Build a RAG chatbot trained on up to 50 pages of your documents

Staff and customers keep asking things that are already written down somewhere: in a manual, a policy PDF, a product sheet. Finding the right page takes longer than asking someone.

I'll build a chatbot that answers from your documents and only from them.

What you get:
- Your documents loaded and indexed, up to 50 pages (PDF, Word, web pages or Google Docs)
- Answers that cite the document and section they came from, so anyone can check
- A clear "that isn't in the documents" reply when the answer isn't there
- A web chat page or a widget for your site
- A test run on 20 real questions from you, with the results shared before handover
- A simple way to add or replace documents later

This is RAG (retrieval-augmented generation). The bot searches your files first and writes the answer from what it finds. I built the same kind of document search for Telego, a project management tool in Telegram that answers questions from project documentation.

The model runs through your own OpenAI or Claude account. For 50 pages the usage cost is usually small, and we'll set a monthly limit together.

Before ordering, tell me the file types and whether any documents are scanned images. Scans need OCR first, which I'll quote separately.