the backend should have a env and app should too so that ai response route over a weak model just to test stuff . mock ai response works but sometimes causes issues . this should fix it cause we can then do rapid test. gemini flash models can be used here