ERC-7730 AI Builder
Hackathon project improving Ledger’s clear signing builder with AI, schema import, and UX fixes.
Problem Statement
This hackathon project enhances Ledger’s Clear Signing efforts by improving theclear-signing-erc7730-builder. Key additions include AI-generated suggestions for metadata and field names using Sourcify contract data, and a UI tab for protocol-based contract search. While the AI integration works, it’s slow and could benefit from background loading or faster models.The project also began porting thepython-erc7730module to TypeScript for a unified codebase, though this remains unfinished. A notable feature added is schema import for continued editing.Feedback highlights include missing boolean field support, irrelevant display of view/pure functions, unclear array rendering, and the need for schema import from files or registries. The generated ERC-7730 schemas are available in the registry folder. More complex data types, especially raw bytes, remain difficult to handle within the standard.
Solution
This is mostly an improvement of an existing project, so it mostly uses existing technologies. While this is against the rules, it's what made most sense for Ledger and contributing to the ecosystem.That being said, AI was added using Vercel's AI NPM package. All the data for a smart contract is fetched from the Sourcify API and injected in a configurable LLM to extract information.
Hackathon
ETHGlobal Cannes
2025
Contributors
- filoozom
11 contributions