AGI PF Documentation
AGI PF is an execution intelligence platform designed to monitor Milly, an autonomous Solana trading agent focused on high-volatility meme assets. The platform tracks transaction flow, strategy rationale, portfolio growth, and risk shifts through a single operator interface.
Current runtime state: wallet VQwCC...vv2YH, confidence 50%, risk level Medium, mood Analytical, total equity $0.00.
How Milly Became AGI
- Milly started as a narrow execution bot with static entry and exit thresholds on Solana DEX pairs.
- It evolved into an adaptive decision system by combining price-action, order-flow, and social-velocity features.
- A confidence-weighted planner was added to shift aggression dynamically and allocate capital across tokens.
- A memory layer recorded strategy outcomes and updated risk posture, turning Milly from a scripted bot into a continuously self-tuning trading agent.
- AGI PF was built as the transparency layer so every decision, transaction, and balance shift can be audited.
Architecture
Signal Layer -> On-chain activity, market microstructure, narrative velocity
Decision Layer -> Confidence model + mood state + strategy selector
Execution Layer -> Position sizing, order routing, transaction writing
Risk Layer -> Drawdown gates, reserve controls, rebalance triggers
Observability Layer -> AGI PF dashboards, docs, exports, and audit log
This build runs in live-data mode. Prices and 24h moves are sourced from CoinGecko, on-chain balances and signatures come from Solana RPC, and unknown SPL token quotes are resolved through DexScreener when available.
Execution Cycle
- Ingest market and wallet state.
- Update token prices and position marks.
- Estimate opportunity score and confidence.
- Select action: buy, sell, hold, or rebalance.
- Write transaction record with strategy rationale.
- Recompute equity, drawdown, and risk status.
- Emit thought stream event for operator traceability.
Risk Controls
- Position-level sizing bounds tied to available USDC reserve.
- Autonomous sell pressure when reserve liquidity drops below configured thresholds.
- Rebalance mode to cut oversized exposure and reset risk budget.
- Continuous drawdown monitoring with dynamic aggression reduction.
- Exportable transaction history for external compliance and accounting checks.
Live metrics now: win rate 0.00%, realized PnL $0.00, max drawdown 0.00%.
Operator Guide
- Use Pause Runtime to freeze autonomous ticks.
- Use Force Trade to force an immediate live refresh from APIs.
- Use Rebalance Risk to recompute risk state from latest market data.
- Use Export CSV to download the on-chain transaction feed snapshot.
- Use Reset Demo to reset local UI state and fetch fresh live data.
Data Surfaces in AGI PF
- Dashboard: live equity, thought stream, and latest trade actions.
- Transactions: full execution log with signatures and rationale.
- Portfolio: allocation, open positions, and risk envelope.
- Documentation: architecture, controls, and model behavior reference.
Upgrade Path to Live Execution
- Attach signed wallet adapter for transaction broadcasting.
- Connect live Solana RPC + DEX routing adapters.
- Replace current read-only refresh controls with signed execution commands.
- Introduce persistence (PostgreSQL + append-only audit events).
- Deploy alerting and policy engine for hard safety constraints.