Realia

Realia is a decentralized protocol that verifies whether an image is real, AI-generated, or modified. In an era where generative AI can produce near-indistinguishable fake visuals, Realia provides verifiable proof of authenticity through a combination of blockchain, autonomous AI agents, and crypto-economic incentives.

Unlike centralized detection tools, Realia uses a multi-agent verification model—each agent independently evaluates image embeddings, stakes PYUSD, and submits on-chain verification results. Consensus among agents determines authenticity, ensuring no single point of failure or trust. Dishonest agents are slashed; honest ones are rewarded.

How It Works:

Minting Flow – Creators pay 1 PYUSD to mint authenticity NFTs for their original images. The backend verifies image originality using AI (aiornot API) and ensures no duplicates exist. Each minted NFT’s metadata is stored on IPFS, and embeddings are indexed by autonomous agents using CLIP + Qdrant. Verification Flow – Anyone can pay 0.05 PYUSD to verify an image. Agents fetch the image, generate embeddings, and compare them with stored NFT fingerprints. Based on similarity (≥0.95 = VERIFIED, 0.75–0.95 = MODIFIED, <0.75 = NOT_VERIFIED), results are posted on-chain. Consensus finalizes the verdict, and rewards are distributed automatically.

Core Innovations:

Economic Game Theory: Agents must stake PYUSD to participate. Honest consensus earns PYUSD; dishonest outliers are slashed. Honesty becomes the most profitable strategy. Autonomous Verification: Five Fetch.ai agents operate independently via Agentverse, maintaining their own Qdrant databases and responding to blockchain events without central coordination. Transparency: All verification events—minting, responses, slashing, and rewards—are indexed and auditable through Blockscout, ensuring full public transparency.

Smart Contracts: Solidity contracts on Arbitrum Sepolia handle minting, staking, and verification.

Frontend: Next.js + TypeScript + Wagmi + RainbowKit + Tailwind for seamless wallet UX.

Backend: Node.js (Express) with Prisma, PostgreSQL, AWS EC2.

AI Layer: Python Flask microservice running CLIP embeddings via sentence-transformers.

Autonomous Agents: 5+ Fetch.ai Agentverse agents performing decentralized verification.

Vector Databases: Independent Qdrant Cloud instances per agent.

Storage: IPFS (Pinata) + AWS S3 for redundancy.

Explorer: Full event indexing on Blockscout.