Design and implement agent memory stores in Azure Cosmos DB

Module
9 Units

Intermediate

Developer

AI Engineer

Azure Cosmos DB

Azure OpenAI Service

Give an agent something to remember. Separate short-term conversation state from long-term memory, distill raw turns into facts a later session can find by meaning, and set the retention, reconciliation, and erasure policies that keep a memory store trustworthy as it grows.

Learning objectives

After completing this module, you're able to:

Design short-term and long-term memory architectures for agent use cases.
Implement conversation state in Azure Cosmos DB for NoSQL and expire it with time to live.
Implement semantic memory by distilling turns into embedded, individually correctable facts.
Retrieve the right memory into a limited context window and inject it safely.
Design memory retention, pruning, consolidation, and erasure policies.

Prerequisites

Experience running vector and full-text queries in Azure Cosmos DB for NoSQL
Familiarity with vector embeddings and calling an embedding API such as Azure OpenAI
Working knowledge of Azure Cosmos DB for NoSQL partitioning, queries, and request units
Programming experience with Python

Introduction min
Understand agent memory architectures min
Implement conversation state min
Implement semantic memory with vector storage min
Optimize memory retrieval and context injection min
Design memory retention, pruning, and consolidation min
Exercise: Build a durable agent memory store min
Module assessment min
Summary min

Start