This blog post is a public copy of our monthly newsletter. If you’d like to receive upcoming updates by email, you can subscribe to our newsletter here.
Highlights from June include new AI capabilities, major releases, community-built projects, and practical guidance from teams running Temporal at scale. You’ll also find answers to real questions from the field, a new hands-on Python tutorial, and opportunities to join us live.
To learn more about these updates and everything else we’ve been working on, keep reading. And as always, we’d love to hear from you. Feel free to share feedback in our Community Slack.
Recently shipped#
-
The AWS Strands SDK Integration is now in Public Preview! Temporal now integrates with AWS Strands Agents, the open-source SDK from AWS for building model-driven AI agents. This integration adds Temporal’s Durable Execution to Strands-based agents.
-
The new Cloud CLI extension is in pre-release! Instead of juggling separate CLIs for self-hosted (
temporal) and Cloud (tcld), you can now use a single CLI, providing a more streamlined experience for working with your Temporal Workflows and managing your Cloud environment. -
Stable IPs for Namespace Endpoints are now generally available. This optional setting on a public Connectivity Rule makes a Namespace Endpoint resolve to a published, fixed set of IP addresses.
-
Custom Roles are now in pre-release. With Custom Roles, you can define granular permissions in Temporal Cloud, giving your team precise control over who can perform specific actions within your account.
-
Temporal OSS v1.31 has been released! Includes support for Serverless Workers (Pre-Release), Standalone Activities (Public Preview), Worker Versioning (GA), Task Queue Priority & Fairness (GA), and more.
Join us live#
Tune in to Vibe Check, our livestream series exploring what it means to build with AI today. We’ll explore the technical details, architectural tradeoffs, and evolving developer experience shaping modern AI systems. Or join us in person at upcoming events around the world:
Builder spotlight#
This month, we’re featuring Lightspeed, a full-featured agent harness designed from the ground up for Temporal. Builder Lukas Buehler has been working on agents for the last few years, and recently set out to build a more robust foundation for them. “But after a few months of building, I realized that I was engineering both a workflow engine and the entire agent harness. So I pivoted and decided to focus entirely on the agent itself.” Check it out!
Dip Patel, a community member, has written a series of posts about how Rapidflare uses Temporal to process hundreds of thousands of documents across 20+ source types. Part 1 covers architectural patterns such as sliding-window fan-out with Signals, Cloud Storage as a data bus for oversized payloads, and page-based Continue-as-New for chunked PDFs. Part 2 dives into operating at that scale, including milestone Heartbeats, checkpoint-based cancellation, and a subtle race condition. These posts include real production anecdotes and concrete implementation patterns to help you scale your own pipelines.
How to Temporal#
Our new Build a Job Queue with Standalone Activities Python tutorial is live!
Learn how to build a durable job queue with Temporal’s new Standalone Activities. You’ll cover:
-
Build durable job queues with automatic retries and recovery
-
Use idempotency keys to make retries crash-safe
-
Deduplicate work using Activity ID reuse
-
Apply concurrency controls and rate limiting for reliable processing
-
Implement heartbeats and checkpointing for long-running jobs
-
Reuse the same Activity inside Workflows as your application grows
And we have a new hands-on coding experience! Give it a try and share your feedback in #python-sdk.
Questions from the field#
Kunal Verma asked in #python-sdk how to address a flood of gRPC cancellation warnings that were causing Activities to wait out their full Start-to-Close timeout before retrying. Community member Saad Mohsin Khan gave a thorough breakdown explaining that the Server only retries an Activity after its attempt has been marked as failed, and recommended shorter timeouts combined with heartbeating to fail faster. Conor McCullough from Temporal then identified the root cause: a bug introduced in SDK version 1.27.0, fixed in 1.27.2. Upgrading to 1.27.2 resolved the issue.
Simon Novak shared a message in #dotnet-sdk about an OverflowException crashing his Workflow after the history grew past 2 GB. Maciej Dudkowski from Temporal explained that the Workflow.CurrentHistorySize property is exposed as a signed int, making an overflow inevitable at that scale, and pointed to Continue-as-New as the standard remedy. Simon confirmed that he mitigated the problem by splitting the long-running Workflow into Child Workflows.