Skip to main content
Every component described here is implemented in this repository — not mocked for demos.
Full final architecture diagram, data flows, packages, deployment, and proof model: see the project README → Architecture section. The diagram and breakdown below are a concise system view.

​
System context (overview)

Live contract: 0xB86dC64573089D8DD89C5686010295bB4412D652 · signer: 0x59ffc8907beaA275F29B466BCB1D9BbfeaDAd165

​
Components

​
Cognition loop

The LangGraph state machine in apps/worker/ameo_worker/graph.py runs: observe → delta_neutral (lp/perps when vol) → reason → plan → guardrail → act → self_heal → log (finalize runs async for 0G + onchain). Each transition emits structured events to logs/events/events_*.jsonl. The replay API (GET /api/cycles/{id}) + UI reconstructs the full verifiable cognition rail (observation through on-chain + 0G proof). See the project README#architecture for the detailed final diagram.

​
Data model (minimal)

​
Security defaults

  • LIVE_ENABLED=false and WORKER_MODE=dry_run by default
  • Max daily notional and per-trade caps before execution
  • Private keys never passed to LLM prompts
  • Slither recommended on contracts before mainnet

​
Deployment

See Deploy guide for production hosting.