sapphire-alpha

name	sapphire-alpha
description	Trading engine — signals, risk, execution, self-improvement loop
type	service
runtime	python
deploy_target	rari1+rari2
dependencies	["sapphire-core","sapphire-telegram"]
entry_point	src/main.py
test_command	pytest tests/

services/alpha

Core trading engine. Receives signals from TradingView webhooks, applies risk filters via sapphire-core, executes on Lighter Protocol, and publishes trade events to the event bus.

Components

• Signal ingestion (TradingView → webhook → alpha)
• Risk kernel (circuit breaker, position sizing, drawdown limits)
• Execution router (Aster DEX + Hyperliquid via rari2)
• Self-improvement: trade metrics → control-plane tasks → agent improvements

Signal Flow (on-prem)

TradingView → webhook.sapphirealpha.xyz (Cloudflare Tunnel)
  → Windows PC:9090 (Ollama enrichment)
  → rari1:18081 + rari2:18081 (Tailscale HTTP, parallel)
  → sapphire-core risk check
  → Redis pub/sub: trading-signals → aster, hyperliquid
  → control-plane:8082 → Telegram notification

Metrics (PnL is king)

• Win rate target: 80%+
• Risk metric: Sortino + Calmar (not Sharpe)
• Max drawdown: 15%
• Pairs: ETH/BTC, SOL/BTC, ZEC/BTC, HYPE/USDT, BTC/USDT

Deploy (on-prem)

Runs on both rari1 and rari2 (port 18081). See infra/pi/ for systemd services. Signal broker: Redis on rari1 (REDIS_URL=redis://100.x.x.x:6379).

# Install Redis on rari1 first:
ssh rari@100.x.x.x 'bash -s' < infra/pi/rari1/setup-redis.sh
# Then deploy control-plane:
bash infra/pi/rari1/deploy-control-plane.sh