OUF!

You never 100% sure the AI model you're interacting with will reply accurately, and the data often seem outdated. The app allows people to create their own assistant with fresh, reliable and verifiable context.

These are pretty huge issues a lot of companies and orgs are facing today. Here's how we fix them:

The app allows users to:

• Upload markdown documents to shape their own assistant
• Edit the documents forming the corpus (context files)
• Token-gate access
• Collectively validate the corpus (context files) using an Open Zeppelin Governor contract: users can verify if the IPFS hash of the context files matches the one that was voted on-chain.
• Receive OUF governance tokens each time you make a query (https://sepolia.arbiscan.io/address/0x7ee0e86167746378f1C5bF947EC08c3B55B6Bb80)

OUF can be used in many different cases, including education, research, businesses of all kinds, orgs, etc. AI safety is improved and generated outputs are way more accurate and up-to-date.

I used:

• Next.js app
• Reown (Web3 Modal)
• Ethers.js for on-chain actions
• Interacts with Fatou API (https://github.com/w3hc/fatou)
• The OUF token contract is deployed to Arbitrum Sepolia: https://sepolia.arbiscan.io/address/0x7ee0e86167746378f1C5bF947EC08c3B55B6Bb80

I integrated with SIWE (SignIn With Ethereum) in combination with Reown (email and social login).

I had to tweak Fatou API (Nest.js-based API) to make sure users seamlessly receive some OUF governance tokens when they ask a question to the assistant.