Spaces:
Running
Running
Khanna, Videh Rakesh Rakesh
fix: load dotenv in ai_forecast, improve top5 5D scoring, harden HF repo IDs
04219e7 | title: PaperTrade | |
| emoji: π | |
| colorFrom: blue | |
| colorTo: green | |
| sdk: docker | |
| pinned: false | |
| app_port: 7860 | |
| # PaperTrade β NSE Indian Equity Prediction Engine | |
| A paper trading system for NSE Indian equities. Predicts short-term price direction (1D/3D/5D) using backtested technical strategies, an ML feature scorer, macro gates, news sentiment, and an LLM-based directional forecast. | |
| ## Recent Reliability Improvements | |
| - Manual trade entries now attempt a best-effort auto-scan at order time to populate missing strategy, timeframe, and prediction context. | |
| - Post-mortems for manual trades now include concrete trade-window price diagnostics (swing, MFE, MAE, trend) to avoid generic commentary. | |
| - Frontend trade submit now waits for in-flight watchlist context fetch before posting, reducing empty post-mortem context. | |
| - Timeframe calibration in the AI forecast path was tightened to use shallow bearish midpoint ranges and safer weak-bear handling. | |