Complete per-user data isolation, bug fixes, and Admin wipe trigger

Gracefully disable YouTube in Cloud Mode

feat: use Invidious API for YouTube transcripts

feat: browser-side YouTube transcript fetching

Replace YouTube fetch with timedtext API - no API keys needed, works from any server

Add YouTube Data API v3 for cloud-reliable transcript fetching with local fallback

Add auto-generated caption support for YouTube: multi-priority fetch, shorts URLs, transcript cleanup

Fix YouTube transcript API: migrate to v1.x instance-based API (fetch + t.text)

Add OCR fallback for scanned/image PDFs using pytesseract + pdf2image

Fix PDF upload empty embeddings error: add validation for scanned/image-only PDFs

fix: Use langchain_core.documents instead of deprecated langchain.schema

fix: Define logger before use in import error handler

fix: Move youtube-transcript-api imports to top level

feat: Enable YouTube URLs in cloud with youtube-transcript-api

fix: Add logger and fix imports for URL ingestion

fix: Add missing logger import

fix: Improve URL ingestion error handling and connectivity check

Fix: Handle AIMessage in all LLM invoke calls

Fix: Handle AIMessage response from ChatHuggingFace

Remove duplicate get_llm() from rag_engine.py - use config.py version

Fix: Use HuggingFace embeddings in cloud mode

Add hybrid deployment: Ollama (local) + Hugging Face (cloud)

Production code cleanup - remove debug statements and duplicates

Add data persistence and comprehensive citation system

Major improvements: Multi-subject plan generation, enhanced quiz/lesson quality, fixed UI layout

feat: Implement AI Plan Generator (Lite Mode) and Focus Mode Layout

Update backend RAG logic, app UI, and add documentation

Initial commit of FocusFlow backend