code\"[[:space:]]*:[[:space:]]*0'; then\n  echo \"ERROR: microservice did not report ready within 120 seconds: ${READY_URL}\" >&2\n  echo \"Check status and logs:\" >&2\n  echo \"  cd ${REPO_ROOT}\u002Fcompose && docker compose ps\" >&2\n  echo \"  cd ${REPO_ROOT}\u002Fcompose && docker compose logs auto-magic-calib-ms\" >&2\n  exit 1\nfi\n\n# Check UI is serving\nUI_STATUS=$(curl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" --max-time 5 \"http:\u002F\u002Flocalhost:${UI_PORT}\")\nif [ \"${UI_STATUS}\" != \"200\" ]; then\n  echo \"ERROR: Web UI returned HTTP ${UI_STATUS}; check docker compose ps and UI logs.\" >&2\n  exit 1\nfi\necho \"Web UI ready: HTTP ${UI_STATUS}\"\n\necho \"Microservice: http:\u002F\u002F${HOST_IP}:${MS_PORT}\"\necho \"Web UI:       http:\u002F\u002F${HOST_IP}:${UI_PORT}\"\n```\n\n## Success Criteria\n\n**Both containers running** — see `docker compose ps` output from Step 5. Both services should show \"Up\" status; microservice should show \"(healthy)\" in the STATUS column.\n\n**Microservice healthy** — Step 6 readiness polling returns `code:0` from `\u002Fv1\u002Fready` and prints the service URL.\n\n**Web UI accessible**:\n- Open browser: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_UI_PORT>`\n- Should display the AutoMagicCalib web interface\n- Should be able to create projects and run calibration\n\n## Key Output\n\n**Microservice**: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_MS_PORT>` (default port 8000)\n- API docs: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_MS_PORT>\u002Fdocs`\n\n**Web UI**: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_UI_PORT>` (default port 5000)\n- Interactive project management\n- File upload interface\n- Calibration configuration\n- Real-time status monitoring\n- Results visualization and download\n\n**Data Persistence**:\n- Projects stored: `$REPO_ROOT\u002Fprojects\u002F`\n- State persisted: `$REPO_ROOT\u002Fprojects\u002Fstate.json`\n\n## Troubleshooting\n\n| Issue | Symptoms | Solution |\n|-------|----------|----------|\n| `pip` not found | `pip: command not found` | Run `sudo apt install -y python3.12-venv python3-pip` then create venv (Step 0) |\n| `huggingface-cli` not found | `huggingface-cli: command not found` | The binary is named `hf` in the venv. Find it with: `find venv ~\u002Fvenv\u002Famc -name hf -type f 2>\u002Fdev\u002Fnull \\| head -1` |\n| `python3 -m venv` fails | \"ensurepip not available\" | Run `sudo apt install -y python3.12-venv` first |\n| Docker permission denied | \"permission denied while trying to connect to docker socket\" | User not in docker group — ask user to run: `sudo usermod -aG docker $USER && newgrp docker`. See: https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002Flinux-postinstall\u002F |\n| `docker login` itself rejected | Step 1 login returns an authentication error | The key is invalid or expired. Ask the user for a current NGC key and log in again before continuing. |\n| Key logs in but can't access an image | The Step 5 access check stops with \"cannot access the required image\" \u002F a 401\u002F403 on `docker manifest inspect`, before any pull | The key authenticates but lacks access to that image's namespace. Re-running login with the same key won't help — ask the user for an NGC key with access to the required namespace, re-run Step 1, then retry Step 5. |\n| VGGT download permission error | \"PermissionError: [Errno 13] Permission denied: 'models\u002Fvggt\u002F.cache'\" | Download VGGT BEFORE setting `chown 1000:1000` on models. Fix: `sudo chown -R $(id -u):$(id -g) models` then re-download |\n| Port already in use | \"address already in use\" | Find available port in 8000-8009 (MS) or 5000-5009 (UI); update `.env` |\n| Microservice readiness timeout | Step 6 reports that `\u002Fv1\u002Fready` did not return `code:0` within 120 seconds | Run `cd $REPO_ROOT\u002Fcompose && docker compose ps`, then inspect logs with `docker compose logs auto-magic-calib-ms` |\n| Permission denied (projects) | \"Permission denied: 'projects\u002F...'\" in MS logs | Run: `sudo chown 1000:1000 -R projects` |\n| Permission denied (models) | \"Permission denied: 'models\u002F...'\" in MS logs | Run: `sudo chown 1000:1000 -R models` |\n| UI can't reach backend | Browser shows connection error | Verify `HOST_IP` in `.env` is the machine's network IP, not `localhost` |\n| VGGT model not found | Warning in MS logs about missing VGGT | Download model (Step 2) or ignore (AMC works without VGGT) |\n| Container exits immediately | Status \"Exited\" | Check logs: `docker compose logs auto-magic-calib-ms` |\n| GPU not available | \"CUDA not available\" in logs | Check NVIDIA runtime: `docker run --rm --runtime=nvidia --gpus all ubuntu:20.04 nvidia-smi` |\n| \"No such file or directory\" when verifying | After launch, can't find compose directory | Working directory persists after `cd compose`. Use absolute paths or run `cd $REPO_ROOT` first |\n\n**Common Fixes**:\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# View logs\ndocker compose logs -f\n\n# View logs for specific service\ndocker compose logs -f auto-magic-calib-ms\n\n# Restart all services\ndocker compose restart\n\n# Stop and remove containers\ndocker compose down\n\n# Update .env and relaunch\ndocker compose up -d\n```\n\n## Stopping the Services\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Stop all services (containers removed, data persisted)\ndocker compose down\n\n# Stop and remove volumes\ndocker compose down -v\n```\n\n## Related Skills\n- `skills\u002Famc-run-sample-calibration\u002FSKILL.md` - Sanity-check the running stack with the bundled sample dataset\n- `skills\u002Famc-run-video-calibration\u002FSKILL.md` - Calibrate from your own pre-recorded MP4s via REST API\n\n\u003C!-- signing marker -->\n","# Skill: Launch AutoMagicCalib Release Containers\n\n## When to Use This Skill\n\nActivate this skill when the user wants to bring up the AutoMagicCalib stack. Typical prompts:\n\n- \"deploy auto calibration\"\n- \"launch auto calibration\" \u002F \"launch AMC\"\n- \"start MS+UI\" \u002F \"bring up the microservice and UI\"\n- \"set up auto-magic-calib\"\n\nPrerequisite: NGC API key (the skill prompts for it). For first-time setup on a fresh machine, this also resolves or clones the `auto-magic-calib` repo after explicit user confirmation.\n\n## Overview\nLaunch the AutoMagicCalib microservice (MS) and UI from pre-built release images via Docker Compose. Use this for production deployments or when you want to run the full stack (backend + web UI) without building images locally.\n\n## Prerequisites\n- Docker and Docker Compose installed\n- NVIDIA Docker Runtime configured (for GPU support)\n- `auto-magic-calib` repo on disk — Step 0b auto-resolves an existing checkout via `git rev-parse --show-toplevel`, or offers to clone `https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib` into `~\u002Fauto-magic-calib` if none is found\n- NGC account with access to NVIDIA container registry\n- **Docker must be runnable without `sudo`** — verify with `docker ps`. If it fails, ask the user to follow the post-install steps: https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002Flinux-postinstall\u002F\n\n> **If `docker ps` fails with \"permission denied\"**, ask the user to run:\n> ```bash\n> sudo usermod -aG docker $USER\n> newgrp docker   # applies group change in current shell without logout\n> ```\n> Then verify with `docker ps` before continuing.\n\n## Instructions\n\n### Step 0: Verify Docker Runs Without sudo\n\n```bash\ndocker ps\n```\n\n- If it succeeds → continue.\n- If it fails with \"permission denied\" → the user is not in the `docker` group. Ask the user to run:\n  ```bash\n  sudo usermod -aG docker $USER && newgrp docker\n  ```\n  Then ask the user to confirm `docker ps` works before continuing.\n\n> **Agent note**: If `docker ps` cannot be run from within the agent sandbox, ask the user to confirm it works (e.g. \"Can you confirm `docker ps` runs without sudo?\") before proceeding.\n\n### Step 0b: Resolve Repo Checkout\n\nThe skill needs `compose\u002F`, `assets\u002Fsdg_08_2_sample_data_010926.zip`, and a `models\u002F` mount point — all of which live in the `auto-magic-calib` repo. If you're already inside a checkout, the skill uses it. Otherwise it asks for explicit user confirmation using the host's question mechanism; if none is available, ask in chat and wait before cloning into `~\u002Fauto-magic-calib`.\n\n```bash\nREPO_URL=\"https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib.git\"\nDEFAULT_CLONE_DIR=\"$HOME\u002Fauto-magic-calib\"\n\n# 1. Already inside a usable checkout?\nREPO_ROOT=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull)\"\nif [ -n \"$REPO_ROOT\" ] && [ -f \"$REPO_ROOT\u002Fcompose\u002Fcompose.yml\" ] && [ -d \"$REPO_ROOT\u002Fassets\" ]; then\n  echo \"✓ Using existing checkout: $REPO_ROOT\"\n\n# 2. Default clone dir already populated from a prior run?\nelif [ -d \"$DEFAULT_CLONE_DIR\u002F.git\" ] && [ -f \"$DEFAULT_CLONE_DIR\u002Fcompose\u002Fcompose.yml\" ]; then\n  REPO_ROOT=\"$DEFAULT_CLONE_DIR\"\n  echo \"✓ Found existing clone at $REPO_ROOT\"\n\n# 3. Nothing on disk — STOP and ask the user for confirmation using the\n#    host's question mechanism; if none is available, ask in chat and wait.\n#    Do NOT clone silently from this block.\nelse\n  echo \"No auto-magic-calib checkout found. Ask the user for confirmation:\"\n  echo \"  Clone $REPO_URL into $DEFAULT_CLONE_DIR? [y\u002FN]\"\n  echo \"On 'y' — run: git clone \\\"$REPO_URL\\\" \\\"$DEFAULT_CLONE_DIR\\\"\"\n  exit 1\nfi\n\ncd \"$REPO_ROOT\"\nexport REPO_ROOT\necho \"REPO_ROOT=$REPO_ROOT\"\n```\n\n> **Agent note**: do **not** clone silently. Ask the user first using the host's question mechanism; if none is available, ask in chat and wait. Example: \"auto-magic-calib repo not found. Clone `https:\u002F\u002Fgithub.com\u002FNVIDIA-AI-IOT\u002Fauto-magic-calib` into `~\u002Fauto-magic-calib`? (or provide your own path)\". Honour an alternate path if the user offers one. The clone is a few hundred MB (compose files + sample dataset + assets).\n\n### Step 0c: Install Python venv (New Systems Only)\n\nOn a fresh system, `pip` and `python3-venv` may not be available. Install them first:\n\n```bash\n# Create a venv for HuggingFace CLI (project-local preferred)\nREPO_DIR=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)\"\nHF_VENV=\"${REPO_DIR}\u002Fvenv\"\npython3 -m venv \"$HF_VENV\" 2>\u002Fdev\u002Fnull || {\n  echo \"ERROR: python3-venv not available.\" >&2\n  echo \"Install it manually: sudo apt install -y python3-venv python3-pip\" >&2\n  exit 1\n}\n\n# Install HuggingFace hub (needed for VGGT download)\n\"$HF_VENV\u002Fbin\u002Fpip\" install --upgrade pip huggingface_hub\n```\n\n> **Note**: Skip this step if a venv with `hf` already exists (check `venv\u002Fbin\u002Fhf` in the repo root or `~\u002Fvenv\u002Famc\u002Fbin\u002Fhf`).\n\n### Step 1: Login to NGC\n\nAsk the user for their NGC API key using the host's question mechanism; if none is available, ask in chat and wait. Then run:\n\n```bash\necho \"\u003CNGC_API_KEY>\" | docker login nvcr.io --username '$oauthtoken' --password-stdin\necho \"✓ NGC authentication complete\"\n```\n\n### Step 2: Download VGGT Model (If Not Already Present)\n\n```bash\nexport REPO_ROOT=$(git rev-parse --show-toplevel)\ncd \"$REPO_ROOT\"\n\nif [ -f \"models\u002Fvggt\u002Fvggt_1B_commercial.pt\" ]; then\n  echo \"✓ VGGT model already present\"\nelse\n  echo \"✗ VGGT model not found\"\n  echo \"Options:\"\n  echo \"  1. Continue without VGGT (AMC only - sufficient for most use cases)\"\n  echo \"  2. Download VGGT model (~4.7GB, requires HuggingFace account)\"\nfi\n```\n\n**To download VGGT** (ask the user for a HuggingFace token using the host's question mechanism; if none is available, ask in chat and wait):\n\n**Step 2a: Accept Model License** (required, one-time):\n1. Visit: https:\u002F\u002Fhuggingface.co\u002Ffacebook\u002FVGGT-1B-Commercial\n2. Log in to your HuggingFace account\n3. Click \"Agree and access repository\" to accept license terms\n\n**Step 2b: Get HuggingFace Token**:\n1. Visit: https:\u002F\u002Fhuggingface.co\u002Fsettings\u002Ftokens\n2. Create new token with \"Read\" access (starts with `hf_...`)\n3. Ask the user for the token using the host's question mechanism; if none is available, ask in chat and wait. Do NOT ask them to run a command\n\n**Step 2c: Download** (pass token via `HF_TOKEN` env var — keeps the secret out of `hf`'s argv \u002F `ps aux`; no interactive login needed):\n```bash\nREPO_DIR=\"$(git rev-parse --show-toplevel 2>\u002Fdev\u002Fnull || pwd)\"\ncd \"$REPO_DIR\"\n\n# Find the HuggingFace CLI binary (named 'hf', not 'huggingface-cli')\nHF_BIN=\"$(find \"$REPO_DIR\u002Fvenv\" ~\u002Fvenv\u002Famc -name hf -type f 2>\u002Fdev\u002Fnull | head -1)\"\n{ [ -z \"$HF_BIN\" ] || [ ! -x \"$HF_BIN\" ]; } && { echo \"ERROR: hf binary not found or not executable; install the hf CLI (Step 0c) or set HF_BIN\" >&2; exit 1; }\n\n# Do NOT use --token on the command line (leaks via ps\u002Fargv). The HF CLI\n# reads HF_TOKEN from the environment automatically.\nHF_TOKEN=\"\u003CHF_TOKEN>\" \"$HF_BIN\" download facebook\u002FVGGT-1B-Commercial \\\n  --local-dir models\u002Fvggt\u002F\n\n# Verify\nls -lh models\u002Fvggt\u002Fvggt_1B_commercial.pt\n# Should show ~4.7GB file\n```\n\n> **Important**: Download BEFORE setting `chown 1000:1000` on the models directory — the current user needs write access during download. Set permissions in Step 4 after download completes.\n\n### Step 3: Configure .env Variables\n\nThe `.env` file at `compose\u002F.env` controls ports and paths. Update it before launching:\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Find available backend port (8000-8009)\nfor port in {8000..8009}; do\n  if ! lsof -Pi :$port -sTCP:LISTEN -t >\u002Fdev\u002Fnull 2>&1; then\n    MS_PORT=$port\n    echo \"Using backend port: $MS_PORT\"\n    break\n  fi\ndone\n[ -z \"$MS_PORT\" ] && { echo \"ERROR: no free backend port in 8000-8009; free one or widen the range.\" >&2; exit 1; }\n\n# Find available UI port (5000-5009)\nfor port in {5000..5009}; do\n  if ! lsof -Pi :$port -sTCP:LISTEN -t >\u002Fdev\u002Fnull 2>&1; then\n    UI_PORT=$port\n    echo \"Using UI port: $UI_PORT\"\n    break\n  fi\ndone\n[ -z \"$UI_PORT\" ] && { echo \"ERROR: no free UI port in 5000-5009; free one or widen the range.\" >&2; exit 1; }\n\n# Get host IP\nHOST_IP=$(hostname -I | awk '{print $1}')\necho \"Host IP: $HOST_IP\"\n\n# Update .env — preserve any keys the user has already set; back up first.\n# .env may contain credentials, so restrict permissions on both the backup and\n# the live file (chmod 600). Add `compose\u002F.env.bak.*` to .gitignore.\nENV_FILE=\".env\"\nif [ -f \"$ENV_FILE\" ]; then\n  BACKUP=\"${ENV_FILE}.bak.$(date +%s)\"\n  cp \"$ENV_FILE\" \"$BACKUP\"\n  chmod 600 \"$BACKUP\"\nfi\ntouch \"$ENV_FILE\"\nchmod 600 \"$ENV_FILE\"\nset_env_key() {\n  local k=\"$1\" v=\"$2\"\n  if grep -qE \"^${k}=\" \"$ENV_FILE\"; then\n    sed -i \"s|^${k}=.*|${k}=${v}|\" \"$ENV_FILE\"\n  else\n    echo \"${k}=${v}\" >> \"$ENV_FILE\"\n  fi\n}\nset_env_key AUTO_MAGIC_CALIB_MS_PORT \"${MS_PORT}\"\nset_env_key AUTO_MAGIC_CALIB_UI_PORT \"${UI_PORT}\"\nset_env_key PROJECT_DIR \"..\u002F..\u002Fprojects\"\nset_env_key MODEL_DIR \"..\u002F..\u002Fmodels\"\nset_env_key HOST_IP \"${HOST_IP}\"\n\n# Keep timestamped .env backups out of git.\nGITIGNORE=\"$REPO_ROOT\u002F.gitignore\"\ntouch \"$GITIGNORE\"\ngrep -qxF \"compose\u002F.env.bak.*\" \"$GITIGNORE\" || echo \"compose\u002F.env.bak.*\" >> \"$GITIGNORE\"\n\necho \"✓ .env updated\"\ncat .env\n```\n\n**Important**: `HOST_IP` must be the machine's network IP (not `localhost`) so the UI container can reach the backend from a browser.\n\nOptional: set `VGGT_MODEL_PATH` only if the VGGT model is mounted at a non-default container path; default is `\u002Ftmp\u002Fvggt_model\u002Fvggt_1B_commercial.pt` inside the MS container.\n\n### Step 4: Set Directory Permissions\n\nThe containers run as UID\u002FGID 1000. The `projects` and `models` directories must be owned by this UID for containers to read\u002Fwrite properly:\n\n```bash\ncd \"$REPO_ROOT\"\n\n# Create projects directory if it doesn't exist\nmkdir -p projects\n\n# Set ownership (required for containers to write calibration outputs).\n# Do this AFTER VGGT download is complete (current user needs write access during download).\n# Get explicit user confirmation before running sudo chown — it recursively changes\n# ownership of $REPO_ROOT\u002Fprojects and $REPO_ROOT\u002Fmodels to UID\u002FGID 1000.\n[ -d projects ] && [ -d models ] || {\n  echo \"ERROR: expected projects\u002F and models\u002F under $REPO_ROOT\" >&2; exit 1;\n}\necho \"About to chown -R 1000:1000 on:\"\necho \"  $REPO_ROOT\u002Fprojects\"\necho \"  $REPO_ROOT\u002Fmodels\"\necho \"(required because containers run as UID 1000). Confirm before proceeding.\"\nsudo chown 1000:1000 -R projects\nsudo chown 1000:1000 -R models\n\necho \"✓ Permissions set\"\n```\n\n### Step 5: Launch Services\n\nBefore pulling, fail fast if the NGC key authenticated in Step 1 but cannot actually access a release image — otherwise `docker compose up` aborts partway with a 401\u002F403 after some work is already done.\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Fail-fast image-access check: confirm the NGC key can reach every release\n# image BEFORE pulling. `docker manifest inspect` checks registry access without\n# downloading layers, and the image list is read from the resolved compose so it\n# tracks the release tag automatically.\nIMAGES=$(docker compose config --images | sort -u)\n[ -z \"$IMAGES\" ] && { echo \"ERROR: no images resolved from compose — check compose\u002F.env and the chosen profile.\" >&2; exit 1; }\nfor img in $IMAGES; do\n  echo \"Checking access: $img\"\n  if ! docker manifest inspect \"$img\" >\u002Fdev\u002Fnull 2>&1; then\n    echo \"NGC login succeeded, but this key cannot access the required image:\" >&2\n    echo \"  $img\" >&2\n    echo \"Provide an NGC key with access to this image's namespace, then re-run Step 1 (login) and retry.\" >&2\n    exit 1\n  fi\ndone\n\n# Start all services (images pulled automatically on first run)\ndocker compose up -d\n\n# Check containers are running\ndocker compose ps\n```\n\n**Expected output**:\n```\nNAME                    IMAGE                                                              STATUS\nauto-magic-calib-ms-1   nvcr.io\u002Fnvidia\u002Fauto-magic-calib:2.0.0           Up (healthy)\nauto-magic-calib-ui-1   nvcr.io\u002Fnvidia\u002Fauto-magic-calib-ui:2.0.0        Up\n```\n\n### Step 6: Verify Services Are Running\n\n```bash\n# Read ports from .env\nMS_PORT=$(grep AUTO_MAGIC_CALIB_MS_PORT $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\nUI_PORT=$(grep AUTO_MAGIC_CALIB_UI_PORT $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\nHOST_IP=$(grep HOST_IP $REPO_ROOT\u002Fcompose\u002F.env | cut -d= -f2)\n\n# Wait for microservice readiness. Cold image pulls or first startup can need\n# extra time after `docker compose up -d` returns.\nREADY_URL=\"http:\u002F\u002Flocalhost:${MS_PORT}\u002Fv1\u002Fready\"\necho \"Waiting for microservice readiness at ${READY_URL} ...\"\nready_response=\"\"\nfor attempt in $(seq 1 24); do\n  if ready_response=$(curl -fsS --max-time 5 \"${READY_URL}\" 2>\u002Fdev\u002Fnull) && \\\n     echo \"${ready_response}\" | grep -q '\"code\"[[:space:]]*:[[:space:]]*0'; then\n    echo \"Microservice ready: ${ready_response}\"\n    break\n  fi\n  if [ \"${attempt}\" -lt 24 ]; then\n    printf \"  [%02d\u002F24] Microservice not ready yet; retrying in 5s...\\n\" \"${attempt}\"\n    sleep 5\n  fi\ndone\n\nif ! echo \"${ready_response}\" | grep -q '\"code\"[[:space:]]*:[[:space:]]*0'; then\n  echo \"ERROR: microservice did not report ready within 120 seconds: ${READY_URL}\" >&2\n  echo \"Check status and logs:\" >&2\n  echo \"  cd ${REPO_ROOT}\u002Fcompose && docker compose ps\" >&2\n  echo \"  cd ${REPO_ROOT}\u002Fcompose && docker compose logs auto-magic-calib-ms\" >&2\n  exit 1\nfi\n\n# Check UI is serving\nUI_STATUS=$(curl -s -o \u002Fdev\u002Fnull -w \"%{http_code}\" --max-time 5 \"http:\u002F\u002Flocalhost:${UI_PORT}\")\nif [ \"${UI_STATUS}\" != \"200\" ]; then\n  echo \"ERROR: Web UI returned HTTP ${UI_STATUS}; check docker compose ps and UI logs.\" >&2\n  exit 1\nfi\necho \"Web UI ready: HTTP ${UI_STATUS}\"\n\necho \"Microservice: http:\u002F\u002F${HOST_IP}:${MS_PORT}\"\necho \"Web UI:       http:\u002F\u002F${HOST_IP}:${UI_PORT}\"\n```\n\n## Success Criteria\n\n**Both containers running** — see `docker compose ps` output from Step 5. Both services should show \"Up\" status; microservice should show \"(healthy)\" in the STATUS column.\n\n**Microservice healthy** — Step 6 readiness polling returns `code:0` from `\u002Fv1\u002Fready` and prints the service URL.\n\n**Web UI accessible**:\n- Open browser: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_UI_PORT>`\n- Should display the AutoMagicCalib web interface\n- Should be able to create projects and run calibration\n\n## Key Output\n\n**Microservice**: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_MS_PORT>` (default port 8000)\n- API docs: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_MS_PORT>\u002Fdocs`\n\n**Web UI**: `http:\u002F\u002F\u003CHOST_IP>:\u003CAUTO_MAGIC_CALIB_UI_PORT>` (default port 5000)\n- Interactive project management\n- File upload interface\n- Calibration configuration\n- Real-time status monitoring\n- Results visualization and download\n\n**Data Persistence**:\n- Projects stored: `$REPO_ROOT\u002Fprojects\u002F`\n- State persisted: `$REPO_ROOT\u002Fprojects\u002Fstate.json`\n\n## Troubleshooting\n\n| Issue | Symptoms | Solution |\n|-------|----------|----------|\n| `pip` not found | `pip: command not found` | Run `sudo apt install -y python3.12-venv python3-pip` then create venv (Step 0) |\n| `huggingface-cli` not found | `huggingface-cli: command not found` | The binary is named `hf` in the venv. Find it with: `find venv ~\u002Fvenv\u002Famc -name hf -type f 2>\u002Fdev\u002Fnull \\| head -1` |\n| `python3 -m venv` fails | \"ensurepip not available\" | Run `sudo apt install -y python3.12-venv` first |\n| Docker permission denied | \"permission denied while trying to connect to docker socket\" | User not in docker group — ask user to run: `sudo usermod -aG docker $USER && newgrp docker`. See: https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002Flinux-postinstall\u002F |\n| `docker login` itself rejected | Step 1 login returns an authentication error | The key is invalid or expired. Ask the user for a current NGC key and log in again before continuing. |\n| Key logs in but can't access an image | The Step 5 access check stops with \"cannot access the required image\" \u002F a 401\u002F403 on `docker manifest inspect`, before any pull | The key authenticates but lacks access to that image's namespace. Re-running login with the same key won't help — ask the user for an NGC key with access to the required namespace, re-run Step 1, then retry Step 5. |\n| VGGT download permission error | \"PermissionError: [Errno 13] Permission denied: 'models\u002Fvggt\u002F.cache'\" | Download VGGT BEFORE setting `chown 1000:1000` on models. Fix: `sudo chown -R $(id -u):$(id -g) models` then re-download |\n| Port already in use | \"address already in use\" | Find available port in 8000-8009 (MS) or 5000-5009 (UI); update `.env` |\n| Microservice readiness timeout | Step 6 reports that `\u002Fv1\u002Fready` did not return `code:0` within 120 seconds | Run `cd $REPO_ROOT\u002Fcompose && docker compose ps`, then inspect logs with `docker compose logs auto-magic-calib-ms` |\n| Permission denied (projects) | \"Permission denied: 'projects\u002F...'\" in MS logs | Run: `sudo chown 1000:1000 -R projects` |\n| Permission denied (models) | \"Permission denied: 'models\u002F...'\" in MS logs | Run: `sudo chown 1000:1000 -R models` |\n| UI can't reach backend | Browser shows connection error | Verify `HOST_IP` in `.env` is the machine's network IP, not `localhost` |\n| VGGT model not found | Warning in MS logs about missing VGGT | Download model (Step 2) or ignore (AMC works without VGGT) |\n| Container exits immediately | Status \"Exited\" | Check logs: `docker compose logs auto-magic-calib-ms` |\n| GPU not available | \"CUDA not available\" in logs | Check NVIDIA runtime: `docker run --rm --runtime=nvidia --gpus all ubuntu:20.04 nvidia-smi` |\n| \"No such file or directory\" when verifying | After launch, can't find compose directory | Working directory persists after `cd compose`. Use absolute paths or run `cd $REPO_ROOT` first |\n\n**Common Fixes**:\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# View logs\ndocker compose logs -f\n\n# View logs for specific service\ndocker compose logs -f auto-magic-calib-ms\n\n# Restart all services\ndocker compose restart\n\n# Stop and remove containers\ndocker compose down\n\n# Update .env and relaunch\ndocker compose up -d\n```\n\n## Stopping the Services\n\n```bash\ncd $REPO_ROOT\u002Fcompose\n\n# Stop all services (containers removed, data persisted)\ndocker compose down\n\n# Stop and remove volumes\ndocker compose down -v\n```\n\n## Related Skills\n- `skills\u002Famc-run-sample-calibration\u002FSKILL.md` - Sanity-check the running stack with the bundled sample dataset\n- `skills\u002Famc-run-video-calibration\u002FSKILL.md` - Calibrate from your own pre-recorded MP4s via REST API\n\n\u003C!-- signing marker -->\n",{"name":5,"owner":24,"license":15,"service":25,"version":26,"metadata":27,"reviewed":35,"description":7,"data_classification":36},"NVIDIA CORPORATION","auto-magic-calib","1.0.0",{"tags":28},[29,30,31,32,33,34],"amc","deepstream","docker","calibration","setup","ngc","2026-04-28","public","main","1c523aba5cdf867a79f078847075492d7778092a","skills\u002Fnvidia\u002Famc-setup-calibration-stack",1,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills","https:\u002F\u002Fskills.x402.wtf","https:\u002F\u002Fgithub.com\u002Fthemusebook","https:\u002F\u002Fmusebook.trade","The official Github of musebook","2026-11-02 23:42:26",{"source":8,"skillName":5,"commands":48,"prompt":58,"zipAvailable":59},[49,52,55],{"tool":50,"command":51},"npx","npx skills add themusebook\u002Fskills --skill amc-setup-calibration-stack",{"tool":53,"command":54},"bunx","bunx skills add themusebook\u002Fskills --skill amc-setup-calibration-stack",{"tool":56,"command":57},"pnpm","pnpm dlx skills add themusebook\u002Fskills --skill amc-setup-calibration-stack","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002FSKILL.md",true,[61,67,71,75],{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"BENCHMARK.md","REFERENCE","参考文档",4489,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002FBENCHMARK.md",{"path":68,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":69,"sourceHtmlUrl":70},"evals\u002Fevals.json",7113,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002Fevals\u002Fevals.json",{"path":72,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":73,"sourceHtmlUrl":74},"skill-card.md",3737,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002Fskill-card.md",{"path":76,"resourceType":77,"resourceTypeDesc":78,"sizeBytes":79,"sourceHtmlUrl":80},"skill.oms.sig","OTHER","其他",4681,"https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fnvidia\u002Famc-setup-calibration-stack\u002Fskill.oms.sig",[82,91,98,105,112,119],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":87,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7AB34SPKVSDRRG009PM1CZH","youtube-clipper","YouTube 视频智能剪辑工具。下载视频和字幕，AI 分析生成精细章节（几分钟级别）， 用户选择片段后自动剪辑、翻译字幕为中英双语、烧录字幕到视频，并生成总结文案。 使用场景：当用户需要剪辑 YouTube 视频、生成短视频片段、制作双语字幕版本时。 关键词：视频剪辑、YouTube、字幕翻译、双语字幕、视频下载、clip video","MIT","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fyoutube-clipper\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M7AB18BH6N7KR7939HRNWVMQ","viem-integration","Integrate EVM blockchains using viem. Use when user says \"read blockchain data\", \"send transaction\", \"interact with smart contract\", \"connect to Ethereum\", \"use viem\", \"use wagmi\", \"wallet integration\", \"viem setup\", or mentions blockchain\u002FEVM development with TypeScript.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fviem-integration\u002FSKILL.md",[],[],{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M7AB12GHGP48WCNXDQQ9222H","v4-security-foundations","Security-first Uniswap v4 hook development. Use when user mentions \"v4 hooks\", \"hook security\", \"PoolManager\", \"beforeSwap\", \"afterSwap\", or asks about V4 hook best practices, vulnerabilities, or audit requirements.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-security-foundations\u002FSKILL.md",[],[],{"skillNo":106,"name":107,"nameZh":6,"description":108,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":6,"sourceHtmlUrl":109,"categories":110,"keywords":111,"publishedAt":19},"01M7AB0XM3PSR7V3Q2ZRR0Q0FB","v4-sdk-integration","App-layer SDK guide for building swap and liquidity experiences directly with the Uniswap v4 SDK. Use when user asks about \"v4 sdk\", \"uniswap v4\", \"v4 swap\", \"v4 liquidity\", \"PoolManager\", \"V4Planner\", \"StateView\", \"PositionManager\", \"pool state\", \"v4 position\", \"uniswap sdk\", or when building swap\u002Fliquidity UX directly with SDKs rather than via the Trading API.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-sdk-integration\u002FSKILL.md",[],[],{"skillNo":113,"name":114,"nameZh":6,"description":115,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":6,"sourceHtmlUrl":116,"categories":117,"keywords":118,"publishedAt":19},"01M7AB0SGA93ASA588DJW4P9WT","v4-hook-generator","Generate Uniswap v4 hook contracts via OpenZeppelin MCP. Use when building custom swap logic, async swaps, hook-owned liquidity, custom curves, dynamic fees, MEV protection, limit orders, or oracle hooks.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fv4-hook-generator\u002FSKILL.md",[],[],{"skillNo":120,"name":121,"nameZh":6,"description":122,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":86,"licenseFileUrl":6,"sourceHtmlUrl":123,"categories":124,"keywords":125,"publishedAt":19},"01M7AAZYH75H4CS03WSW2EZ8DM","testing","For testing with Light Protocol programs and clients on localnet, devnet, and mainnet validation.","https:\u002F\u002Fgithub.com\u002Fthemusebook\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Ftesting\u002FSKILL.md",[],[]]