sudarsan2507-hue/DemoSearch — reverse-engineered prompt

I want this turned into a usable Python research demo for adaptive search on uncertain knowledge graphs. There are a lot of experiment scripts here, but not much documentation, so please make it easy to run from one main entry point instead of having to guess which file does what.

The app should let me run a basic search baseline and a few smarter search variants, compare how reliably they find good paths when the graph is noisy or incomplete, and save clean outputs in the results folder, like metrics, summaries, and a few simple plots. If several scripts are overlapping, please tidy that up so the project feels coherent.

Give me sensible defaults, but also let me change things like search budget, beam size, embedding choice, verifier or shield settings, and random seed. Make sure it installs and runs cleanly from the requirements file, handles missing inputs gracefully, and add a short README with exact commands for reproducing a couple of experiments. If needed, look up current docs or common research conventions online.