← back to work

mindsync

fastapi · qdrant · next.js — rag document q&a delivered over whatsapp, with an admin dashboard and cited answers

view live →airagfull-stack

the problem

Campus regulations, schedules and announcements live in PDFs that nobody reads, so students ask staff the same questions over and over. The answers already exist in writing — they are just not reachable from where students actually are.

what i built

Administrators upload the official documents, including scans, which are OCR'd, chunked and embedded into Qdrant. Students then ask questions in plain language over WhatsApp and get an answer with a citation back to the source document, so nothing is invented. The admin dashboard covers document management and categorisation, usage analytics, an editable system prompt, and a RAG playground for testing retrieval before it reaches students. The WhatsApp side runs through self-hosted WAHA with typing indicators and rate limiting. LiteLLM sits in front of the model layer so providers can be swapped without touching the retrieval code.

stack

fastapipythonpostgresqdrantlitellmtesseract ocrnext.jsshadcn/uitanstack querydocker compose

interested in working together?

get in touch →