Skip to main content

​
Evidence Packet Pilot Runbook

Use this flow for beta design-partner pilots. It packages an existing Physical AI or VLA evaluation output into a signed evidence packet and, when explicitly approved, a hosted reviewer share.
This is not a self-serve production SaaS claim. Hosted shares should be used in approved pilot/staging environments only when the target commit and environment have fresh external proof for the hosted-share path.

​
Scope

WorldFlux does:
  • triage an already-extracted run folder or adapter audit input locally;
  • create or review claim.json and protocol.json;
  • generate, sign, and verify an evidence package;
  • render a reviewer-safe memo;
  • publish an approved hosted share when the customer explicitly approves upload.
WorldFlux does not:
  • store model weights or private keys;
  • take custody of raw videos, checkpoints, datasets, or credential-bearing logs;
  • act as the official benchmark owner;
  • certify deployment safety or legal/regulatory conformity;
  • production-claim live provider/runtime execution unless explicitly marked.

​
Commands

worldflux audit import run-folder \
  --from <extracted_run_dir> \
  --dry-run \
  --report import_report.json \
  --customer-report import_report_public.md

worldflux audit import run-folder \
  --from <extracted_run_dir> \
  --select-run <candidate_id> \
  --output audit_input.json \
  --report import_report.json \
  --customer-report import_report_public.md

worldflux claim create --template openpi-libero --output claim_pkg/

worldflux audit run audit-input \
  --from audit_input.json \
  --claim claim_pkg/claim.json \
  --protocol claim_pkg/protocol.json \
  --output evidence_pkg/

worldflux audit sign evidence_pkg/
worldflux audit verify evidence_pkg/
worldflux evidence inspect evidence_pkg/
Keep import_report.json as private operator output. Use import_report_public.md when a customer needs the triage status. Render the reviewer memo. This deployment memo is the canonical reviewer artifact for Design-Partner Ready Beta: 
worldflux dd render \
  --evidence evidence_pkg/ \
  --investor "Reviewer" \
  --portfolio-company "Customer" \
  --decision-context "pilot evidence review" \
  --output outputs/dd \
  --tone-profile deployment_evidence
Publish only after explicit approval: 
worldflux audit publish evidence_pkg/ \
  --share \
  --cloud-run-id <cloud-run-uuid> \
  --confirm-public-share-upload \
  --approval-file public_share_approval.json \
  --password-env WORLDFLUX_SHARE_ACCESS_CODE

​
Before Sending

Confirm the reviewer URL, expiry, approval fields, recommendation, missing evidence labels, verification status, and revocation owner. Attach known limitations that state what the package proves and does not prove. Passport is deferred for this beta flow. If a customer asks for a passport, provide the deployment memo; any future Passport must be a thin alias or wrapper over the memo after a separate implementation plan.

​
Support Docs

Use the repository support docs before the first reviewer send:
  • docs/demos/design-partner-security-overview.md
  • docs/runbooks/design-partner-incident-runbook.md
  • docs/runbooks/design-partner-operator-runbook.md
  • docs/demos/design-partner-onboarding-questionnaire.md
  • docs/demos/design-partner-reviewer-memo-template.md