Governed AI execution runtime for accountable action under human authority.
RIO sits between intelligent systems and consequential action. It provides policy review, authorization boundaries, execution gating, receipt generation, and ledger-backed proof.
This repository now contains the active RIO gateway runtime plus SPG-M pattern-governance integration.
The active runtime is the gateway in:
gateway/
Core pipeline:
Signal → Proposal → Policy → Authorization → Execution Gate → Execute → Receipt → Ledger
SPG-M integration:
Pattern signal → SPG-M intake/review → RIO governance context → authorization boundary
- accepts proposed intents,
- evaluates policy and risk,
- requires human authorization when needed,
- blocks unauthorized execution,
- issues bounded execution tokens,
- generates cryptographic receipts,
- writes audit events to a persistent ledger,
- exposes public API v1 endpoints,
- verifies receipt/ledger lineage.
SPG-M — Symbolic Pattern Governance Module — is a non-executing pattern-governance layer.
It can:
- receive ambiguous pattern signals,
- classify consequence class,
- produce gate and routing metadata,
- preview RIO policy review,
- pass optional review metadata into live governance,
- increase review weight when needed.
It cannot:
- approve,
- execute,
- issue tokens,
- generate receipts,
- create memory,
- bypass the Execution Gate.
Implemented SPG-M surfaces:
GET /spgm/status
POST /spgm/intake
POST /spgm/policy-review
POST /govern
POST /api/v1/intents/:id/govern
Read in this order:
SYSTEM_RUNTIME_MAP.mdgateway/README.mddocs/SPG_M_RUNTIME_PLACEMENT.mddocs/SPG_M_GATEWAY_INTAKE_CONTRACT.mddocs/SPG_M_POLICY_CONTEXT_BRIDGE.mdgateway/spgm/VERIFY_INTAKE.mdgateway/spgm/VERIFY_POLICY_REVIEW.mdgateway/spgm/GOVERN_REQUEST_BRIDGE.mdgateway/spgm/API_V1_GOVERN_BRIDGE.mdgateway/spgm/CI_VERIFICATION.md
From gateway/:
npm test
npm run test:spgm
npm run test:spgm:policy-review
npm run test:spgm:govern
npm run test:spgm:openapiSPG-M tests are also covered by GitHub Actions:
.github/workflows/spgm-gateway-tests.yml
| Repository | Role |
|---|---|
rio-protocol |
protocol specification |
rio-receipt-protocol |
receipt proof layer |
rio-system |
active gateway/runtime |
language-intake-mvp |
language governance / crossing detection |
Human-led. Machine-operated. Governed by protocol. Proven by receipts.