aisearch-openai-rag-audio

Overview of aisearch-openai-rag-audio

What it is

This repository provides a simple example implementation of the VoiceRAG pattern. It is designed to power interactive voice generative AI experiences using Retrieval Augmented Generation (RAG) with Azure AI Search and Azure OpenAI's gpt-4o-realtime-preview model. The application uses the browser's microphone for voice input and plays audio responses from the GPT-4o Realtime API.

Key features

• Voice interface: Captures voice input via the browser's microphone and processes it with the Azure OpenAI GPT-4o Realtime API.
• RAG (Retrieval Augmented Generation): Uses Azure AI Search to retrieve relevant documents from a knowledge base and sends them to the GPT-4o Realtime API for response generation.
• Audio output: Plays the generated response as audio using the browser's audio capabilities.
• Citations: Displays the search results that were used to generate the response.

Use cases

This pattern is suitable for applications that require a voice-based user interface combined with retrieval-augmented generation. Example use cases include interactive voice assistants that answer questions based on a specific knowledge base, such as internal company policies or product documentation.

Getting started

To deploy the app to Azure Container Apps, run the following commands after cloning the repository and logging into Azure:shellazd auth loginazd env newazd upTo run the app locally, create an app/backend/.env file with the required environment variables and then start the app:shell./scripts/start.shFor detailed instructions, see the repository README.

License and pricing

This project is licensed under the MIT license. Pricing for the Azure services used (Azure Container Apps, Azure OpenAI, Azure AI Search, Azure Blob Storage, Azure Monitor) varies by region and usage; see the Azure pricing calculator for estimates.

Based on the project's public README and repository metadata.