Lower ZeroGPU reservation for anonymous users

Update custom Server UI

Cap generated section output

Tighten separate section prompts

Generate appointment sections separately

Serialize llama.cpp warmup and generation

Warm llama.cpp model on page load

Tune llama.cpp latency

Add llama.cpp GPU verification logs

Tighten 4B report prompt

Switch Space to llama.cpp MedGemma 27B GGUF

Update medication search and patient-friendly copy

Update app UI and model settings

Fix ZeroGPU inference argument pickling

Deploy initial Gradio Space