TradeBot — Algorithmic Trading System

Autonomous, multi-agent trading system built on the RBI (Research → Backtest → Implement) pipeline. Agents operate independently, communicate through a shared SQLite blackboard, and are governed by a risk-policy framework that spans four tiers from conservative to aggressive.

Note

Development is temporarily paused on TradeBot (as of 2026-03-13). Active focus has shifted to AugurSight — a companion system for autonomous prediction market trading (Polymarket, Kalshi). AugurSight targets a lower barrier to initial alpha and will stress-test the shared homelab infrastructure before TradeBot goes live. The long-term plan is to integrate AugurSight as a subsidiary prediction market module within TradeBot's portfolio, with TradeBot's Manager Agent acting as the parent orchestrator across both systems. TradeBot resumes when AugurSight reaches stable production.

---

Architecture

Manager Agent
│
├── Research
│   ├── Strategy Agent       — generates and scores trade strategies
│   ├── Market News Agent    — monitors news sentiment and catalysts
│   ├── Predictions Agent    — ML models: Monte Carlo, ARIMA, LightGBM,
│   │                          HMM, Prophet, LSTM, TFT + Ensemble layer
│   └── Crypto Liquidation   — liquidation heatmaps and whale signals
│
├── Backtest Agent           — validates strategies before live deployment
│
├── Brokers
│   ├── Alpaca               — paper trading (stocks + crypto)
│   ├── IBKR                 — live trading (stocks, options, futures)
│   └── OKX                  — live crypto (US endpoint)
│
├── Portfolio Tracker        — live risk management, position sizing, stop-loss
└── Analytics Agent          — performance tracking, model accuracy surveillance

---

Asset Coverage

Class	Assets	Candle Size
Crypto	BTC, ETH, SOL, XRP, ADA, DOGE	6h
Stocks	AAPL, GOOG, NVDA, META, NFLX, ASTS	1d

---

Key Components

Predictions Agent (Active — v1.0.0)

ML forecasting pipeline that generates probabilistic BUY/SELL/HOLD signals for all 12 assets. Signals are regime-gated by an HMM detector that runs first every cycle and determines which models activate.

Model	Role	Regime
HMM	Regime classifier — runs first, gates all others	Always
Monte Carlo	Price distribution / risk bands (P5/P50/P95)	Always
ARIMA/SARIMA	Mean-reversion detection	BEAR/NEUTRAL
LightGBM	Next-candle direction (primary entry signal)	BULL/NEUTRAL
Prophet	Seasonality decomposition, trend confirmation	BULL/NEUTRAL
LSTM	Nonlinear pattern sequences	BULL/NEUTRAL
TFT	Multi-horizon quantile forecast (1/6/14 candles)	All
Ensemble	Confidence-weighted signal synthesis	All

→ See agents/research/predictions/README.md

Risk Policy Framework

Four-tier risk policy (HIGH → LOW) governs leverage, position sizing, and drawdown limits. The Manager Agent enforces the active policy across all brokers and the Portfolio Tracker.

→ See agents/manager/README.md

Blackboard Architecture

All agents share state through tradebot.db (SQLite, WAL mode). No direct agent-to-agent calls — everything is event-driven via the event_log table and ZeroMQ pub/sub bus.

---

Project Layout

TradeBot/
├── agents/
│   ├── common/          — shared ORM models, enums, database session
│   ├── manager/         — orchestrator, risk policy, event routing
│   ├── research/        — strategy, news, predictions, liquidations
│   ├── backtest/        — walk-forward validation engine
│   ├── brokers/         — Alpaca / IBKR / OKX integrations
│   ├── portfolio_tracker/
│   └── analytics/
├── data/
│   ├── datasets/        — historical OHLCV CSVs (500 weeks crypto, 1000 weeks stocks)
│   ├── backtests/       — backtest results
│   ├── state/           — model weights, runtime state
│   └── tradebot.db      — shared blackboard (SQLite)
├── config/              — environment and system configuration
├── docs/
│   ├── DATA_SCHEMAS.md  — all database table schemas
│   └── changelogs/      — versioned release notes
└── scripts/             — data collection, utilities

---

Brokers

Broker	Mode	Assets
Alpaca	Paper trading	Stocks, Crypto
IBKR	Live trading	Stocks, Options, Futures
OKX	Live trading	Crypto (US endpoint)

---

Documentation

Document	Description
agents/manager/README.md	System orchestration, risk policy, full agent map
agents/research/AGENTS.md	Research domain agent responsibilities
agents/research/predictions/README.md	Predictions Agent — models, DB schema, usage
docs/DATA_SCHEMAS.md	All database table schemas (v2.0)
docs/changelogs/	Release changelogs