prefect

It's only been a little over a week since the last release, but there are a lot of new features we're excited to announce!

This release includes:

• Cancellation of flow runs
• Pause and resume of flow runs
• Logging of print statements
• Agent-level flow run concurrency limits
• Notification blocks for PagerDuty and Twilio
• Improved reporting of crashed infrastructure
• Improved handling of failures during HTTP requests
• ... and dozens more fixes and enhancements!

Thanks a bunch to our five external contributors ❤️

See the release notes for examples and more details.

This release includes big updates to the Prefect Cloud login experience as well as improvements to logging (e.g. task run logs on remote workers are sent to the API now).

See the release notes for details!

This release includes new agent options, many bug fixes, and the announcement of a newprefect-vault collection.

See the release notes for details.

This release adds timeouts for task runs, colored logging, obfuscation of secret settings, dark mode for documentation, and more! It also includes usability updates in the UI, performance improvements for the scheduler, and lots of bug fixes.

See the release notes for more details 🎉

This release adds status displays for work queue health to the UI and support for waiting for upstream tasks before starting a subflow!

It also includes:

• Scheduler performance improvements
• Documentation for our versioning scheme
• Three new collections: GitLab, Alerts, and Fivetran!

See the full release notes for more details.

This release includes bug fixes and enhancements to the core library. The highlights are:

• Support for manual flow run retries
• Improved server performance when retrying flow runs with many tasks
• Status checks in work queues
• Support for Python 3.11

See the release notes for details.

This release includes bug fixes and enhancements to the core library, but the highlights are in our collections. There's a new infrastructure block for running flows in containers on Azure, a new key-value store collection, and the ECS infrastructure block has been designated as ready for production.

See the release notes for details.

By the way, if you haven't checked out 2.6 yet, you should now! We launched first-class configuration of results in 2.6.0 and every release since includes additional improvements to result handling.

This release fixes a bug in 2.6.2 where Task.with_options could fail due to a null value for the new cache_result_in_memory setting.

This release includes some major enhancements, including support for compound recurrence rule schedules and private GitHub repositories. There are also fixes and enhancements for the new results feature; notably, we now support reducing memory consumption by optionally always retrieving results from storage.

As always, the highlights above are a small subset of the changes. See the release notes for everything else!

This release patches a bug where flow and task return values of {} or [] could be incorrectly coerced to None.

👋 This release includes some big features like first-class configuration of results, waiting for failed tasks, specifying an agent's work queues with a prefix string, and setting parameters when running deployments from the CLI. There are also a bunch of fixes and performance improvements.

Check out the release notes for examples and more details.

This release includes the ability to easily orchestrate flow runs for deployments from other flow runs! If you're currently using OrionClient.create_flow_run_from_deployment, or if that's been too complicated, you'll be excited to see our sync-compatible run_deployment function that links created runs as subflows. There are a bunch of other enhancements and fixes too! Check out the release notes for details.

This release disables block protection. With block protection enabled, as in 2.4.3 and 2.4.4, client and server versions cannot be mismatched unless you are on a version before 2.4.0. Disabling block protection restores the ability for a client and server to have different version.

Block protection was added in 2.4.1 to prevent users from deleting block types that are necessary for the system to function. With this change, you are able to delete block types that will cause your flow runs to fail. New safeguards that do not affect client/server compatibility will be added in the future.

Note this release also restores block memoization by default, reverting the change in 2.4.4 as it did not patch the issue as intended.

When running a server with this version, the client must be the same version. This does not apply to clients connecting to Prefect Cloud.

This release attempted to fix a bug affecting a subset of users on 2.4.3 where block schemas cannot be saved. This release did not address the issue as intended and is fixed in 2.4.5.

When running a server with this version, the client must be the same version. This does not apply to clients connecting to Prefect Cloud.

This release includes several performance enhancements, CLI quality of life improvements, and numerous fixes - full release notes here

This release includes some fixes to bugs with blocks in 2.4.1, see the release notes for details.

Thanks to our users who found and helped fix these issues!

This release includes bug fixes and a lot of enhancements. We're sure you'll be excited about:

• Microsoft Teams notifications
• Improved orchestration for duplicate flow runs
• A CLI command to generate a Kubernetes manifest for agents
• Performance improvements for blocks

Check out the release notes to see all of the changes.

Changes

Enhancements

• Add get_latest to TaskRunView — #6749
• Update agents to allow logs to be disabled with PREFECT__CLOUD__SEND_FLOW_RUN_LOGS — #6833

Fixes

• Fix bug where scheduler_comm.close_rpc() was not awaited in DaskExecutor — #6793
• Fix bug where the Prefect diagnostics logger could duplicate output — #6832

Task library

• Add validation for node type id and instance pool id in Databricks NewCluster — #6853

This release includes updates to deployments for using custom infrastructure blocks as well as many enhancements and bugfixes.

Read the full release notes for details.

Changes

Fixes

• Fix bug where flows with emoji characters pass flow.validate(), but fail to extract from file - #6375
• Fix FlowRunView task run query errors when all task runs in the flwo run are cached - #6572
• Update FlowRunView to avoid retrieval of cached static tasks during latest calls - #6572
• Update FlowRunView to be robust to multithreaded usage - #6572

Task Library

• Add state handler to log to Snowflake - #5671
• Fix default idempotency token for DatabricksSubmitMultitaskRun - #6412
• Fix custom tag type for Databricks NewCluster - #6573
• Fix empty file handling in S3List task - #6028

Contributors

• Dennis Hinnenkamp
• Edmondo Porcu
• Mate Hricz
• Stéphan Taljaard