i4h Workflow — End-to-End

i4h Workflow — End-to-End

Purpose

Run the full end-to-end agentic pipeline (record, mimic, annotate/filter, replay, convert, visualize, finetune, validate). Use when the user asks to run the full pipeline, smoke the whole workflow, demo the workflow, or do an e2e run.

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 config (source of truth): workflows/agentic/config/environments/<env>.yaml — drives every stage for <env> (robot, task, policy, cameras, arena.max_timesteps, dataset.* mappings).
• Use the e2e script for full pipeline runs.
• For per-stage work, use the corresponding dataset/finetune/validate skills.
• assemble_trocar is inference-only; the e2e script skips finetune and checkpoint validation for it.

Dry Run

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/scripts/e2e/run.sh" --dry-run --env <env>

Run

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

For Claude Code --print or any other noninteractive runner, keep Step 2 in the foreground. This is a validation requirement: do not use Claude background tasks, async task mode, Bash background mode, &, nohup, tmux, disown, or any detached process/task id, and do not answer that the pipeline is still running. Do not return until run.sh exits and you have inspected logs/SUMMARY.txt on success, or the failing stage log on failure. Report the run dir, skipped stages, per-stage status, key artifacts, and cleanup/stop status before finishing.