@openrng/core
VEO-2 types, schema, and shared primitives for the OpenRNG ecosystem.
VEO = Verifiable Execution Object — a standard format for recording, proving, and explaining AI decisions.
Install
npm install @openrng/core
Quick Start
import { createVEO, validateVEO, createVEOHash } from '@openrng/core';
const veo = createVEO({
provider_id: 'my-service',
execution: {
prompt_hash: 'sha256-of-prompt',
output_hash: 'sha256-of-output',
model_id: 'gpt-4o',
latency_ms: 412,
cost: { total_tokens: 1500, cost_usd: 0.003 },
},
confidence: { score: 850, grade: 'AA' },
});
const { valid, errors } = validateVEO(veo);
const hash = createVEOHash(veo);
Object Classes
| VEO-2A | Raw Execution | Single AI call (chat, completion, inference) |
| VEO-2B | Composite Execution | Multi-step chains, agent pipelines |
| VEO-2C | Anchored Execution | With blockchain proof / Merkle anchor |
| VEO-2D | Governed Execution | With policy assertions, human approvals |
VEO Lifecycle
created → signed → anchored → indexed → verified
Full Spec
Related Packages
@openrng/core | Types, schema, validation (this package) |
@openrng/verify | Verify any VEO object |
@openrng/auto | Auto-instrument AI SDK calls |
License
MIT — OpenRNG