aisearch-openai-rag-audio

Why we included this project

Voice assistants that answer from your own documents are tricky to build because the audio path and the retrieval path have to work together in real time. This sample shows one working way to combine them: the browser captures speech, a Python backend relays it to OpenAI's realtime audio API, and a search tool pulls relevant passages from Azure AI Search so the model grounds its spoken answer in your knowledge base. The repo includes infrastructure as code, a Dockerfile for Azure Container Apps, and dev container and Codespaces setups, so you can go from clone to a running app without much setup. It is a reference implementation rather than a finished product, but the architecture diagram and the middle-tier code give you a concrete starting point to adapt, especially if you are already working in the Azure ecosystem.