Inspiration
Could a flood-response fund publish its trigger and recipients before the water rises, then release aid against a verifiable river reading? OMI explores that question with the Niger at Lokoja, Nigeria, and lets people inspect other rivers too.
What it does
OMI connects public river-flow data to a donor-funded, rule-based payment contract. A fund manager creates a site, registers recipient wallets, and funds the contract. Submitted readings include a hash of the source response and enter a one-hour challenge period. Afterward, anyone can settle an eligible reading; a daily automation can also settle it.
For Lokoja, the contract supports a 30% early payment if the reported forecast median reaches the flood threshold within 15 days, then releases the remaining cover after two consecutive reported days at or above that threshold. Payout decisions are deterministic contract rules; an AI model does not authorize payments.
How we built it
The web app uses Next.js, React, TypeScript, and ethers.js. Solidity contracts run on Arc testnet with test USDC. River inputs use GloFAS and Open-Meteo data. The prototype includes a historical replay, a fund page, registered test wallets, and a settlement path.
Challenges we ran into
Archived forecasts were unavailable for the September 2022 replay. That means we could demonstrate the observed-flow trigger and payment path, but not the forecast-based early-payment path. We state that limitation clearly instead of implying that the forecast scenario was tested.
Accomplishments we're proud of
The deployed prototype replays recorded September 2022 river-flow data and shows one test USDC paid to each of two test wallets. The replay reaches the full-payment condition on September 23, six days before the modeled flow peak on September 29. These are test tokens and test wallets; no aid was sent to real flood victims.
What we learned
Modeled flow is not a substitute for local gauge observations. A real pilot would need independent data reporting, a security review, adequate funding, recipient consent, and a community partner. The current reporter and guardian remain trusted roles.
What's next for OMI
Validate the trigger with local gauge data and community partners, audit the contract, and test a forecast-backed early-payment replay before attempting a real pilot.
Built with
Solidity, TypeScript, Next.js, React, ethers.js, Arc testnet, test USDC, GloFAS, Open-Meteo, Vercel, and GitHub Actions. Development used AI assistance; payment decisions remain deterministic contract rules.
Log in or sign up for Devpost to join the conversation.