Audit-Proof Identity

Inventory variance, triage, and 0G proof under one mark.

AuditEvent generator+OpenClaw triage+0G proof

ARKA

Physical operations, turned into reviewable proof.

ARKA compares business intent against physical inventory movement, creates an AuditEvent, routes the case through an OpenClaw-compatible triage layer, and seals important proof packages through 0G Storage and 0G Chain.

Open Operator Console Check Truthful Status

AuditEvent

CASE-STATE-C

REQUEST_EXPLANATION

Expected90g3 Protein Shakes

Recorded99gWhey movement out

Proof pathREAL

1AuditEvent created

2Proof package canonicalized

30G Storage root returned

4AuditProofRegistry anchor written

storage root: 0x65b50bcd81377e...

chain tx: 0x4c378bb1801a7057...

Operator note

Usage is above expected range. Ask for explanation after owner approval. Do not rewrite the AuditEvent facts.

Architecture

One loop, six strict responsibilities.

ARKA stays narrow: it creates, triages, investigates, seals, and anchors AuditEvents. It does not become a POS, ERP, warehouse suite, CCTV AI system, or HR punishment workflow.

Backend

Creates AuditEvent

Calculates expected usage, actual movement, variance, status, severity, and recommended action.

Database

Stores evidence

Keeps operational evidence and ProofRecord metadata fast and local to the app workflow.

OpenClaw

Layer-1 triage

Operates on AuditEvent facts and appends owner-safe recommendations without rewriting reconciliation truth.

Dashboard

Layer-2 investigation

Shows scenario runs, expected-vs-actual comparison, triage state, and proof lifecycle in one console.

0G Storage

Sealed packages

Stores selected redacted proof packages for important cases, not daily operational queries.

0G Chain

Proof anchors

Registers compact anchors that point to 0G Storage packages without putting private case data on-chain.

Demo world

Protein bar scenario, built for repeatable judging.

The MVP uses one concrete world: Protein Shake orders, Whey Protein movement, and a 30g usage rule. The point is not to simulate a full business. The point is to show how a mismatch becomes an auditable case.

Product: Protein Shake
Usage rule: 30g Whey / serving
Handler: Joni
Cashier: Nina

State A

Clear usage

AUTO_CLEAR

Business intent90g expected

Physical movement90g recorded

Shows ARKA does not interrupt the owner when business intent and physical movement match.

State C

Request explanation

REQUEST_EXPLANATION

Business intent90g expected

Physical movement99g recorded

Creates a moderate review case and prepares an owner-approved clarification request.

State D

Critical review

ESCALATE

Business intent90g expected

Physical movement160g recorded

Raises an immediate review case when movement is materially above the expected range.

Truthful status

The landing page does not overclaim the demo.

ARKA shows what is real, what is partial, and what remains outside the current MVP. Chain anchors do not prove blame; they prove that a sealed package was registered for review.

Deployed 0G Galileo registry

0xEA4a472F0123fC9889650be807A1FF5EF780029F

AuditEvent generation

PARTIAL

A/C/D demo cases create order-linked AuditEvents through the local Next.js flow.

OpenClaw triage

PARTIAL

Deterministic fallback is active; local fork, skill loading, plugin load, and one model inference smoke are verified.

0G Storage

REAL

A live proof package upload has been verified through the dashboard proof service.

0G Chain

REAL

AuditProofRegistry anchoring is verified on 0G Galileo Testnet.