LangChain on Azure - Building Scalable LLM Applications

Install docker and the docker CLI, choose the correct docker desktop version for Mac, Windows, or Linux, and use docker ps to verify the daemon and available images.

Learn to install the Azure CLI across Windows, macOS, and Linux, choosing 32- and 64-bit installers, using PowerShell or Homebrew, and verify with az --version.

Install Visual Studio Code, the recommended IDE for this course, to access Jupyter Notebook support and syntax highlighting for Python, Docker, JavaScript, HTML, shell, and script.

Create and link a new Azure resource group to a subscription, name it, and choose a region—understanding data residency, latency, and cost implications before review and creation.

Create an Azure Cognitive Search service, explore its vector database capabilities, and use the ACS SDK to insert and query data. Then build an index with Python in Jupyter notebooks.

Set up a Python virtual environment, install notebooks' dependencies, and connect a Jupyter kernel to Azure cognitive search using .env endpoint and key.

Create and populate a restaurant index in Azure Cognitive Search using the Python SDK, defining simple and complex fields, uploading documents, and performing searches with LangChain's vector retrieval.

Learn to integrate Azure cognitive search with LangChain by vectorizing documents with OpenAI embeddings and using a retrieval augmented generation chain to answer opening hours.

Learn to set up Azure blob storage for unstructured data, create storage accounts and containers, and apply hot, cool, and archive tiering with lifecycle policies.

Learn to manage Azure blob storage with the Python SDK, upload and delete blobs, and build a Lang Chain-driven vector store with Azure cognitive search for scalable LLM apps.

Configure Azure Database for Postgres flexible server and install the PG vector extension to create a vector store for scalable LLM applications.

Index text with pkg vector in LangChain on Azure by configuring environment variables, connecting to a Postgres vector store, and using the record manager with incremental, full, or none cleanup.

Connects peg vector with the indexing API and Azure blob storage to load, split, and index documents, adjust metadata to base file names, and manage document changes.

learn to use pgvector as a retriever with the Lang Chain Expression Language, turning a vector store into a retriever and building a context-aware retrieval QA pipeline.

Set up a docker-compose multi-service app with Postgres, back end, front end, and upload services; configure volumes, environment vars, and endpoints for a vector store and blob storage.

Walk through the angular frontend and its docker multi-stage build. See how a FastAPI proxy forwards requests to the backend, with environment configs for production hosts and ports.

Walk through the back end code of a LangChain on Azure app, setting up blob storage, Postgres vector store, and OpenAI embeddings, with a Q&A chain.

Create an Azure container registry, build Docker images, push them to the registry, and configure an admin user and access keys so the app service can pull images securely.

Build and push docker images for back end, front end, and upload service to Azure container registry using the provided script, then configure app services to use these private images.

Deploy dockerized front-end apps on Azure App Services for simple, scalable web deployments. Configure websites_port to 3000 so the front end loads on port 80 via Azure.

Deploy the backend services on Azure app service and configure docker single container. Set environment variables for API key, PG vector, and blob storage, then monitor with uvicorn logs.

Deploy and configure an Azure upload service and front-end, set environment variables and proxies, push docker images, connect blob storage, and trigger an Azure function to update the vector store.