ARKA

ARKA is an audit-first inventory intelligence system that helps small businesses prove what happened between a business transaction and physical inventory movement. Instead of trying to become a full ERP, warehouse system, or POS replacement, ARKA focuses on one core problem: turning orders, usage rules, and actual stock movements into a clear AuditEvent. That AuditEvent explains whether the physical usage of an item matches what the business expected to happen.

For example, if a POS order says one product was sold, ARKA checks the expected inventory usage against the actual movement detected from stock, scanner, sensor, or simulated input. It can support consumable inventory like whey, chicken, flour, or oil, and countable inventory like lamps, spare parts, medicine, or filters by translating units into expected physical quantities such as weight. ARKA then classifies the result with variance, severity, proof status, and recommended next steps.

The goal is to give business owners a trustworthy case file for each audit loop: what was ordered, what should have moved, what actually moved, whether the difference was acceptable, and whether it needs action. Downstream systems like the dashboard, OpenClaw agent, Telegram alerts, and 0G proof layer all consume the AuditEvent instead of raw data.

ARKA is built as a TypeScript monorepo with separate workspaces for the web app, shared types, backend logic, agent layer, database layer, smart contracts, tests, and documentation. The frontend uses Next.js, React, and TypeScript, while the root workspace scripts handle building, linting, typechecking, testing, and OpenClaw verification across the project.

The system is centered around an AuditEvent pipeline. ARKA takes order data, product or menu data, usage rules, recipes, and inventory movement records, then compares the expected quantity against the actual quantity moved. Usage rules define how a sold product maps to physical stock, including units, tolerances, and weight conversions. This allows ARKA to support both service-window workflows, such as food, beverage, and repair businesses, and direct-out workflows, such as retail, pharmacies, lamps, filters, and spare parts.

OpenClaw is used as the agent layer for triage and owner-facing reasoning. It does not change reconciliation facts. Instead, it reads backend-generated AuditEvents and decides whether a case should be auto-cleared, silently logged, sent for explanation, or escalated. This keeps the system safer because the backend owns the audit truth, while the agent helps business owners understand and act on the result.

ARKA also includes a 0G proof architecture. The local database acts as fast operational memory, 0G Storage stores audit packages, and 0G Chain serves as a proof registry rather than the main database. Smart-contract tooling is prepared with Hardhat, ethers, and TypeScript. For the hackathon demo, ARKA can run simulated inventory movements and deterministic agent fallbacks, while still being structured for future integrations with real scanners, sensors, Telegram alerts, and verifiable proof storage.