Unreal AI

OpenAI-compatible, on-chain metered AI compute network. Unreal unifies access to AI models, provides transparent on-chain receipts for every call, and lets independent datacenters monetize capacity under common meters and SLAs.

• Products: OpenAI-compatible API and Console for enterprises, an Art platform for creators, and a ranked provider marketplace.
• Billing: Deterministic metering with UNREAL-denominated rates across input/output tokens and media units (images, video, audio).
• Mission: Make AI feel like a vending machine—choose a model, pay for exactly what meters record, and receive auditable receipts.

Refer to the founding backstory in aidocs/info-about-founding/backstory.md for deeper context.

---

Why Unreal

• Unified access: One OpenAI-compatible interface across multiple models and providers.
• Verifiable billing: Each inference produces an on-chain receipt containing cryptographic usage meters.
• Predictable pricing: Governance-set bands keep rates sustainable for providers and transparent for buyers.
• Enterprise ready: Keys, wallets, project-centric workflows, audit-friendly receipts.

Product Surface

• OpenAI-compatible API: One-line migration from OpenAI SDKs with standardized metering and settlement in UNREAL.
• Console: Wallet-connected key management, developer airdrops, and business registration for managed access.
• Art platform: High-speed surface that proved UX and supply; now a funnel and live showcase for network latency and cost.
• Marketplace: Ranked provider directory with regional routing and lanes (Priority vs Flex) tied to SLA performance.

Pricing & Metering (T1)

• Token meters: UNREAL per 1M input tokens (R_in) and per 1M output tokens (R_out).
• Media meters: Clearly defined units for images, video, and audio. Rates validated against provider costs.
• Settlement: Prepay in UNREAL or top up via fiat/stablecoin bridges; every call writes a verifiable on-chain receipt.

Architecture (High Level)

• Router logs deterministic meters per call and writes receipts (e.g., Merkle-hashed usage) on-chain.
• Provider program validates costs and SLAs; marketplace placement rewards reliability over raw price.
• Governance oversees pricing bands, dispute rules, and SLA definitions.

Quickstart (This Repository)

This repo contains the Unreal website (React + TypeScript + Vite).

Prerequisites

• bun 1.2.5+ (or newer)

Install and run

bun install
bun run dev

Build and preview

bun run build
bun run preview

Project layout

• src/ — React app source
• public/ — static assets
• aidocs/ — internal docs, including the founding backstory

Using the OpenAI-Compatible API

Unreal speaks the OpenAI API so you can migrate with a baseURL change. Example (Node):

import OpenAI from "openai"

const client = new OpenAI({
  baseURL: process.env.UNREAL_API_BASE, // e.g. https://api.unreal.example/v1
  apiKey: process.env.UNREAL_API_KEY,
})

const response = await client.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "Hello, Unreal!" },
  ],
})

console.log(response.choices[0]?.message?.content)

Or via cURL:

curl -X POST "$UNREAL_API_BASE/chat/completions" \
  -H "Authorization: Bearer $UNREAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [
      {"role": "user", "content": "Hello, Unreal!"}
    ]
  }'

Note: Actual models, lanes, and rate cards are subject to governance bands and provider availability.

Roadmap (At a Glance)

• Launch T1 metered billing with cryptographic receipts and high-RPM enterprise lanes.
• Onboard 20–50 providers with validated rate cards; open the ranked marketplace with regional routing.
• Expand multimodal coverage and activate governance for pricing bands and dispute workflows.

Contributing

We welcome issues and PRs. Please read CONTRIBUTING.md for branching, commits, linting, and PR guidelines.

Docs

• Founding backstory: aidocs/info-about-founding/backstory.md
• Additional internal docs live under aidocs/.