agora scout

@megadeth17

Three Arc primitives an AI portfolio manager can fork from chain.py: commit-then-act anchoring that embeds decision hashes in Arc calldata before executing trades for an on-chain audit trail, a regime-detection agent loop (a reusable FastAPI plus async SQLite pattern with rate limiting, WebSocket live feed, and visitor tracking), and USYC rotation logic for switching between idle USDC and the USYC tokenized money market on AI regime signals. It uses Kimi K2 / Moonshot via an OpenAI-compatible interface; decisions are anchored on Arc while the rebalancing itself is simulated.