A chatbot that answers from your own documents instead of making things up.

This is for a business whose knowledge sits in places nobody can search — product manuals, policy documents, SOPs, past proposals, support history, a messy shared drive. Your team asks the same questions every week and someone senior keeps answering them.

RAG is the technical name for the method: before the model answers, the system retrieves the relevant passages from your own documents and answers from those. The practical result is that it cites where the answer came from, and when your documents do not contain the answer, it says so instead of inventing one.

What gets built:

Your documents are processed and indexed into a vector database. The assistant retrieves the right passages for each question and answers from them, with the source shown. It handles follow-up questions in context. You can add new documents later without a rebuild.

Where it lives is your choice: a widget on your website, an internal tool for the team, or inside the chat platform your team already uses.

How the work runs:

Week 1 — document ingestion, indexing and retrieval, then the first version of the assistant.

Week 2 — we test it against the questions your team actually asks, measure where it gets them wrong, and tune retrieval until it is right. This second week is what separates a usable assistant from a demo.

Built with Pinecone or pgvector for retrieval, Supabase for storage, and Claude or OpenAI for the answers.

30 days of support after launch.