Autonomous-RL-Trading-Bot

This project is a professional-grade autonomous trading assistant that combines Reinforcement Learning, Technical Analysis, and Market Sentiment Analysis. The system scans market data, calculates advanced technical indicators, and generates AI-driven buy/sell signals.

Key Features

• PPO-Based Decision Engine: Autonomous decision-making using deep neural networks trained with the Stable Baselines3 PPO algorithm.
• Market Intelligence: Real-time financial news scanning, Reddit discussion tracking, and sentiment analysis using FinBERT.
• Advanced Risk Management: Implementation of Kelly Criterion, Volatility Scaling, and dynamic Position Sizing to protect capital.
• Integrated Alert System: Real-time push notifications of AI decisions and trading signals via Telegram.
• Live Paper Trading: Strategy testing using live market data within a simulated paper trading environment.

Technical Stack

• RL Environment: Custom Gymnasium implementation (StockTradingEnv).
• Data Source: Real-time and historical price data provided via yfinance.
• Technical Indicators: Calculation of RSI, MACD, Bollinger Bands, ATR, OBV, and other quantitative metrics.
• UI/UX: Interactive dashboard built with Streamlit and advanced financial charting using Plotly.

Quick Start

1. Install Dependencies:

   pip install -r requirements.txt

Technical Notes and Usage

• Execution Command: The command streamlit run app/app.py initializes a local web server and triggers essential setup functions such as apply_styles() and init_session_state() within the main application file.
• Dependencies: It is critical that all libraries listed in requirements.txt—specifically stable-baselines3 and torch—are installed correctly to prevent errors during model loading in rl_agent.py.
• Sidebar Integration: The user interface utilizes sidebar components for core functionalities including model training (train_agent), asset selection, and Telegram notification configuration via the SignalBroadcaster.
• Directory Structure: The application expects a specific directory structure for saving and loading AI models; specifically, it utilizes the models/trained/ path for storing .zip and .pkl files.