ConversationRelay makes it easy to connect your users to an AI agent, but how do you ensure that agent is giving your user the right information and not hallucinating? OpenRAG is an open-source stack for retrieval-augmented generation (RAG) built on OpenSearch for search, Docling for parsing of unstructured data, and Langflow for agent coordination.

In this talk, you'll see how to build a robust agent that can answer the toughest questions about your data. You'll learn how to set up OpenRAG and connect it to ConversationRelay, how to tweak your ingestion and search pipelines to improve the results, and how to use Conversational Intelligence to observe and validate the behaviour of your agent. Overall, you'll see how to combine a hackable RAG stack with the Twilio platform to create knowledgeable, intelligent agents you can trust.