I Got Tired Of Parsing Websocket Dumps For My RL Bot, So I Structured A Clean SOL Level 2 Dataset.

Hey guys. I’ve been training reinforcement learning models and pulling my hair out trying to get clean, historical order book depth data without paying institutional API fees. I finally set up my own pipeline to capture L2 depth and imbalance for SOL. I formatted it specifically for backtesting and feature engineering. You can grab the dataset here:https://imbalancelabs.com/datasets/sol-historical-orderbook. What specific L2 features (besides standard imbalance and spread) are you feeding into your models lately?