QuantAgent v2

Modular AI-powered trading engine. Event-driven architecture with pluggable signal agents (LLM + ML), exchange adapters, and a conviction-based execution pipeline.

Architecture

SENTINEL (persistent, code-only, per symbol)
  │ monitors price, computes readiness, manages SL/TP
  │
  └── ANALYSIS PIPELINE (spawns TraderBot)
        ├── 1. DATA LAYER      (OHLCV, indicators, flow, parent TF)
        ├── 2. SIGNAL LAYER    (3 LLM + N ML agents via SignalProducer)
        ├── 3. CONVICTION      (meta-evaluator, scores 0-1)
        ├── 4. EXECUTION       (action selection + mechanical safety)
        └── 5. REFLECTION      (async, post-trade rule distillation)

Setup

# Clone
git clone https://github.com/Alishark14/quantagent_v2.git
cd quantagent_v2

# Create venv
python3.12 -m venv .venv
source .venv/bin/activate

# Install
pip install -e ".[dev]"

# Configure
cp .env.example .env  # fill in API keys

# Run tests
pytest tests/

Commands

# Start the full server (BotRunner + API)
python -m quantagent run

# Run database migrations
python -m quantagent migrate

# Seed dev database
python -m quantagent seed

API

The API runs on port 8000 by default. Auth via X-API-Key header.

Endpoint	Method	Description
/v1/bots	POST	Create bot config
/v1/bots	GET	List user's bots
/v1/bots/{id}	GET	Bot detail + last cycle
/v1/bots/{id}	DELETE	Stop a bot
/v1/bots/{id}/analyze	POST	Trigger manual analysis
/v1/trades	GET	Recent trades (filters: bot_id, symbol, limit)
/v1/trades/{id}	GET	Trade detail
/v1/positions	GET	Open positions
/v1/rules	GET	Active reflection rules
/v1/health	GET	System health (no auth)

Tech Stack

• Python 3.12+, async/await throughout
• Claude API (Anthropic) for signal + conviction agents
• FastAPI + uvicorn for the API layer
• PostgreSQL (asyncpg) / SQLite (dev fallback)
• Alembic for database migrations
• CCXT for exchange connectivity (Hyperliquid primary)

Project Structure

engine/          Pure library — zero web/DB imports
sentinel/        Persistent market monitors
exchanges/       One adapter per exchange
llm/             LLM provider abstraction
storage/         Repository pattern (PostgreSQL / SQLite)
tracking/        Observability + data moat capture
api/             FastAPI web layer
quantagent/      CLI + BotRunner

License

Proprietary. All rights reserved.