Skip to content

Use Case Cookbook

These recipes are copy-pasteable paths for common WorldForge work. Each one names the command, the expected output, the artifact to keep, the first triage step, and the non-claim boundary. Use the matching demo workflow when you want a preserved showcase run for issue or release evidence.

Recipe 1: First Local World

Field Value
Command uv run python scripts/demo_showcases.py run first-run --workspace-dir .worldforge/demo-showcases --overwrite
Expected output status: passed, one mock world, one object, one prediction, and preflight status
Artifact .worldforge/demo-showcases/first-run/exported-world.json
First triage step run uv run worldforge world preflight --state-dir .worldforge/demo-showcases/first-run/worlds
Boundary mock provider only; no physical-fidelity or real-runtime claim

Recipe 2: Provider Diagnostic Issue Bundle

Field Value
Command uv run python scripts/demo_showcases.py run diagnostics-issue-bundle --workspace-dir .worldforge/demo-showcases --overwrite
Expected output skipped provider diagnostic with safe_to_attach: true
Artifact .worldforge/demo-showcases/diagnostics-issue-bundle/issue-bundle/issue.md
First triage step open evidence_manifest.json and confirm safe_to_attach: true
Boundary fixture diagnostic only; do not paste raw provider credentials or signed URLs

Recipe 3: Robotics Replay Before Prepared-Host Work

Field Value
Command uv run python scripts/demo_showcases.py run robotics-replay --workspace-dir .worldforge/demo-showcases --overwrite
Expected output selected candidate index, candidate costs, policy result, score result, and event phases
Artifact .worldforge/demo-showcases/robotics-replay/robotics-replay-manifest.json
First triage step run uv run worldforge-demo-lerobot before scripts/robotics-showcase --health-only
Boundary deterministic replay only; robot hardware, controllers, safety checks, and checkpoints stay host-owned

Recipe 4: Remote Media Dry Run

Field Value
Command uv run python scripts/demo_showcases.py run remote-media-dry-run --workspace-dir .worldforge/demo-showcases --overwrite
Expected output redacted Cosmos success and Runway expired-artifact events
Artifact .worldforge/demo-showcases/remote-media-dry-run/remote-media-events.json
First triage step inspect provider event target, message, and metadata for redaction before live smoke
Boundary fixture-backed dry run only; no paid API calls or artifact-retention guarantee

Recipe 5: Adapter Author Scaffold

Field Value
Command uv run python scripts/demo_showcases.py run adapter-author --workspace-dir .worldforge/demo-showcases --overwrite
Expected output generated provider, generated test, docs stub, runtime manifest stub, and workbench report
Artifact .worldforge/demo-showcases/adapter-author/generated-provider/
First triage step replace placeholder fixtures and run the generated provider test before promotion
Boundary scaffold is intentionally fail-closed and incomplete; it is not evidence of real provider behavior

Recipe 6: Batch Eval With Budget Failure

Field Value
Command uv run python scripts/demo_showcases.py run batch-eval --workspace-dir .worldforge/demo-showcases --overwrite
Expected output eval job passes, benchmark job returns controlled exit_code: 1, and both runs are preserved
Artifact .worldforge/demo-showcases/batch-eval/batch-host/runs/<run-id>/run_manifest.json
First triage step inspect the benchmark report and copied budget before changing thresholds
Boundary mock provider and impossible budget only; no scheduler, durable storage, or release budget claim

Recipe 7: Stdlib Service Host Smoke

Field Value
Command uv run python scripts/demo_showcases.py run service-host --workspace-dir .worldforge/demo-showcases --overwrite
Expected output readiness is ready, one mock prediction request is summarized, and server shutdown is recorded
Artifact .worldforge/demo-showcases/service-host/runs/<run-id>/results/summary.json
First triage step run uv run python examples/hosts/service/app.py --provider mock --port 8080 and inspect /readyz
Boundary reference host only; auth, deployment, uptime, dashboards, and rollback remain host-owned
Field Value
Command uv run python scripts/demo_showcases.py run rerun-gallery --workspace-dir .worldforge/demo-showcases --overwrite
Expected output status: skipped with missing rerun extra and a gallery layer manifest
Artifact .worldforge/demo-showcases/rerun-gallery/rerun-gallery-manifest.json
First triage step install worldforge-ai[rerun], then run uv run --extra rerun worldforge-demo-rerun
Boundary checkout-safe manifest only; visual .rrd files require the optional Rerun runtime

Recipe 9: Failure Recovery Lab

Field Value
Command uv run python scripts/demo_showcases.py run failure-lab --workspace-dir .worldforge/demo-showcases --overwrite
Expected output missing credential, corrupted state, unsafe metadata drills, preflight, and recovery commands
Artifact .worldforge/demo-showcases/failure-lab/failure-lab-report.json
First triage step read recovery_commands before touching real .worldforge state
Boundary mutates only the lab workspace; no real credentials, optional runtimes, or user state are used

Recipe 10: Full Showcase Evidence Sweep

Field Value
Command uv run python scripts/demo_showcases.py run all --workspace-dir .worldforge/demo-showcases --format json --overwrite
Expected output all ten workflows report passed or intentional skipped, and top-level status is passed
Artifact .worldforge/demo-showcases/<workflow>/runs/<run-id>/run_manifest.json
First triage step open the failed workflow's workflow-result.json, then its preserved run_manifest.json
Boundary integration evidence only; optional runtimes, provider credentials, robots, and physical-fidelity claims stay outside the checkout path