cdf-image-attribute-augmentation

name: cdf-image-attribute-augmentationdescription: Run the CDF Orchestration Image Attribute Augmentation DAG on Kubernetes - person-crop clothing augmentation, attribute search, and augmented dataset generation. Select for requests about image attribute augmentation, person attribute search, person re-identification data, clothing augmentation, attribute captions, augmentation payloads, run status, or result retrieval. Runs environment setup first when controller readiness is unknown. Not for video or defect-image generation.version: "1.0.1"license: CC-BY-4.0 AND Apache-2.0metadata:   owner: NVIDIA  service: cosmos-data-factory  version: "1.0.1"  reviewed: "2026-10-09"  author: NVIDIA  tags:     - physical-ai    - cdf-orchestration    - image-attribute-augmentation    - cosmos

CDF Orchestration — Image Attribute Augmentation

Run the Image Attribute Augmentation DAG end to end: person-crop input preparation, cosmos image-edit augmentation, cosmos post-processing, event and person attribute search, augmented dataset generation, and result retrieval.

Component details

This workflow skill and its references are the primary guide for configuring and running the workflow. Consult a component skill only when the requested configuration or diagnosis requires component-specific details not covered here. Routine payload authoring with documented fields, preflight, triggering, monitoring, and result retrieval do not require component skill access.

When additional component guidance is needed, consult that component's documentation, and do not invent keys or values. The links below are pinned to the latest main commit of each component repository as of 2026-10-20; a component skill installed separately may describe a different version.

Use only documentation at these commits; do not substitute a moving branch or another version.

If the repository is inaccessible, continue tasks supported by this workflow's guidance. Identify the unavailable reference and the unresolved component-specific detail only when that missing guidance prevents completing the requested work; do not invent component settings or treat component access as a prerequisite for ordinary workflow use.

Apply component guidance through this workflow's supported payload fields and DAG constraints. Component configuration and the Airflow trigger payload are different contracts; component capabilities do not automatically become supported DAG options. Keep execution through this workflow's existing procedure. If the pinned guidance differs from the deployed component version or DAG integration, report the mismatch before proposing the affected settings.

DAG selection

The workflow builds one DAG per compute platform from airflow/dags/workflows/image_attribute_augmentation_dag/:

Kubernetes is the only platform whose manifest is checked in, so image_attribute_augmentation_dag_k8s is the only DAG this repository registers. A DAG is registered only if its manifest exists; a missing manifest means the DAG is absent from Airflow rather than broken. List the DAGs Airflow actually loaded before triggering, and never name a DAG ID that is not in that list.

There is a single end-to-end pipeline — there are no augmentation-only or labeling-only DAG variants. If a user asks for augmentation without attribute search, tell them the checked-in DAG does not offer that flow rather than inventing a DAG ID.

Manual payload entry in the Airflow UI

If the user wants to enter their own payload directly in the Airflow UI rather than have you construct and trigger one, your job is limited to getting them to the UI: confirm controller readiness, ensure make port-forward is running (see airflow-direct-api.md), and report the reachable URL. Do not render a payload, run preflight, or trigger a run yourself in this case — the user is doing that from the UI. Resume monitoring (step 6 below) once they tell you a run has been triggered; you can find it via the Airflow API without needing the payload they used.

Scope

Before building any payload, collect all of the following from the user. Never invent an endpoint URL or bucket path — use only values the user supplied.

If the user does not provide a required value, ask for it explicitly before proceeding. Never invent one. Values the user supplied earlier in the session, including for a previous run, may be reused; list each reused value in the payload shown for confirmation in step 3, so the user confirms it again before anything is submitted.

Always run the following readiness checks before triggering a run. The checks are short-circuiting — stop at the first failure and route to the environment-setup skill immediately.

Before any check, establish the cluster connection. The cluster is reached only through credentials the user supplies — they are never part of the repository. Check whether the cluster credential file path is already exported in the shell environment; if not, ask the user for the absolute path before running any cluster command. Never assume a path or fall back to any on-disk default — see setup-and-preflight.md for the full procedure.

The controller (Airflow) and DAG compute tasks run on the same cluster unless a different remote cluster connection was configured. GPU capacity is checked on this cluster.

1. Controller pods — check that the Airflow controller pods (not DAG task pods) are Running. DAG task pods in Pending or Failed state are normal and must not be mistaken for controller failures:

   kubectl get pods -n sdg-workflow -l "release=sdg-workflow-controller"
   

   All pods matching the release=sdg-workflow-controller label must be Running. If the namespace is absent, this is a first-install condition — route to the environment-setup skill, do not diagnose further.

2. Airflow API — reachable only if check 1 passes. First establish AIRFLOW_URL from the Kubernetes ClusterIP. That works from a cluster node; from a workstation, fall back to a port-forward as described in airflow-direct-api.md:

   AIRFLOW_URL="http://$(kubectl get svc -n sdg-workflow \
     sdg-workflow-controller-api-server \
     -o jsonpath='{.spec.clusterIP}'):8080"
   

   Then confirm the target DAG is loaded and is_paused: False. See airflow-direct-api.md for the full auth + check sequence. If the API is unreachable, route to the environment-setup skill.

3. Pools — only if check 2 passes. Derive the required pool names from the manifest; do not assume them. Pool names are deployment-specific: they come from the manifest's deployment.profiles.*.pool and are created by deploy/values.yaml's airflowPools.pools, so a customer on different hardware or with renamed pools will not match whatever this repository ships today. See airflow-direct-api.md for the derivation snippet, which also adds the mode-dependent augmentation pool the DAG selects at runtime (external_image_edit_service_pool for external, default_pool for internal — set in the DAG's CosmosTaskGroup(...) wiring, not in the manifest). On the currently checked-in manifest this resolves to default_pool, k8s_rtx6000_1, and external_image_edit_service_pool. Every derived pool must exist with open slots.

4. Compute-cluster GPUs — check the cluster (using the cluster connection established above):

   # Per node: allocatable GPUs minus GPUs requested by pods in every namespace.
   kubectl get nodes -o json > /tmp/iaa-nodes.json
   kubectl get pods -A --field-selector=status.phase!=Succeeded,status.phase!=Failed \
     -o json > /tmp/iaa-pods.json
   python3 - <<'PY'
   import json
   alloc = {n["metadata"]["name"]: int(n["status"].get("allocatable", {}).get("nvidia.com/gpu", 0))
            for n in json.load(open("/tmp/iaa-nodes.json"))["items"]}
   used = dict.fromkeys(alloc, 0)
   for p in json.load(open("/tmp/iaa-pods.json"))["items"]:
       node = p["spec"].get("nodeName")
       if node in used:
           for c in p["spec"].get("containers", []):
               used[node] += int(c.get("resources", {}).get("requests", {}).get("nvidia.com/gpu", 0))
   for node in sorted(alloc):
       print(f"{node}: {alloc[node] - used[node]} free / {alloc[node]} allocatable")
   print("cluster free GPUs:", sum(alloc.values()) - sum(used.values()))
   PY
   

   The compute cluster is shared — other users' runs and other namespaces may hold GPUs. Report GPUs as free-versus-total from this calculation, not just allocatable. Pending pods are not counted because they hold no node yet; mention any that are waiting for GPUs, since they will compete for the next free one. External mode needs no GPUs for inference — every task pod (augmentation, cosmos_post_processing, event_and_person_attribute_search) runs on a CPU profile, unlike EVG's k8s_gpu_task-profiled auto-labeling stages.

   Internal mode deploys two GPU endpoints, not three. service_lifecycle has exactly two services — vlm_service and image_edit_service — because one deployed Qwen serves both the multimodal and the text-only traffic: in internal mode the payload model assigns llm_service_url = vlm_service_url and llm_model = vlm_model. There is no separate llm_service endpoint to deploy. So one replica of each needs two allocatable GPUs, more only if a replicas value is raised. Derive this from the manifest rather than trusting this number — count components whose deployment_profile resolves to a profile with configuration.gpu: true, summing gpu_count where set (default 1).

5. Stale failed pods — before triggering, check for accumulated failed pods in the compute namespace and report them. They are retained by design and do not affect run correctness, but they consume namespace quota and clutter log searches:

   kubectl get pods -n sdg-workflow \
     --field-selector=status.phase=Failed \
     -o custom-columns='NAME:.metadata.name,AGE:.metadata.creationTimestamp,DAG:.metadata.labels.dag_id'
   

   Clean up only pods whose dag_id label matches a run you own, after confirming with the user.

Document each check result explicitly.

If any check fails: invoke the environment-setup skill automatically — do not wait for the user to say "set up" or ask them to name the skill.

If the user's request implies first-time or explicit deployment ("deploy", "install", "set up", "reinstall", "redeploy", "full setup"): invoke the environment-setup skill even if all checks pass, and confirm the planned commands first.

If all checks pass and the user only wants to run the workflow: proceed directly to payload and trigger.

Bundled tools

• scripts/upload_images.py: validate/upload local <person_id>/<image>.(jpg|jpeg|png) data.
• scripts/payload.py: render or validate a standalone ImageAttributeAugmentationDagPayloadConfig-compatible JSON.
• scripts/payload_schema.py: payload field allowlists and defaults imported by payload.py. It is a module, not an entry point; do not run it directly.
• scripts/summarize_results.py: summarize a downloaded augmented_data.json dataset.
• scripts/workflow.py: drive the SDG webserver API — submit a run, poll its status, retrieve results, or cancel a single named run by ID (cancels only that run; does not touch cluster resources or other runs). Requires WEBSERVER_ENDPOINT and NGC_API_KEY. Prefer the Airflow API path below for normal operation.

Run the bundled scripts/ commands from this skill directory; they need only Python and, for uploads, boto3. The Airflow pool check reads the DAG manifest, so it needs the CDF Orchestration checkout: set REPO_ROOT to it, as shown in airflow-direct-api.md#preflight-direct-path. Credentials must be inherited from the shell that launched the agent; never ask the user to paste secret values into the prompt.

Procedure

1. Determine the input source.

   • For local data, validate before upload:

     python scripts/upload_images.py --path /path/to/crops --validate-only
     

   • Then upload while preserving the hierarchy:

     python scripts/upload_images.py \
       --path /path/to/crops --destination-path image-attribute-augmentation/my-run
     

   • For an existing storage URL, use it unchanged after confirming it contains person-ID subdirectories. Each immediate subdirectory of input_path is treated as one person ID, and its images are combined into a single horizontal strip per person.

2. Select service mode.

   • external requires explicit VLM, LLM, and image-edit endpoint URLs. The LLM and VLM URLs may legitimately differ here — only internal mode collapses them onto one endpoint.
   • internal lets the DAG's service lifecycle deploy the services in-cluster. It deploys two endpoints, not three: vlm_service and image_edit_service, with the single Qwen VLM also serving the text-only LLM traffic.
   • Choose service mode independently from controller placement. A local controller may use external inference endpoints.
   • Keep nested service mode and output directory consistent with the top level.
   • On Kubernetes each deployed endpoint claims one GPU from the manifest's GPU pool, so internal mode needs at least two allocatable GPUs (more if any replicas value is raised); external mode needs none for inference.

3. Read payload-contract.md, then render a payload from the values collected above. Every deployment-specific value — endpoint URLs, bucket paths, input paths — must be one the user supplied; never invent one.

   External:

   python scripts/payload.py render \
     --input-path s3://bucket/input/person-crops/ \
     --output-directory s3://bucket/output/image-attribute-augmentation/ \
     --service-mode external \
     --vlm-url https://vlm.example/v1 \
     --llm-url https://llm.example/v1 \
     --image-edit-url https://image-edit.example/v1 \
     --max-imgs 10 --num-augmentation 3 \
     --variable-distribution assets/variable-distribution.json \
     --output /tmp/iaa-payload.json
   

   Internal:

   python scripts/payload.py render \
     --input-path s3://bucket/input/person-crops/ \
     --output-directory s3://bucket/output/image-attribute-augmentation/ \
     --service-mode internal \
     --max-imgs 10 --num-augmentation 3 \
     --output /tmp/iaa-payload.json
   

   Add --cosmos-evaluator-url <arbitrator-root-url> to use an existing evaluator, or --checker-override metropolis.attribute_verification=false to turn the checker off for one run. The renderer and validate reject unknown fields and any distribution that omits one of the six clothing attributes; fix the payload instead of removing the field that was rejected.

   Show the user the rendered payload (or its validated contents) and get explicit confirmation before proceeding. Only continue to preflight and triggering if they confirm; if they want changes, re-render and re-confirm.

4. Preflight the DAG through the Airflow API. Check that the DAG is loaded, required pools have slots, and controller pods are healthy — see airflow-direct-api.md#preflight-direct-path. Confirm credential presence with set/unset only; never print values, lengths, prefixes, suffixes, hashes, or fingerprints. Then, from the repository root, confirm the deployment maps every bucket the payload reads or writes:

   python3 skills/cdf-orchestration-preflight/scripts/storage_check.py --payload /tmp/iaa-payload.json
   

   It must print Storage mapping: ready. An UNMAPPED line means the run would fail after its GPU services start; report the unmapped field and the configured prefixes to the user. Using that bucket requires changing the controller's storage configuration and redeploying it; with the user's authorization, route that change to orchestration setup, then rerun the check.

5. Submit exactly one DAG run. Pass the payload from step 3 as conf.payload — see airflow-direct-api.md#trigger-a-run for the full request shape and lost-response recovery. Record and return the DAG_RUN_ID, input path, output directory, and service mode.

6. Immediately after triggering — without waiting to be asked — monitor the run until it reaches a terminal state (success or failed). Poll the Airflow API every 60–120 seconds, with an upper bound so a run that never leaves queued or running is reported instead of polled forever:

   # Poll run state for up to 12 hours (720 polls x 60 s); raise MAX_POLLS for very large runs.
   MAX_POLLS=720
   for i in $(seq 1 "$MAX_POLLS"); do
     RESPONSE=$(curl -s --max-time 30 -H "Authorization: Bearer $TOKEN" \
       "$AIRFLOW_URL/api/v2/dags/$DAG_ID/dagRuns/$DAG_RUN_ID")
     STATE=$(RESPONSE="$RESPONSE" python3 -c "import json, os; print(json.loads(os.environ['RESPONSE']).get('state', 'unknown'))" 2>/dev/null || echo "unreachable")
     echo "$(date -u +%H:%M:%S) $STATE"
     case "$STATE" in success|failed) break ;; esac
     sleep 60
   done
   

   unreachable or unknown means the request failed (expired token, timeout, lost port-forward); fix the connection rather than counting it as progress. If the loop ends without a terminal state, or the run stays queued for more than about 30 minutes, stop polling and report the state with the per-task breakdown below: tasks waiting on pool slots or GPUs are the usual cause.

   For a per-task breakdown when state is running or failed, see airflow-direct-api.md.

   Stop polling as soon as the run state is success or failed. Use the polling loop that fits your runtime — a shell while loop, a background process, or a tool-native scheduler. Do not block the user waiting for each poll; report state changes as they occur.

   Tell the user they can also watch progress live in the Airflow UI. make port-forward runs in the foreground and never exits, so start it as a background job — and prefer that the user runs it in their own terminal, since an agent-owned forward dies with the session. Resolve the host's real address rather than reporting a placeholder or localhost, which is meaningless from another machine:

   HOST_IP=$(hostname -I | awk '{print $1}')
   echo "Airflow UI: http://$HOST_IP:8080"
   

   Use the configured Airflow login. The documented default login is defined in deploy/values.yaml under airflow.createUserJob.defaultUser (not webserver.defaultUser); deployments may override it. Do not include resolved credentials in chat or generated scripts. Change the default login before production use.

   To stop an in-progress run: open the Airflow UI, find the active DagRun, locate the running task, and mark it Failed (task menu → Mark Failed). This triggers the DAG's shutdown path, cleaning up Deployments, Services, and GPU pods. Do not delete the DagRun or the DAG — that bypasses cleanup and leaves stale cluster resources.

7. After the run reaches success or failed, if the user has not already requested or authorized downloading and analyzing the results, ask: "Would you like to download and analyze the results?" Wait for confirmation unless that authorization is already present in the session.

   When authorized, use whatever AWS credentials are already available in the shell environment (standard AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY / AWS_DEFAULT_REGION, an AWS profile, or instance role). Never ask the user to paste credentials into the prompt. Run artifacts live under <output_directory>/<run_id>/, where <output_directory> is the payload value and <run_id> is the dag_run_id from step 5. The final dataset is in augmented_dataset/; sync its contents into one local directory and summarize that directory:

   aws s3 sync "<output_directory>/<run_id>/augmented_dataset/" /tmp/iaa-results/
   python scripts/summarize_results.py --results-dir /tmp/iaa-results
   

   The summarizer reads the canonical augmented_data.json at the top of the dataset and ignores the per-scene copies under <scene>/sidecars/.

   To inspect intermediate augmented images instead, sync <output_directory>/<run_id>/cosmos/ and read output_metadata.json from each <person_id>/<augmentation_index>/ folder.

   Read outputs.md before interpreting files.

8. Report execution status and output quality separately. A success run state means every task finished; it does not mean every scene is complete. Before calling a run fully successful:

   • Fetch the generate_augmented_dataset task log (map_index=-1; see airflow-direct-api.md) and look for Missing optional output metadata and Ignoring ... cosmos skipped sidecar warnings. A missing metadata file leaves that scene with no selected attributes and generic person queries.
   • If results were downloaded, read the Quality warnings section from summarize_results.py. It lists entries without selected attributes, entries missing any clothing attribute, and skipped augmentations.

   If either check finds a problem, report the run as succeeded with quality warnings, list the affected entries, and do not describe the dataset as complete.

Guardrails

• Never invent endpoint URLs, bucket paths, or input paths — use only values the user supplied. Always ask for every deployment-specific value before building a payload. If a required value is missing, stop and ask — do not substitute a guess.
• Preserve explicit user inputs and endpoint/model selections throughout the session.
• Do not submit if payload validation, local dataset validation, the storage mapping check, or Airflow preflight fails.
• Do not show AWS credentials, Airflow bearer tokens, or S3 signed URLs.
• Report a file as saved only after a successful write and an existence check. Otherwise, describe the content as proposed and not saved.
• Do not start multiple runs unless the user explicitly requests them.
• Ask for a dataset location if none was supplied; this workflow has no implicit demo dataset.
• Do not invent augmentation-only or labeling-only DAG IDs — only the DAG listed above exists.
• Only offer a platform whose manifest exists and whose DAG is loaded in Airflow.

References

• Read setup-and-preflight.md for environment, storage, and policy requirements.
• Read payload-contract.md when creating or changing a payload.
• Read outputs.md when retrieving or interpreting results.
• Read troubleshooting.md after validation, API, or runtime errors.
• Read airflow-direct-api.md for all Airflow interactions: preflight, triggering, monitoring, per-task breakdown, and log retrieval.