Building AI Applications with Azure Cosmos DB JavaScript SDK
summary
A live support chatbot for CityJS uses Azure Cosmos DB JavaScript SDK as a vector store, stores vectors and source text together in one document, and uses Azure OpenAI embeddings (ADA) plus GPT-4 for generation. LangChain and llamaIndex integrations provide a retriever that returns the top 3 documents, the app ingests a one‑page conference PDF to build embeddings, and the demo answers questions (lunch at 12:40) while the SDK reports ~500k weekly downloads and supports flat/quantized/disk ANN indexing.
AI-written summary · about
chapters (14)
- 0:00Introduction to demo
- 2:36Building AI applications overview
- 3:40Azure Cosmos DB NoSQL features
- 4:42Cosmos DB as vector store
- 6:29Use cases for Cosmos DB
- 7:33RAG pattern explained
- 8:37JavaScript SDK introduction
- 9:39AI frameworks for JavaScript
- 11:24Connecting components with RAG
- 12:27Creating vector store from PDF
- 13:34Full RAG workflow
- 14:59Code walkthrough overview
- 16:36Chat API implementation
- 19:24Q&A and closing
related talks
12mBuilding Generative AI Apps with Azure Cosmos DB
50mHow to design and build AI applications with vector search using Azure OpenAI & Azure Cosmos DB
19mCosmic Database Powering OpenAI’s ChatGPT
03h 12mAzure Cosmos DB Conf 2024
17mAzure Cosmos DB in AI Workflows: From Knowledge Retrieval to Microsoft Fabric Integrations
15mAzure Cosmos DB as Vector Database
01h 16mAzure SQL ❤️ Open AI
37mBuilding an Enterprise Knowledge Management System using Hybrid Search in Azure Cosmos DB