Design patterns for scalable multi-agent applications with Azure Cosmos DB

In this session, we'll explore design patterns for developing scalable multi-agent applications. We'll focus on leveraging Azure Cosmos DB as the single database to manage agent transactional data, operational application data, vector embeddings, text data, and more! We'll see why Azure Cosmos DB is the database of choice for any-scale AI apps.

Key topics include:

Agent-to-Agent Transactions: Discover methods to facilitate seamless and efficient transactions between agents, ensuring data consistency and reliability.

Statefulness: Learn how to maintain state across interactions, enabling agents to remember and build upon previous exchanges.

Chat History: Explore techniques for storing and retrieving chat histories, allowing agents to provide contextually relevant responses.

Retrieval-Augmented Generation (RAG): Understand how to implement patterns like vector Search with DiskANN for low-latency, highly scalable vector search, and Hybrid Search to enhance the relevancy of agent contexts and responses.

Function Calling: Learn how to enable structured queries from Cosmos DB to expand scenarios beyond simple RAG approaches.

Multitenancy: See how to support multitenant applications, ensuring data isolation and security for different users or clients.

Scalability: Examine features that handle large volumes of data and high-throughput workloads with Azure Cosmos DB's Dynamic Autoscale, which is essential for multi-agent systems.

Build Multi-Agent Applications in Python Using LangGraph & Azure Cosmos DB

Multi-agent systems represent the cutting edge of dynamic and scalable applications, enabling autonomous collaboration between specialized agents to solve complex problems. In this hands-on workshop, you’ll learn how to harness the power of LangGraph in Python to create a robust, AI-driven system.

This workshop is designed to guide developers through every step of building durable multi-agent applications, including:

• Understanding Multi-Agent Architectures: Explore the core principles of multi-agent systems, their benefits, and how they can simulate human-like collaboration.
• Agent Specialization and Orchestration: Learn how to define agent roles and tasks to enable seamless collaboration.
• Knowledge and Memory: Implement long-term memory using Azure Cosmos DB with vector search and agent state persistence to ensure durability and context-awareness in agent interactions.
• Monitor: Learn how to measure and evaluate multi-agent applications to ensure correctness and efficiency.

By the end of the workshop, participants will have built a fully functional, multi-agent application, equipped with intelligent, task-oriented agents capable of collaborating autonomously. Regardless of your current skills with agents, this workshop will take your AI development skills to the next level.

There are pre-requisites for this workshop. Please see, https://aka.ms/agent-workshop-python-prereqs for more details.

CORENOTE: Building Agentic apps in Fabric with Cosmos DB and User Data Functions

Build agentic apps faster with Microsoft Fabric as your intelligent backend. Add persistent memory with Cosmos DB and fast semantic search over OneLake.

Create real-time, context-aware experiences with User Data Functions, seamless Cosmos DB–OneLake mirroring, and Real-Time Intelligence.

How to model and partition data in a distributed NoSQL database to achieve cloud scale

The era of cloud computing has ushered in a new era where applications now demand unprecedented levels of scale and size. Relational databases which have been with us for 50 years were never designed to ingest vast quantities of data per second, grow to Petabytes in size, or provide 99.999% availability. Enter distributed NoSQL databases which were designed specifically to deal with these and other unique challenges brought about by the cloud.

In this session we will take the relational data model for a simple e-commerce application and migrate it to a horizontally partitioned, NoSQL database. Along the way you will learn the concepts, techniques and technologies needed to model for a horizontally partitioned NoSQL database that will allow your application to achieve millisecond response times with near unlimited size and scalability.

Build high performance AI apps using Azure OpenAI & Azure Cosmos DB

Azure Cosmos DB is a distributed NoSQL database providing unmatched availability, performance and scalability and now offers vector database capabilities developed by Microsoft Research for building AI applications with industry-leading performance and scale.

In this session we’ll dive straight into vector embeddings, and what makes them different and so powerful for search. We will show how Azure Cosmos DB enables you to store your data and vectors together and perform blazing fast vector search at enormous scale, using native SDK's as well as Semantic Kernel and LangChain in both C# and Python. We will then tie all this together with more demos showing how Azure Cosmos DB can be leveraged for RAG, chat history, and semantic caching to elevate your generative AI applications to the next level. If you want to learn how to build the next generation of AI-enabled applications, attend this session to accelerate your AI readiness for the future!

Targeting developers. Features available at Microsoft BUILD in May 2024

Architecting highly resilient applications in Azure: A technical guide for developers

Dive into the technical strategies and architectural patterns essential for designing applications that achieve zero downtime, zero data loss, and seamless scalability on Azure.

This session will provide developers with hands-on guidance to implement key resilience patterns such as Active-Active replication, Saga transactions, and Global Secondary Indexing. Learn how these patterns are operationalized in Azure Kubernetes Service (AKS), Azure Cosmos DB, and other critical Azure services. Gain actionable insights on automating deployments, monitoring distributed systems, and optimizing for performance.

By the end of this session, developers will have the tools to architect and build robust, fault-tolerant systems tailored to business-critical demands.

The audience for this talk are architects and developers.