Two Millennia of Compute Later
Last month, we orchestrated two millennia of compute. Our users are running dbt projects with 1,000+ models and routing flows to fleets of GPUs.
When you scale to that level, you shouldn't have to build custom tooling just to keep the lights on. We've spent the quarter turning some of the most effective patterns we've seen from high-scale teams into first-class features.
Infrastructure Debugging
Debug Kubernetes and ECS failures without leaving the Prefect UI.
A bad image, a pod eviction, an OOM: kubectl can stay closed. We hardened our workers to surface infrastructure failures directly in your flow run logs, and added a real-time Infrastructure Drawer with CPU and memory metrics so you can see why a run failed. No external cloud logs required.
Read the docs →Managed Automations
Automatically clean up runs stuck in "Running" or "Cancelling."
Infrastructure breaks: a container fails to bootstrap, a teardown hangs, a worker dies mid-flight, and you end up with a zombie. Opt in to system defaults that detect those states and move them to terminal states on the thresholds you set, so you can stop writing cleanup scripts.
Read the docs →Config Plugins
Ship your org's Prefect settings as a pip-installable library.
Package auth, credentials, and compliance logging into a plugin, and a new author gets a correctly wired environment from a single pip install on day one, with no setup wiki to chase. One install. Day-one ready.
Read the docs →Rate Limit Controls
Visible usage dashboards and workspace-level limit allocation.
We simplified rate limits into two buckets: Orchestration and Events. We cranked the limits up (up to 4x for logs) and launched a dashboard so you can see which deployment or API key is burning through your quota before you hit a 429.
Read the docs →CLI Startup Performance
Snappier interactions for developers and AI agents.
We overhauled the CLI with lazy loading and deferred imports, cutting overhead across every command. The 15% drop in startup time makes the developer loop feel much snappier and provides a significant efficiency gain for AI agents calling Prefect in a loop.
Now in Open Beta
Default Result Storage
Set your S3 or GCS bucket once at the account level.
Tired of configuring storage on every single flow? Account-level defaults propagate down to every workspace automatically, so new flows inherit the right bucket without manual setup.
Read the docs →Assets Explorer
A list-first catalog for materialization history and impact analysis.
The asset graph shows the big picture; history is harder. The new Catalog is searchable, and Blast Radius Analysis surfaces every downstream flow that depends on an asset before you triage a failure.
Quickstart
- In Prefect Cloud, open the Assets page from the left sidebar.
- The page now opens to the Catalog tab by default. The graph view is still available via the toggle at the top.
- Use the search bar to find an asset by name or path.
- Click into any asset to open its detail page.
Cross-Team Assets
Connect workflows across workspaces without sharing access.
Team A can publish an asset that Team B subscribes to. Team B sees the status and history and can trigger their own workflows from the events. Without ever touching Team A's workspace.
Quickstart
Cross-Team Assets work in two steps: publish in the producing workspace, then subscribe in each consuming workspace.
In the producing workspace:
- Open the asset's detail page from the Assets Catalog.
- Click the Publish toggle in the top right.
- The asset is now available for subscription from any other workspace in the same account.
You need permission to update assets in the producing workspace to publish.
In the consuming workspace:
- Open Assets and switch to the External tab.
- Find the published asset in the list. Assets available for subscription appear here automatically.
- Click the asset, then click Subscribe in the top right.
Once subscribed, the asset's full history and ongoing events appear in the consuming workspace.
Workspace Personalization
Pin flows, deployments, and creators to filter out the noise.
In a shared workspace, yours are the only flows that matter. You can now pin resources to your sidebar, see profile photos of who last edited a resource, and filter tables to just what you own. These changes are visible in the UI.
Getting started
If you have suggestions or requests for what you'd like to see, check out our GitHub discussions. Our Community Slack is also a great place to connect with others and get help with your workflows.