subrat-dwi/prag-personal-rag

PRAG is a personal assistant that answers questions about you using your own documents. The idea is simple: instead of digging through old files, photos, messages, and emails to find something like your college roll number or the exact date you started an internship, you just ask PRAG and it looks it up for you. You feed it your personal documents in formats like PDF, Word files, images, markdown, and plain text. For scanned documents or photos, it uses OCR to pull out the text, then cleans up any messy results with a language model. It can also sync directly with Google Drive, checking file checksums so it only reprocesses files that actually changed, and removing old data automatically when a file is deleted. When you ask a question, PRAG combines two kinds of search: one based on meaning, which catches questions phrased differently than the original document, and one based on exact keyword matching, which is better for things like ID numbers or specific terms. It also tries to correct spelling mistakes in your question and figures out whether you are asking for a plain fact or something that needs pulling information together from multiple places. Every answer includes a citation showing which document it came from, along with a link back to the file in Google Drive if it is stored there. The language model itself can run locally through Ollama for privacy, or switch to a faster hosted option for deployment, changed with a single setting. The whole system, including its web interface that installs like a mobile app, is packaged with Docker so it can be set up the same way locally or in production.