Amazon Bedrock Agents enable developers to build conversational applications beyond single-turn interactions. With the newly introduced memory feature, agents can retain context across sessions, summarize past conversations, and deliver more natural and personalized responses, without requiring developers to manually manage conversation history.
In this Cloud Lab, you’ll build a complete multi-user conversational application using Amazon Bedrock Agents with session memory maintenance. The goal is to demonstrate how Bedrock Agents can recall user goals, assist actions across multiple sessions, and how this integrates into a real-world web application.
You’ll start by creating an Amazon DynamoDB table to manage user accounts, including login credentials and session tracking for each user. This ensures every user who signs in receives a unique session ID that ties directly to their conversation history.
Next, you’ll create a Bedrock Agent and enable the memory feature with session summarization. This ensures that every user interaction is tied to their session ID, allowing the Agent to recall prior conversations and generate context-aware responses.
Finally, you’ll integrate the agent into a Flask-based web application that allows users to sign up, log in, and log out securely. Once authenticated, users can interact with the Bedrock Agent through a chat interface. Each query is routed with the correct session ID, allowing the agent to retrieve stored summaries and continue the conversation naturally. Users can log out and return later, with conversations picking up where they left off.
By the end of this Cloud Lab, you will have created a fully functional multi-user conversational AI application powered by Amazon Bedrock, DynamoDB, and Flask, that can maintain session memory across conversations.
The following is the high-level architecture diagram of the infrastructure you’ll create in this Cloud Lab:
