gabsh/mlens — reverse-engineered prompt

Build me a small full stack app for benchmarking movie review sentiment models on the Stanford IMDB dataset.

I want it to download the data, train a bunch of embedding and classifier combinations, save the trained models and benchmark results, and keep experiment history I can inspect locally. Then give me a backend API where I can paste in a review and get back the predicted sentiment plus a simple word level explanation of why the model thinks that, using LIME style highlights or scores.

On the frontend, make a clean Vue app where I can compare model performance, browse the benchmark results, and test predictions in a little inference screen. It should feel like a simple dashboard, not just raw JSON. Please wire it so I can run the training flow, start the API, and run the frontend locally without a lot of manual setup. If there is already deployment stuff for containers and a small VPS cluster, keep that structure in place too. Look up current docs online if you need to.