Demo: Runtime Governance Trace to Evidence

This demo walks through the core GA operator story:

runtime trace -> policy decision -> explanation / grounding / scope / RBAC / lineage -> operator package

Script

Run:

python examples/runtime_governance_demo.py

What it does:

1. Loads and activates a production runtime policy bundle.
2. Registers scope and RBAC manifests.
3. Creates a trace with one low-grounding answer path.
4. Emits signed policy, explanation, grounding, scope, RBAC, and lineage records.
5. Inspects the trace through sf_operator.
6. Exports a signed operator evidence package to examples/artifacts/runtime_governance_operator_package.json.

Expected Outcome

The example is intentionally tuned so the grounding path triggers human review:

• policy action: human_review
• scope decision: allowed
• RBAC decision: allowed
• grounding result: below threshold
• operator outcome: escalated for review

Relevant APIs

• sf_policy.load_bundle()
• sf_policy.activate()
• sf_scope.evaluate_with_policy()
• sf_rbac.authorize_with_policy()
• sf_rag.assess_grounding_with_policy()
• sf_explain.generate_with_policy()
• sf_lineage.record_with_policy()
• sf_operator.inspect_trace()
• sf_operator.export_package()

CLI Follow-Up

After running the script, inspect the same trace from the CLI:

spanforge operator inspect trace-phase7-runtime --format json
spanforge operator export trace-phase7-runtime --output operator-package.json