i4h-workflow

por nvidia

Overview of `workflows/agentic/` (IsaacLab-Arena + GR00T/openpi). Use when the user asks what i4h workflow is, what's supported, or where to start.

npx skills add https://github.com/nvidia/skills --skill i4h-workflow

i4h Agentic Workflow

Purpose

Orient on the agentic workflow before touching a specific stage: which envs/robots/policies are supported, how the workflows/agentic/ subprojects fit together, and which per-stage skill to invoke next. This skill routes — it runs no pipeline stage itself.

Base Code

These steps drive the i4h-workflows base code (the workflows/agentic/ tree). To reuse an existing checkout, set I4H_WORKFLOWS to its path (no clone happens). Otherwise this resolves the current repo, or clones to ~/i4h-workflows — pick that default without prompting. Run every command below from the resolved root:

# Resolve the i4h-workflows base code (provides workflows/agentic/).
ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"
if [ ! -d "$ROOT/workflows/agentic" ]; then
  ROOT="${I4H_WORKFLOWS:-$HOME/i4h-workflows}"
  [ -d "$ROOT/workflows/agentic" ] || git clone https://github.com/isaac-for-healthcare/i4h-workflows "$ROOT"
fi
export I4H_WORKFLOWS="$ROOT"; cd "$ROOT"

Basics

  • Env YAMLs at workflows/agentic/config/environments/<env>.yaml are the source of truth.
  • Each pipeline stage has its own skill. Compose them or use [[i4h-workflow-e2e]] for full runs.
  • For env authoring or scene-edit source changes, load skills/i4h-workflow/references/repo-map.md before generating code.

Supported Envs

Fallback when --list-envs fails (setup incomplete):

EnvRobotPolicy
scissor_pick_and_placeSO-ARM 101GR00T N1.5 (N1.7 alternative)
locomanip_tray_pick_and_placeUnitree G1GR00T N1.6 (shared policy.locomanip.*)
locomanip_push_cartUnitree G1GR00T N1.6 (shared policy.locomanip.*)
assemble_trocarUnitree G1 + Dex handsGR00T N1.5 (inference-only)
ultrasound_liver_scanFranka-style armopenpi PI0
surgical_reach_psmdVRK PSMGR00T N1.5 or scripted state machine
surgical_reach_dual_psmdVRK dual PSMGR00T N1.5 or scripted state machine
surgical_reach_starSTARGR00T N1.5 or scripted state machine
surgical_lift_blockdVRK PSMGR00T N1.5 or scripted state machine
surgical_lift_needledVRK PSMGR00T N1.5 or scripted state machine
surgical_lift_needle_organsdVRK PSMGR00T N1.5 or scripted state machine

Run

Run the step below before answering. Each step is a separate bash call; variables persist in the local agent's tmux session.

Step 1 — list supported envs

REPO_ROOT="${I4H_WORKFLOWS:-$(git rev-parse --show-toplevel 2>/dev/null)}"; [ -d "$REPO_ROOT/workflows/agentic" ] || REPO_ROOT="$HOME/i4h-workflows"
"${REPO_ROOT}/workflows/agentic/policy/run.sh" --list-envs

If the command fails, use the Supported Envs fallback table above.

Subprojects

DirectoryPurpose
workflows/agentic/arena/IsaacLab-Arena envs, scenes, tasks, teleop, record, replay
workflows/agentic/policy/Policy daemons and train dispatchers
workflows/agentic/dataset/HDF5 → LeRobot conversion and visualization
workflows/agentic/mimic/HDF5 trajectory expansion
workflows/agentic/annotator/VLM success labels and filtering
workflows/agentic/cosmos/Optional Cosmos Transfer video augmentation
workflows/agentic/common/Shared config, messaging, robot constants

Skill Index

SkillPurpose
i4h-workflow-setupInstall / sync / check third-party deps
i4h-workflow-createAdd a new env
i4h-workflow-scene-editEdit an existing scene / task / camera
i4h-workflow-dataset-teleopRecord human demos
i4h-workflow-dataset-replayReplay HDF5 episodes
i4h-workflow-dataset-mimicExpand HDF5 demos with noise
i4h-workflow-dataset-annotateVLM label / filter episodes
i4h-workflow-dataset-convertConvert HDF5 to LeRobot
i4h-workflow-finetuneTrain supported envs
i4h-workflow-validateRoll out / evaluate policy checkpoints
i4h-workflow-e2eRun the full pipeline
i4h-lerobot-vizOpen the LeRobot HTML viewer

Prerequisites

  • For any hands-on stage, set up the workflow first — see [[i4h-workflow-setup]] (component venvs and third-party checkouts present).
  • Env YAMLs at workflows/agentic/config/environments/<env>.yaml are the source of truth.

Limitations

  • Overview/routing only — each pipeline stage has its own skill; this one performs no recording, training, or rollout.

Troubleshooting

  • Error: env not found / unsupported — Cause: typo, unregistered env, or setup incomplete. Fix: run workflows/agentic/policy/run.sh --list-envs; if that fails, run [[i4h-workflow-setup]] first; otherwise check the Supported Envs table.

Final Response

The response is incomplete unless it includes all three parts below, each separated by a blank line. Do not reply with prose only or run sections together without spacing.

  1. Orientation — one short paragraph: env YAMLs, subprojects, sim-to-policy pipeline.

  2. Supported envs — heading plus markdown table from Step 1 (--list-envs output: env, stack, description) or the Supported Envs fallback table.

  3. Available skills — heading plus the full Skill Index table copied into the response.

Then (after another blank line) recommend [[i4h-workflow-setup]] if setup is not done, and name one next stage skill matched to the user's goal.

Más skills de nvidia

compileiq-debug
nvidia
Úsalo cuando algo esté mal: Search() se cuelga, todas las evaluaciones devuelven INVALID_SCORE, las puntuaciones no mejoran, cada configuración devuelve el mismo número, errores de ptxas…
create-github-pr
nvidia
Crear solicitudes de extracción de GitHub usando la CLI gh. Usar cuando el usuario quiera crear un nuevo PR, enviar código para revisión o abrir una solicitud de extracción. Palabras clave de activación -…
nemoclaw-maintainer-cross-issue-sweep
nvidia
Escanea otros issues abiertos para encontrar aquellos que un PR dado también podría corregir o romper accidentalmente. Genera oportunidades de corrección adyacente y riesgos de contradicción con archivo:línea…
fhir-basics
nvidia
Enseña a los agentes cómo funcionan las APIs de FHIR R4, qué recursos están disponibles, cómo consultarlos con parámetros de búsqueda y cómo analizar correctamente todos los formatos de respuesta…
compileiq-validate-result
nvidia
Usar DESPUÉS de que una Búsqueda haya finalizado y ANTES de reclamar cualquier aceleración o enviar un ACF. Carga el CSV de dump_results, extrae los mejores K candidatos (de un solo objetivo)…
changelog-audit
nvidia
Auditar el CHANGELOG.md de Warp antes de un lanzamiento: recuperar entradas perdidas, ordenar por impacto en el usuario, refinar el lenguaje de las entradas, ajustar saltos de línea y (en modo rama de lanzamiento) incrementar comparación…
maintain-dynamic-plugins
nvidia
Mantener los cargadores de plugins dinámicos de NeMo Relay, manifiestos, SDKs nativos de Rust, protocolo de trabajador gRPC, SDK de trabajador Python, documentación, pruebas y cobertura del flujo de trabajo de lanzamiento
dgx-diagnose
nvidia
Diagnostica problemas comunes de la DGX Station GB300: fallos de CUDA, direccionamiento incorrecto de GPU, errores de contenedores vLLM/SGLang, problemas de estado MIG, errores de NVLink/Fabric Manager,…