Select your cookie preferences

We use essential cookies and similar tools that are necessary to provide our site and services. We use performance cookies to collect anonymous statistics, so we can understand how customers use our site and make improvements. Essential cookies cannot be deactivated, but you can choose “Customize” or “Decline” to decline performance cookies.

If you agree, AWS and approved third parties will also use cookies to provide useful site features, remember your preferences, and display relevant content, including relevant advertising. To accept or decline all non-essential cookies, choose “Accept” or “Decline.” To make more detailed choices, choose “Customize.”

AWS Logo
Menu
Supercharging Amazon Bedrock Flows with AWS Multi-Agent Orchestrator

Supercharging Amazon Bedrock Flows with AWS Multi-Agent Orchestrator

Amazon Bedrock Flows is an intuitive visual design tool that enables users to create complex AI workflows through a user-friendly drag-and-drop interface. By seamlessly integrating with AWS Multi-Agent Orchestrator, it simplifies the coordination of AI agents, making sophisticated AI solutions more accessible to organizations of all technical levels.

genaimulti-agent-orchestratorbedrockbedrock-flows
Anthony Bernabeu
Amazon Employee
Published Dec 3, 2024
Comments
I've been working with Amazon Bedrock Flows lately - it's this really cool feature from Bedrock that lets you visually design your AI workflows. If you're like me and prefer visuals over code, you'll love it! Drag, drop, connect nodes, and boom - you've got an AI flow running.
But after using it for a while, I hit some limitations that Bedrock Flows alone couldn't solve.
Two big ones stood out:
  1. No memory across conversations (each request starts fresh)
  2. No built-in way to manage multiple flows
Let me show you how I solved these using the Multi-Agent Orchestrator framework.

The Memory Problem 🧠

As of December 3, 2024, Amazon Bedrock Flow doesn't have built-in memory support. That means if you're building a conversational AI, your flow can't remember what was said earlier in the conversation - not ideal!
But here's where the Bedrock Flows Agent comes to the rescue. It adds chat history support to your flows, making them actually conversational. Pretty neat, right?

Case 1: Single Flow Agent 🎯

For our example we have created a tech agent flow as followed:
tech-flow
The tech_agent prompt node will extract history and question from the Flow input.
Here is the prompt we used for this tech_agent node:
In this example we are not going to use a classifier since we only have one agent and we don’t need to use the full orchestrator logic.
Here is a diagram that shows how everything connects together within the multi-agent-orchestrator:
Our tech agent flow, enhanced by chat history context, accurately recognized that the question was related to AWS Lambda. This allowed it to provide the correct response to our follow-up query.
Pretty impressive, right? 🤩

Case 2: Multiple Flows with Orchestrator 🔄

Let’s create another Bedrock Flows, for health request:
Similar to the tech_agent flow, question and history are extracted from the Flow input node.
Here is the prompt we defined for this health_agent node:
When you have multiple flows, you'll want to use the orchestrator to route requests to the right flow:

The Result? 🎉

Now you have:
  • Flows that integrates conversation history
  • Smart routing between different flows
  • A clean way to mix flows with other agent types

Wrap Up

By using the Multi-Agent Orchestrator framework, we've basically given our Bedrock Flows superpowers! The flows can now remember conversations and work together seamlessly. No more awkward "sorry, what were we talking about?" moments or confusion about which flow should handle what.
Give it a try and let me know how it works for you! 🚀

Wants to know more?

Head out to our documentation with practical examples:

Any opinions in this post are those of the individual author and may not reflect the opinions of AWS.

Comments

Comments