Final hackathon submission: README, Field Notes, UI overhaul, fine-tuned model

Reproducción inline de canciones y cuentos, mic como botón de terminar

Add real song playback, precache stories, improve LLM fallback

Add Sofia color-change, Cuentos frontend section, Ollama local backend

Tune ZeroGPU queue duration; document per-visitor daily quota

Simplify startup flow: drop scripted greeting, start from child's first turn

Fix ZeroGPU import order: spaces before torch

Add retry + curated-content fallback for ZeroGPU allocation failures

Pivot LLM runtime: llama.cpp (CPU) -> transformers + ZeroGPU

Fix LLM perf on free CPU tier: pin n_threads=2, warmup at startup

Rename persona Lumi -> Sofía, prep for HF Space deploy

Initial commit: Lumi pipeline migrated to llama.cpp, local-first TTS