I am building a cross-platform desktop application (Windows/Mac/Linux) similar in concept to NotebookLM, but fully offline. The app should: • Import documents (PDFs + scanned images) • Run OCR locally (Tesseract) • Split text into chunks • Generate embeddings locally • Store embeddings in SQLite (or local vector DB) • Use llama.cpp for local LLM inference • Enable chat with documents (RAG architecture) • Show source citations in answers Stack requirements: • Electron + React + TypeScript • SQLite (with vector search support preferred) • Local llama.cpp integration • Tesseract OCR integration • Fully offline (no cloud APIs)