I need an AI-driven indicator that helps me make same-day decisions on U.S. equities. The tool has to blend three information streams—historical price action, live market ticks, and real-time news sentiment—into clear, low-latency buy/sell signals suitable for a fast intraday workflow. Core expectations • Data: pull and preprocess historical OHLCV, stream live quotes, and mine headline sentiment in real time (Python, pandas, WebSocket feeds, NLP libraries such as spaCy or transformers are fine—use what you are comfortable with). • Model: train a lightweight yet accurate model (e.g., LSTM, Transformer, XGBoost) that can update incrementally during the session without full retraining. • Output: visual overlay or dashboard with entry/exit cues, confidence scores, and stop-loss suggestions; it must refresh in seconds, not minutes. A TradingView Pine v5 script, a lightweight desktop app, or an MT5 plug-in are all acceptable—pick the environment you can deliver fastest. • Validation: share back-tests on at least two years of 1-minute equity data plus a live forward test (paper account is fine) that runs for one trading day to show latency and stability. • Handover: commented source code, model weights, environment.yml/requirements.txt, and a concise setup guide I can follow on Windows. Acceptance criteria 1. Average signal latency under two seconds from quote arrival. 2. Sharpe ratio ≥ 1.5 in the back-test after commissions. 3. No external paid APIs beyond what I already have (Alpaca data and NewsAPI). If you’ve built intraday AI indicators before and can hit these benchmarks, I’m ready to start immediately and will be available for quick feedback as you iterate.