TradeLix AI

TradeLix AI is a FastAPI + React trading-agent PoC for the Alpaca AI Trading Agents Hackathon. The system runs a linear multi-agent pipeline: news, sentiment, options context, market features, technical indicators, order blocks, institutional flow, account state, risk, decision, and execution gate. The key design principle is: LLMs propose, deterministic code authorizes. Groq-powered agents and optional ReAct loops can research headlines, inspect compact market signals, reason about BUY/SELL/HOLD, and explain their proposal, but they cannot place trades. Every proposal must pass a deterministic execution gate that checks paper mode, buying power, exposure, duplicate working orders, kill switch, schedule window, and plan validity. Execution is disabled by default and only sends Alpaca paper orders when explicitly armed. The dashboard streams the agent pipeline over SSE, displays candlestick charts and smart-money overlays, shows gate verdicts, supports dry-run previews, and exposes the audit trail so users can inspect why a trade was allowed, blocked, or skipped.