andreakone/Kronos ? reverse-engineered prompt

Build me a working open source project called Kronos for forecasting financial candlestick data from historical market prices.

I want the app or library to take in OHLCV style time series data with timestamps, run it through the model, and return future predictions in a simple format people can use right away. Please include a clean Python API for loading a pretrained model and tokenizer, making single and batch forecasts, and handling the usual preprocessing and normalization behind the scenes. If it makes sense, add a small web demo where I can upload or choose market data and see the forecast visually, similar to the BTC USDT demo in the README.

Also include the fine tuning scripts so I can adapt the model to my own data, plus a few basic tests and example notebooks or scripts showing how to use it. Keep the code easy to run locally with clear setup instructions, and look up current docs online if you need to fill in any missing details.