How to design and build AI applications with vector search using Azure OpenAI & Azure Cosmos DB
summary
Shows a working RAG pipeline using Azure OpenAI for embeddings and chat completions plus Cosmos DB VOR as the vector store, with a Java demo that loads three documents, indexes them, and queries by embedding similarity. Demonstrates 7s→580ms semantic caching speedup, support for HNSW and DISN indexes (DISN via Cassandra offering), LangChain integration, and a Cosmos DB co-pilot that translates natural language to queries. Announces vector indexing coming to the NoSQL API and emphasizes prompt engineering, retrieval tuning and trade-offs between accuracy, latency and cost.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 8:11Defining AI and LLMs
- 10:20Modern AI challenges
- 11:46Key concepts for AI apps
- 16:03Vector embeddings and search
- 20:37Azure Cosmos DB for MongoDB vCore
- 24:31AI apps with RAG pattern
- 26:37Demo: Simple AI app
- 33:01Demo: Complex AI app with LangChain
- 36:35Query Copilot for Cosmos DB
- 38:47Demo: Semantic caching
- 42:39Resources and closing
- 43:21Q&A
related talks
12mBuilding Generative AI Apps with Azure Cosmos DB
15mAzure Cosmos DB as Vector Database
03h 12mAzure Cosmos DB Conf 2024
11mScalable RAG with Azure Cosmos DB and DiskANN
01h 16mAzure SQL ❤️ Open AI
40mUnleashing the Power of Vector Search with Azure SQL Database
17mAzure Cosmos DB in AI Workflows: From Knowledge Retrieval to Microsoft Fabric Integrations
48mThe power of AI and Copilot for Azure Databases