RAG with Azure OpenAI (AI Search, CosmosDB, GraphRAG & more)

Explore the evolution from artificial intelligence to generative ai and large language models, powered by transformer architecture, and learn how foundation models become fine-tuned applications like Copilot and ChatGPT.

Learn the basics of generative AI jargons—tokens, system prompts, and user prompts—and how token counts and the chat completions API drive cost and latency.

Explore how Azure OpenAI provides private tenant isolation, secure endpoints, and integrated model deployments (GPT, text embedding) via the Azure OpenAI studio, differentiating it from OpenAI's independent platform.

Explore the what, why, and how of retrieval augmented generation, grounding enterprise chatbots in private data via three steps: retrieval, augmentation, and generation.

Explore the Azure OpenAI horizon for RAG systems, from pro code deployments to Copilot Studio, and learn how Azure Search and Cosmos DB store documents and vector embeddings.

Compare Azure AI Studio and Microsoft Copilot Studio to guide citizen developers toward low-code copilot creation and pro-code production, using orchestration, SDK integration, and model catalogs.

Deploy an Azure OpenAI resource in portal.azure.com, create a Sweden Central resource group, and configure primary and secondary keys and endpoints for bearer token API calls.

Explore deploying GPT 35 turbo 16 K in the Azure AI Foundry portal and chat in the chat playground, using system prompts and prompt engineering to shape responses.

Explore the image generation model in Azure OpenAI with a hands-on lab, deploying Dall-E 3, setting content safety, and testing prompts to generate varied images.

Explain the recent ui rebranding from Azure OpenAI Studio to the Azure AI Foundry portal, and confirm that model deployment and the chat playground remain largely unchanged.

Learn to wire Python code to the Azure OpenAI GPT engine using the chat completions API, instantiate an Azure OpenAI client, and call client.chat.completions.create with system and user prompts.

Learn to generate text embeddings in Python using Azure OpenAI's text embedding model, configure environment variables, and obtain 1536-dimensional embeddings for rag architecture.

Master prompt engineering by crafting precise prompts to configure intelligent agents and guide outcomes, focusing on the four parts—goal, context, expectations, and source—for reliable results.

Explore prompt engineering techniques such as chain-of-thought prompting, zero-shot prompting, and few-shot prompting, plus best practices for configuring AI agents, structuring outputs, and breaking tasks into steps.

Learn to implement retrieval augmented generation with Azure AI Search, using vector embeddings and hybrid search across sources like one lake storage, blob storage, and Cosmos DB to answer queries.

Learn to build a hybrid search with vector and keyword queries using Azure AI Search and OpenAI embedding models, indexing PDF reviews from blob storage and enabling citation-backed Q&A.

Clone Rag with Azure OpenAI repo in VS Code, configure environment variables, and run Python notebook to generate vector embeddings and retrieve three documents with cosine similarity for GPT summarization.

Build a multimodal rag pipeline with Azure AI Search to answer audit questions using PDFs, images, and tables while preserving layout and citations.

Build a simple multimodal rag pipeline for image verbalization in Azure, wiring up blob storage, Azure OpenAI, and Azure Search to index images and generate embeddings.

Build an end-to-end rag chatbot with Azure OpenAI, Azure search, and embeddings to retrieve and cite content from PDFs and images, guided by a lab workflow that emphasizes responsible AI.

Explore advanced multimodal rag for structured documents, preserving layout with the document intelligence layout API, and build an Azure search index using a multi-service account and invoices.

Build an advanced multimodal document intelligence RAG pipeline in a VS Code lab, configure environment variables, run a Jupyter notebook, and generate invoice citations with bounding boxes.

Learn how Azure AI content understanding extracts and labels unstructured data—from PDFs, images, audio, and video—via content and field extraction analyzers, to power RAG pipelines with vector databases.

Compare pro field extraction pricing with standard field extraction on Azure AI content understanding, and explain how large language model reasoning, verification, and redo cycles affect token costs.

Explore azure ai content understanding through a graphical user interface workflow, sample analyses, and exercises for document, invoice, audio, and video analysis, with prebuilt and custom analyzers and endpoint usage.

Practice Azure AI content understanding in a code-first lab using Python notebooks to call the API, set endpoint and key, and run prebuilt analyzers for documents, images, audio, and video.

Create a custom analyzer in Azure AI content understanding, define the schema and fields, test with invoices, then run analysis and call it via a Python notebook.

Set up Azure storage and blob storage, create Azure AI vector index, upload MP3, MP4, PNG, PDF, and deploy GPT-4 chat and text embedding 002 in the AI Foundry project.

Set up environment variables in this hands-on lab and prepare Azure blob data to build multimodal data sets and index them in Azure search for rag with Azure OpenAI.

Execute a hands-on rag pipeline that retrieves relevant content from Azure AI Search via vector embeddings and similarity search, then generate a response with an LLM.

Explore how Azure Cosmos DB supports RAG by providing globally distributed, low-latency NoSQL storage with multi-api support, vector search, and scalable throughput.

Explore the RAG architecture for Azure Cosmos DB for NoSQL API, storing JSON documents with dish details, generating and indexing vector embeddings with DiskANN, and using cosine similarity for retrieval.

Engage in a hands-on lab that builds a RAG workflow with Azure Cosmos DB for NoSQL API, vector embeddings via Azure OpenAI, and a GPT-based summary.

Architect RAG with Azure OpenAI and Cosmos DB MongoDB core by storing documents and vector embeddings, and use LangChain to vectorize queries and retrieve results for generation.

Execute a hands-on lab to build a RAG pipeline with LangChain on Azure Cosmos DB MongoDB vcore, using 1536-d embeddings, IVF vector indices, and cosine similarity.

Explore a real-life vector embeddings use case beyond Rag, demonstrating a dockerized Azure app that semantically maps queries to images stored in Cosmos DB, with similarity scoring.

Deploy Azure resources to enable vector search using GPT-4 vision and text embeddings with Azure OpenAI and Cosmos DB. Configure storage for icons and enable anonymous access.

Clone azure icon search repo and run notebook. Convert svg icons to png and generate vector embeddings with GPT-4 vision, then store results in Cosmos DB for top-5 retrieval.