issues fixed in three files

changed llama cpp version

new way of building application

extra system dependencies for PyAV

numpy sqlite fixed

python 3.10 to 3.11

Production-ready AARA: persistent models, structured logging, rate limiting

Refine browser voice flow and docs

Add WhatsApp voice-note replies

Remove legacy app and reset browser sessions

Deploy browser voice app to HF Space

Fix XTTS license gate

Align runtime to Hugging Face models

Fix: Compatible Gradio + Llama/XTTS open-source stack, auto-download models

Fix: Upgrade to Gradio 5.1.0+ for huggingface_hub 1.9.0 compatibility (HfFolder removed in newer versions)

Fix: Remove Gradio from requirements.txt, pin Gradio 4.39.0 in Dockerfile (HfFolder fix)

Comprehensive fix: Auto-install Gradio, clean Dockerfile, minimal deps

Simplify: Minimal requirements.txt for HF Spaces

Fix: Remove vocabulary.json and allow source builds for Gradio

Fix: Replace llama-cpp-python with ctransformers (no compilation needed)

Optimize: binary wheels only, skip build-time DB init

Fix llama-cpp-python: use pre-built wheels instead of building from source

AARA Voice Agent - Code only, WAV files via web UI