Hi everyone! I'm the maker of LocalRAG. I built this because I wanted to chat with my PDF documents using AI, but I wasn't comfortable uploading sensitive files to cloud services like ChatPDF. Tax returns, medical records, client contracts — these shouldn't have to leave my device just to ask a question about them. **The core idea:** All document processing happens on your iPhone/iPad. Text extraction, chunking, and TF-IDF indexing are 100% local. Only the relevant text chunks (not the full document) are sent to Claude AI when you ask a question. **Technical approach:** - NLTokenizer for sentence-level chunking (500 chars, 80 char overlap) - TF-IDF cosine similarity + bigram matching for hybrid search - 3-stage PDF extraction: PDFKit → copy-protection bypass → Vision OCR - Supports 9 document formats natively **Business model:** Free tier (5 questions/day with Haiku), paid subscriptions for more usage, or bring your own Anthropic API key for unlimited access with any Claude model. I'd love to hear your feedback. Privacy-first AI is still a niche, but I think it's an important one.

LocalRAG!

AI knows everything — except what's in your documents

AI knows everything — except what's in your documents