yansari/arabic-pronunciation-trainer

fix: Add uid 1000 user and writable cache dirs for HF Spaces runtime

fix: Add robust audio format support (WebM/Ogg/M4A) and production bug fixes

fix: Convert numpy bool to Python bool for JSON serialization

fix: Resolve cache permissions and processor loading issues

chore: Trigger HF Space rebuild

fix: Load model from HuggingFace Hub when local files not available

fix: Add python-multipart for FastAPI file upload support

fix: Add missing torch dependency and fix matplotlib cache permission

feat: Add HF Spaces configuration and optimize Dockerfile

feat: Add smart audio extraction and API documentation

Update .gitignore to exclude PNG files for HF Space compatibility

Merge: Resolve conflicts, keep smart extraction implementation

feat: Add smart audio extraction with hybrid center+energy approach

Configure for Google Cloud Run deployment

Fix model download: disable symlinks for proper tensor loading

Add error handling for microphone access

Download model from HF Hub on Docker build

Add python-multipart for file uploads

Fix: Add torch to requirements and fix matplotlib permissions

Restore original HF Space configuration (Docker SDK)

Revert README to original - remove HF Space metadata that broke deployment

Add Hugging Face Space metadata to README

Remove binary PNG for HF Space compatibility

Optimize /verify_letter endpoint response for production

Add /verify_letter endpoint for automated testing

Add pronunciation hints system with detailed anatomical guidance

README: full project description & Linux-proven quickstart

Ignore hf_export artefacts

Make install.sh cross-platform (Windows Git Bash + Linux/macOS), add HF Python fallback

Update README.md

Fix run.sh: robust venv activation and execution

Initial commit: code, scripts, and diagnostics (model via HF)