AIGovOps Lantern™¶
The human-carried companion to AIGovOps Beacon. Beacon signs. Lantern reads.
Lantern turns Beacon evidence bundles — machine-signed NDJSON streams of policy receipts — into language each role actually wants to see. Same source of truth, four different lenses.
Quick start¶
pip install git+https://github.com/aigovops-foundation/aigovops-lantern.git@v0.1.0-alpha
lantern --version
lantern read bundle.ndjson -f markdown -r auditor
Three commands¶
lantern read— render a Beacon NDJSON bundle as a human summarylantern diff— compare two bundles with a role-targeted narrativelantern explain— resolve a Unified Control ID against the umbrella-govops registry
See the CLI reference for every flag and example.
Four role lenses¶
| Role | Focus |
|---|---|
engineer |
Pipeline integrity, gate failures, CI signal |
compliance |
Control coverage, UCID mapping, framework crosswalks |
auditor |
Chain of custody, signature presence, evidence completeness |
regulator |
Framework citations (NIST AI RMF, EU AI Act, ISO 42001) |
Three output formats¶
text— Rich panel for terminals (default)markdown— drop into a PR comment or emailjson— machine-readable, conforms to a published schema
Design principles¶
- Beacon is the boundary. Lantern reads what Beacon signs; it never claims to verify.
- Role is a first-class input. Same bundle, four narratives.
- Schemas are the contract. Every JSON output validates against a published schema.
- Friendly failure. Bad input never crashes — it raises a
BundleErrorwith an actionable message.
Where to go next¶
- CLI reference — every command and flag
- API · Data model — what a bundle looks like, what a UCID looks like
- API · Actions — the
event_typevocabulary - API · Flows — sequence diagrams for every journey
- API · OpenAPI — the planned v0.2 web view contract
- Python API — for tool authors embedding the renderer
- Roadmap — what's coming in v0.2, v0.3, v1.0