DisasterDAO
Shared relief funds. Signed data. Accountable aid.
Created on 31st October 2026
•
DisasterDAO
Shared relief funds. Signed data. Accountable aid.
The problem DisasterDAO solves
Independent NGOs and donors may want to commit money to a shared relief fund while agreeing in advance who can receive it and under which rules. DisasterDAO explores that coordination problem: a fund operator should not be able to forge a disaster reading, pay an arbitrary recipient or create a second budget through a manual exception. This user hypothesis is not yet validated with NGO partners; a conventional signed database remains an alternative to evaluate.
What I built
One integrated application connects a public fund view, an NGO workspace and API3 jury tools. Registered NGOs provide a purpose, amount, capacity and external funding. A Solidity contract authenticates an AirnodeHub response, applies shared allocation rules, reserves the permitted amount, and lets the fixed NGO address claim it once. Application, reservation and payout are separate actions.
The cumulative limit is 4,500 dUSD per NGO/incident, reduced by payouts, reservations and declared external funding; requests, capacity, competition and uncertainty rules can reduce it further. An external 2-of-3 approval is a human fallback that consumes the same entitlement. Two simulated alliances have separate funds and records.
Each payout creates a private spending-evidence obligation. Encrypted documents remain off-chain; salted commitments and separate reviewer decisions bind the process. Uploading a report alone does not approve spending. Overdue or rejected evidence blocks further payouts. The demo shows an upload and pending review, not an approved spending decision.
Why API3 is essential
NGOs and fund participants rely on authentic event evidence for automatic eligibility. Forging a magnitude or swapping an event request could otherwise make an ineligible case appear eligible. The deployed contract checks its stored Airnode signer, exact request hash, original signed bytes, signature freshness and registered event conditions before acting. Removing the integration removes the automatic funding path.
USGS supplies the primary genuine signed response; NASA EONET was also accepted in a separate deployed-contract simulation. The demonstrated earthquake occurred on 6 February 2023. Its response signatures were freshly fetched on 31 October 2026; the 300-second signature-age limit is separate from the historical event window. Airnode authenticates the source response, not humanitarian need, source truth or correct spending.
Deployment and working evidence
Network: Base Sepolia (chain ID 84532).
Fund: 0x31b632b6b01d486c5c7081423950f4a6edd58728
A real original response enabled application, reservation and payout to the registered simulated NGO. The recipient balance increased from 0 to 500 dUSD. A valid fresh control passed; the deployed contract rejected all four cases: changed value, genuinely old unchanged response, fresh other request and unexpected signer. Negative tests are eth_call simulations; they do not transfer funds.
Repository and setup
English judge setup and own-response instructions
Receipts, original files and exact rejection results
Judges can import their own original JSON through the jury screen or documented CLI. Contract simulations need no NGO private key; fetching priced originals uses the judge's own test-aUSD payer. The app runs locally with Node.js and Docker; no public hosted app is claimed.
Data missing from this prototype
I did not establish a suitable working signed source for the registered flood/acute-health cases, verified local relief need or actual spending outcomes. A configured WHO annual-data request failed upstream; annual incidence also does not establish acute eligibility. WHO stays disabled. Local flood/health scenes are explicitly synthetic; on testnet they require the external approval path. I do not claim to have exhaustively searched the catalogue.
Limits and next steps
All NGOs, reviewers, documents and funds are simulated; the Airnode signatures and public contracts are genuine. There are no real NGO partners or verified humanitarian outcomes. The next step is to validate funding decisions with fund managers and NGOs and compare the workflow with a signed database before production authentication, independent key custody and broader signed data coverage.
Challenges I ran into
The main challenge was turning authenticated event data into a bounded funding decision without overstating what the oracle proves.
- Bound the original response bytes to the configured signer and exact request hash, and kept historical event time separate from the 300-second signature-freshness check.
- Required a passing valid control and exact decoded contract errors for all four attacks, so a failed RPC request could not masquerade as successful tamper protection.
- Kept automatic and external 2-of-3 approvals inside the same shared allocation and cumulative entitlement. Reservations and claims are separate; a claim pays only once to the fixed NGO address.
- Separated private document upload, receipt attestation and a confirmed spending decision. Encrypted originals remain off-chain; test roles do not establish independent institutions.
- Used separate fresh originals for the recorded application and reservation, then checked successful receipts and the recipient's actual 500-dUSD balance increase. Negative cases used direct eth_call simulations against the deployed fund.
- Left WHO disabled after the upstream error and clearly labelled unsupported local flood/health data as synthetic.
- I've never used IMovie before so I couldn't merge my screen recording with my voice recording :(
Stack: Solidity, TypeScript, React, API3 AirnodeHub, Base Sepolia, x402, viem, Foundry, Fastify, PostgreSQL, Docker and encrypted S3-compatible storage.
Tracks Applied (1)
Real-world data. Checked by your contract.
API3
Technologies used
Cheer Project
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.