ujjwaltwri/fairlens ? reverse-engineered prompt

Build me a web app called FairLens that helps people check whether a machine learning model or dataset is showing hidden bias. I want a simple dashboard where someone can pick one of the built in example datasets or upload their own CSV, run an audit, and then see clear results about fairness problems like disparate impact, parity gaps, and other bias checks in plain language.

It should also train a basic model, show an explanation of what features are influencing the decisions, and suggest a few ways to reduce the bias. After the audit, the app should generate a downloadable PDF and JSON report, and keep track of completed jobs so the user can come back later and view the results. If it makes sense, include a small chat or assistant view that can answer questions about the finished audit.

Use the existing backend and frontend setup, and look up current docs online if you need to.