A chatbot that answers from your documents, and says when it doesn't know.
X1Y1 builds retrieval-augmented generation (RAG) systems for startups. For a fixed $1,500 and in 2 weeks, we index up to 500 of your documents and ship a chatbot or API that answers from them, cites its sources, and says 'I don't know' when the answer isn't there. You also get a 30-question test set so you can see the accuracy yourself.
What you get
- Up to 500 documents indexed
- Hybrid search (keyword plus vectors), so exact terms and IDs still match
- Reranking, so the right passage reaches the model
- Answers that cite their sources, and an honest "I don't know"
- A 30-question test set so you can see the accuracy
- A chatbot or an API, whichever fits your product
Who it is for
- Startups adding "ask our docs" or support answers to their product
- Teams whose current RAG demo gives confident wrong answers
- CTOs who want accuracy they can measure before they ship
Questions
Why does my RAG system hallucinate?
Usually because of retrieval, not the model. The right chunk is never retrieved, retrieved chunks contradict each other, or the chunk is correct but missing context. Hybrid search, reranking and chunking on document structure fix most of it. Changing the model rarely does.
How much does a RAG system cost to build?
Our RAG offer is a fixed $1,500 for up to 500 documents, delivered in 2 weeks, including a 30-question test set to measure accuracy.
How do you measure RAG accuracy?
With a test set of real questions and known answers from your documents. We build 30 of them with you, so you can re-run the test after every change and see whether answers got better or worse.
Who owns the code?
You do. All code and docs we write are yours.