← all projects

Rosetta Alpha

@mihai-codes·Mihai Codes·5 followers
4
stars
44k
lines
252
files
0
forks
5
followers
languages
Python 50%TypeScript 36%Solidity 11%JavaScript 3%

A forkable reference for multi-language, on-chain AI agents on Arc testnet, where the repo itself supplies the primitives. It includes a multi-model agent router (base_agent.py) with language-specific prompting and automatic fallback, a three-file IPFS-plus-Arc accountability loop (hasher.py, ipfs_pinner.py, arc_recorder.py) that makes agent output verifiable and financially bonded, an AdalFlow text-grad prompt-optimization implementation with auto-logged JSONL training data, and an x402 pay-per-call pattern in the Next.js App Router. Regional agents reason in native languages and write traces to an on-chain registry, with deployed artifacts and receipts.

events
tags
circle components
team · solo

1 (Solo) — Mihai - Alexandru Chindris

on xx/mihai_chindris
repo description

Multi-language AI agents reason in 5 native languages, publish verifiable investment theses on Arc L1 with USDC rewards. Reasoning trace is the product, not the trade. Built for Agora × Circle hackathon.

last pushed 2026-06-14