Improve production ASR fallback handling

Add real-time voice architecture demo with HF asset recommendations

Phase 4 - Add real HF LLM architecture generation

Phase 3 - Add Hugging Face ASR transcription