ozturkoktay/insurance-llm-framework ? reverse-engineered prompt

Build me a simple web app for experimenting with open source language models for insurance work.

I want a place where I can load a model, write and save insurance specific prompts, generate responses, and compare how different models handle the same task. It should help with things like policy summaries, claim reply drafts, risk notes, customer messages, underwriting help, and compliance checks. I also want an evaluation area where I can score outputs with common text metrics and do side by side comparisons, plus a way to create small benchmarks from sample insurance data and track results over time.

Make the interface easy to use, with a clean dashboard for model selection, prompt editing, generation, evaluation, and benchmark runs. It should work well on both CPU and GPU, let me choose basic settings like quantization, and let me save or export generated outputs. If you need current docs for any model or library, look them up online first.