Add-it-ya/RL_Optimal_Trade ? reverse-engineered prompt

Build me a Python project for reinforcement learning based optimal trade execution.

I want an app that simulates a market and lets an agent decide how much of a large order to trade at each step so it can beat simple strategies like TWAP, VWAP, POV, Random, and a classic Almgren Chriss baseline. It should support both discrete and continuous actions, use a Gymnasium style environment, and include a simple limit order book simulator with spread, imbalance, volatility, and market impact. Please add a few RL agents, including DQN, Double DQN, PPO, A2C, and SAC, plus training, evaluation, and backtesting scripts.

I also want metrics like implementation shortfall, execution cost, inventory path, and average fill price, along with charts, a research report, and a Streamlit dashboard to explore results. If helpful, look up current docs online for any library details you need. Make it work as a clean, production style repo with tests, config files, and a quick smoke run that trains and evaluates everything end to end.