JCHETAN26/Volatix-AI — reverse-engineered prompt

Build me a working local demo of Volatix AI that feels like a real time market surveillance tool. I want it to ingest live or mock trade data over websockets, calculate fast market features like order flow imbalance and realized volatility, send those into a stream, score suspicious activity with a simple ML anomaly model, and then have an agent flow that investigates the event, writes an audit trail, and can mark high confidence cases as frozen.

I also want a clean web control board where I can watch the stream, see alerts appear live, inspect why something was flagged, and review the agent decisions and history. If real market keys are not set up, use the mock ticker so the whole thing still works out of the box. Use the hosted Postgres setup mentioned in the repo for logs and metadata, and wire up the vector store piece only as far as needed for the investigation flow.

Please get the repo into a runnable state with sensible defaults, setup scripts, and a clear path to launch everything locally. Look up current docs online if you need to.