This post has been republished via RSS; it originally appeared at: Microsoft Tech Community - Latest Blogs - .
We recently concluded our first Microsoft AI Chat App Hack, a hackathon which challenged developers to build applications using RAG (Retrieval Augmented Generation) in order to answer questions in custom domains. You can browse through the winners and submissions to see what developers built.
To help developers understand the new world of generative AI, we held 17 live streams across four languages. If you missed the hackathon this time around, you can still watch the stream recordings from the links below and start building an AI chat app today.
If you're an educator or student organizer, you could even run your own hackathon! We've provided links to the slide decks and demos. Post in the discussion forum if you're looking for any additional resources to help your event.
English streams
|
Building a RAG Chat App in Python Want to learn how to create a chat app on your own data using Python? This session shows you how to use RAG (Retrieval Augmented Generation), a powerful technique for combining knowledge retrieval with LLMs. You'll learn how to combine OpenAI with Azure AI search and Azure Document Intelligence to ingest and vectorize data, and then deploy a frontend to chat with that data on Azure App Service. |
|
Ready to customize your RAG app for your own data and organization? This session shows you how to bring in your own documents and websites. It also includes tips on customizing the frontend to add your own branding, and ideas for additional features you might want to bring into your chat app. |
|
Azure AI Search Best Practices When you're building an AI chat app using RAG, the best way to get great answers is to get great retrieval (the "R" in RAG!). In this session, learn about best practictices from an Azure AI Search engineer for indexing and querying documents. You'll discover the differences between text search, vector search and hybrid search, and see the benefits of the semantic reranker. You'll also learn about custom analyzers, integrated vectorization, and more. |
|
Connecting a RAG chat app to Azure Cosmos DB Learn how to use RAG with Azure Cosmos DB for MongoDB vCore. This session demonstrates how to efficiently store and retrieve transactional and vector data together. It also walks through creating a low-code RAG application in Azure OpenAI Studio with just a database, coding the retrieval component in Python, and explains when to choose Azure Cosmos DB for MongoDB vCore for your RAG implementations. |
|
Ready to see the future of AI? You can now build AI chat apps that can answer questions based on images - like photos, graphs, diagrams, and illustrations. Learn how to use the Azure Computer Vision multi-modal embedding API, Azure AI search to index and query images, and then use GPT-4-Turbo with Vision to answer questions about images. This cutting edge technique can be an amazing approach for document types that are image heavy, like financial charts. |
|
Web Components are the new standard way to build reusable custom elements for web sites, and our team has created a set of web components for you to build AI chat apps. This session shows you how to use a frontend built entirely of web components, and how to easily swap your current frontend with a modern web component based frontend. |
|
Access Control in RAG Chat Apps When you're building an AI chat app for internal documents, you need to think about access control: which of your users can access which documents? This session shows a sophisticated approach to access control that stores documents in a secure ACL'd datastore, remembers ACLs in Azure AI search, and only sends user-visible documents to OpenAI. This session also shows you how to set up authentication for your app and how to block unwanted access. |
|
Chat Completion API Tools & Functions in RAG Chat Apps This session shows off a relatively new way to retrieve knowledge and direct a conversational flow, using the Azure OpenAI function calling feature. This session will give you the skills to extend your AI chat to cover more use cases, like structured queries and user question pre-processing. |
|
This session shows you how to use various GPT-based metrics and tools to evaluate a RAG chat app. We try out changes like different prompts, various search parameters, and alternate LLM options, to see how that affects the overall metrics. When building a RAG chat app, you absolutely need an evaluation step in your pipeline to build confidence that your chat app will provide helpful, grounded, and relevant answers for your users. |
|
Continuous Deployment of your Chat App Do you want to learn how to deploy your chat app to Azure without any hassle? Join this live session with an Azure Developer CLI software engineer and discover the best practices for continuous deployment of your chat app. We'll focus on GitHub actions, but you can use the same approach for Azure DevOps or other CIs. |
|
Content Safety for Azure OpenAI A special session from a Microsoft MVP on Responsible AI and the Azure OpenAI content safety filters. Learn how to change the filter levels and use them in your OpenAI chat apps. |
|
Building a Chat on Your Business Data Without Writing a Line of Code A special session from a Microsoft MVP on the Azure OpenAI Studio "On Your Data" feature, a low-code approach to developing a RAG chat app. |
|
The final session, showcasing many submissions and demos from Microsoft's AI-In-a-Box team. |
Spanish streams
Las dos primeras sesiones, presentadas por Bruno Capuano.
|
Creación de una aplicación de chat RAG en Python ¿Quieres aprender a crear una aplicación de chat con tus propios datos usando Python? Esta sesión le muestra cómo utilizar RAG (Generación Aumentada de Recuperación), una técnica poderosa para combinar la recuperación de conocimientos con LLM. Aprenderá cómo combinar OpenAI con la búsqueda de Azure AI y Azure Document Intelligence para ingerir y vectorizar datos, y luego implementar una interfaz para chatear con esos datos en Azure App Service. |
|
Personalización de la aplicación RAG + Chat ¿Listo para personalizar su aplicación RAG para sus propios datos y organización? Esta sesión le muestra cómo traer sus propios documentos y sitios web. También incluye consejos sobre cómo personalizar la interfaz para agregar su propia marca e ideas para funciones adicionales que quizás desee incorporar a su aplicación de chat. |
Portuguese streams
As duas primeiras sessões, apresentadas por Pablo Lopes.
|
Criando um aplicativo RAG Chat em Python Quer aprender como criar um aplicativo de bate-papo com seus próprios dados usando Python? Esta sessão mostra como usar RAG (Retrieval Augmented Generation), uma técnica poderosa para combinar recuperação de conhecimento com LLMs. Você aprenderá como combinar o OpenAI com a pesquisa de IA do Azure e o Azure Document Intelligence para ingerir e vetorizar dados e, em seguida, implantar um front-end para conversar com esses dados no Azure App Service. |
|
Personalização do aplicativo RAG + Chat Pronto para personalizar seu aplicativo RAG para seus próprios dados e organização? Esta sessão mostra como trazer seus próprios documentos e sites. Ele também inclui dicas sobre como personalizar o front-end para adicionar sua própria marca e ideias para recursos adicionais que você pode querer trazer para seu aplicativo de bate-papo. |
Chinese streams
|
我们将首先来学习如何使用 Python，以及 Azure OpenAI Service 构建我们的第一个 RAG 应用，入门生成式人工智能。 |
|
这一期，我们将结合 Embedding，学习如何定制化我们的 RAG 应用。 |