Switch to faster-whisper for 4x CPU speedup

Fix: skip OpenAI Whisper API for Nvidia, use tiny model by default

Speed up CPU transcription: remove thread limit and disable fp16

Fix TimestampMapper instantiation argument error

Fix critical issues: API key race conditions, unused pipeline stages, port mapping

chore: commit remaining local changes - quiz/topic extractors and helper utilities

fix: notes generation works in all modes - fixed _llm_complete removal, _build_notes_from_chunks fallback, translate_notes, _summarize_huggingface

fix: save outputs inside final_notes folder to align with download endpoints

fix: import os at top level of all service modules to avoid NameError

fix: propagate runtime and environment-based OpenAI API key to all downstream services

Update default port to 7860 for Hugging Face Spaces compatibility

Limit PyTorch to 1 thread to fix freeze on Render

Add OpenAI Whisper API for fast transcription on Render

Fix memory leak and background task blocking

Updated project files

Updated project files

initial commit