🐛 Correctly unset loaded state after unloading

✨ Implement lazy loading for models and correct tokens counting

✨ Add Gemma 4 E2B model integration and update service to support multiple models

⚗️Move spaces.GPU to app.py endpoints

🩺 Add debug statement for Llama text generation

👔 Update model to immediately initialise instead of lazy loading

✨ Implement model gateway with Llama model integration and /generate API endpoint