A Botpress chatbot grounded in a vector knowledge base, built to answer customer questions accurately instead of guessing.
Problem
Generic scripted chatbot flows couldn't answer specific product questions, so support tickets kept piling up and website visitors left without the answer that would have turned them into a lead.
Solution
Combined Botpress's conversation flows with the OpenAI API and a vector database so every answer is grounded in the company's actual documentation, cutting support tickets by 50% and lifting lead capture by 25%.
Architecture
Source content is chunked and embedded into a vector store; each incoming message is embedded and matched against it at query time, and the retrieved context is passed to OpenAI to generate a grounded response inside the Botpress flow.
