Announcing Long-Running Execution Flow Support in Amazon Bedrock Flows

Transform Your Generative AI Workflows with Asynchronous Capabilities

Solution Overview

Key Benefits of Long-Running Execution Flows

Real-World Application: Dentsu’s Easy Reading Project

Creating a Long-Running Execution Flow

Testing Your Workflow

Conclusion and Next Steps

About the Authors

Unlocking New Possibilities with Amazon Bedrock’s Long-Running Execution Flows

In the dynamic world of generative AI, the demand for efficient workflow management has surged. Today, we are excited to announce the public preview of long-running execution (asynchronous) flow support within Amazon Bedrock Flows. This groundbreaking feature allows businesses to link various Amazon Bedrock components—like foundation models, prompt management, knowledge bases, and guardrails—to build and scale predefined generative AI workflows.

Addressing Industry Needs

As organizations across sectors strive to create more intricate applications, many have voiced the need for processing larger datasets and executing complex workflows that exceed a typical runtime. Customers want to transform entire books, manage extensive documents, and orchestrate sophisticated workflows without stressing over time limits. To meet these needs, we’re extending the workflow execution time from just 5 minutes (synchronous) to a remarkable 24 hours (asynchronous).

The Benefits of Long-Running Execution Flows

With the introduction of Amazon Bedrock long-running execution flows, businesses can now:

1. Run Long Workflows: Execute workflows seamlessly in the background for up to 24 hours, decoupling the execution process from immediate user interaction.

2. Handle Large Payloads: Process large and resource-intensive tasks over a day instead of the previous 5-minute constraint.

3. Manage Complex Use Cases: Orchestrate intricate, multi-step decision-making workflows that integrate with various external systems.

4. Enhance Builder Experience: Create and manage workflows effortlessly through both the Amazon Bedrock API and console.

5. Gain Observability: Enjoy a smoother user experience by tracking flow execution status and accessing detailed traces for each workflow node.

A Real-World Example: Dentsu’s Easy Reading Application

To illustrate the application of these new features, consider Dentsu, a leading advertising agency. They needed to handle complex multi-step generative AI processes that required extended execution times. One notable project, the Easy Reading application, transforms books into more readable formats tailored for individuals with intellectual disabilities.

With the new long-running execution flows, Dentsu can now:

• Process expansive inputs and complex tasks that were previously constrained by the 5-minute limit.
• Integrate a variety of external services into their AI workflows.
• Support both quick and long-running tasks efficiently.

Victoria Aiello, Innovation Director at Dentsu Creative Brazil, commented, “Amazon Bedrock has been amazing to work with and demonstrates value to our clients. The long-running execution flows allow us to process an entire book at once, streamlining our workflow and saving significant time.”

Visualizing the Workflow

Dentsu’s efficient book processing application begins when a client uploads a book to Amazon S3. This upload triggers a flow that processes multiple chapters, applying necessary accessibility transformations and formatting. This long-running execution flow not only ensures efficient processing but also provides continuous status updates throughout the transformation.

Creating Your Long-Running Execution Flow

To get started with implementing these features, follow these prerequisites:

• Ensure you have access to Amazon Bedrock services.
• Familiarize yourself with the Amazon Bedrock console and API.

Step-by-Step Guide

1. Navigate to Amazon Bedrock, and under Builder tools, select Flows.
2. Click Create a flow and input a name—e.g., easy-read-long-running-flow.
3. Utilize various nodes to build your workflow, customizing it according to your application requirements.

Testing Your Flow

You can test your new flow using a fictional book, such as “Beyond Earth: Humanity’s Journey to the Stars.” Input the appropriate chapter prefixes and relevant metadata to initiate the long-running flow.

Amazon Bedrock Flows will automatically manage the execution while providing vital performance metrics and execution traces, which can be viewed in the console or sent to Amazon CloudWatch for better observability.

Conclusion

Amazon Bedrock’s long-running execution flows are a game-changer for developers and organizations looking to harness the full potential of generative AI. By allowing complex workflows to run efficiently over extended periods, we’re addressing critical challenges in the ever-evolving landscape of AI application development.

Join us today in exploring these exciting features within the Amazon Bedrock console or APIs. We look forward to seeing the innovative applications you will build with these new capabilities. Please share your experiences with us through AWS re:Post or connect with the generative AI builder community at community.aws.

About the Authors

The blog is authored by a team of seasoned AWS professionals, each specializing in various aspects of generative AI and application development. Their collective expertise spans engineering, architecture, and product management, ensuring that you receive the most accurate and valuable information in the field.

Embrace the future of AI workflows with Amazon Bedrock Flows—where ambition meets possibility!