How answers work (RAG)
How your chatbot answers from your own documents and website, with citations.
schedule4 min readNivi Chat answers using retrieval-augmented generation (RAG). Instead of guessing from general knowledge, your bot searches the knowledge base you built from your own content, then writes an answer grounded in what it found — and cites the source it used. That is what keeps replies accurate and on-brand.
What your bot learns from
Everything you add on the Train page becomes searchable knowledge:
- Documents — PDFs, Word docs and text files such as product guides, policies and manuals.
- Website — paste a URL and Nivi Chat reads the page content.
- FAQ / text — paste answers to common questions directly.
How a question gets answered
- 1
Search your knowledge
When a visitor asks something, Nivi Chat finds the most relevant passages from your sources. - 2
Answer from what it found
The bot writes a reply grounded in those passages, rather than inventing an answer. - 3
Cite the source
Where it can, the reply points back to the content it used, so answers are traceable.
Frequently asked questions
What is RAG (retrieval-augmented generation)?
It is a method where the chatbot first retrieves relevant passages from your own content, then generates an answer based on them. This keeps answers grounded in your material instead of general guesses.
Will the chatbot make things up?
RAG reduces that by answering from the sources you provide and citing them. Accuracy depends on the quality and coverage of the content you train it on.
Which plan includes a knowledge base?
A document knowledge base is available from the Starter plan (₹999 per month) and up. The Free plan includes the website chat widget.
Can I update the knowledge later?
Yes. Add or remove sources any time on the Train page, and your chatbot uses the latest knowledge right away.