Workbook

Technical Enablement / Trainer

You own: customer + internal training — how new users get productive on GeneFlow. You translate workbooks into hands-on sessions.

What's new for you

A complete role-by-role workbook library — this directory
A working soak test as a teaching aid: scripts/soak-tests/geneflow-end-to-end.py
A live dashboard demoable at /dashboard/geneflow/ml/*

Course catalog (suggested)

GeneFlow Foundations (2 hours, all roles)

What is GeneFlow? Where it fits.
The lifecycle: train → register → serve → monitor.
Hands-on: train one run, register a version, promote to Staging.
Where to ask questions (workbook for your role, #geneflow Slack).

GeneFlow for ML Engineers (4 hours)

Mirror 01-ml-engineer.md end-to-end.
Lab: deploy your run as an endpoint, save a drift baseline, run a check.

GeneFlow for AI / Prompt Engineers (4 hours)

GeneFlow for MLOps / SRE (3 hours)

GeneFlow for Migration (90 min)

Mirror 16-migration-engineer.md.
Lab: dry-run import from a test MLflow instance.

GeneFlow for Governance (90 min)

Sandbox

For workshops, spin up a sandbox tenant:

genedata tenants create SANDBOX-$DATE --display-name "Training Sandbox $DATE" \
  --auto-delete-after 7d \
  --quota-runs-per-hour 100

Each attendee gets a PAT with geneflow:write scoped to SANDBOX-*.

Demo data

Use the soak test to seed:

GENEFLOW_TENANT_ID=SANDBOX-2026-05-12 \
  python scripts/soak-tests/geneflow-end-to-end.py --keep

This creates:

An experiment with 2 runs (3 metrics each)
A registered model with 2 versions
A prompt with 2 versions
An eval set
An import job (dry-run, doesn't actually pull from anywhere)

Common workshop pitfalls

Cluster IP issues — if attendees are on corp VPN that blocks pod IPs, fall back to kubectl port-forward for the dashboard.
PAT confusion — give them ready-baked tokens; don't make them issue their own in the workshop.
Y.js collab demos — pair them up; the editor really shines with 2+ users.

Where to go next