Support answers backed by your documents
Upload your documentation once. Answers are retrieved from your knowledge base, cited to the source document, and passed to a human when the answer isn't there.
AI
OpenAI embeddings
RAG
Similarity search in Postgres
Selected
Every answer maps to a source chunk
The RAG pipeline
From raw document to grounded answer
Each stage runs server-side and reports back in real time, so you can see where an answer came from.
Upload
PDF, DOCX, Markdown and text files land in your knowledge base.
Process
Documents are cleaned and split into overlapping, context-aware chunks.
Embed
Each chunk becomes a 1536-dimension vector using OpenAI embeddings.
Store
Vectors are indexed in Postgres with pgvector for fast similarity search.
Retrieve
Every question pulls the most relevant chunks with a scored match.
Answer
The model answers strictly from retrieved context, with citations.
The team 👋
Built by two students
Design, retrieval and backend — shipped end to end. 🚀
Vikas Gupta
🎨 UI/UX & RAG
🎓 3rd year
Vaibhav Srivastav
⚙️ Front-end
🎓 3rd year
Try it with the sample knowledge base
Five pre-loaded documents cover products, pricing, refunds, security and support. Ask a question, then upload your own files.
Open the RAG Assistant