Sumamasonia/behaviorci ? reverse-engineered prompt

Build me a small app for testing whether an AI product changed in a bad way after I update the model or prompt.

I want to define test cases in YAML, run them against any HTTP endpoint that takes a prompt and returns an output, and then get a clear report showing what changed compared with the last run. It should catch more than just score drops, like tone changes, reasoning changes, output quality, and cases where something still passes but the style drifted.

Please include a simple web dashboard, a REST API, and a command line tool so I can sync test suites and run them easily. Use free local defaults if possible, like SQLite and a local model for judging, so I can try it without any paid services or API keys. A sample customer support test suite would be helpful too.

If you need to, look up current docs online for any local model or library details.