#3309 fix(pyproject.toml): Exclude gcsfs 2024.2.0 due to bug Pull request opened by pingsutw ## Tracking issue NA ## Why are the changes needed? There is a race condition in gcsfs 2024.2.0 ## What changes were proposed in this pull request? Exclude gcsfs 2024.2.0 in pyproject.toml ## How was this patch tested? Run a workflow on the internal cluster ### Setup process ### Screenshots ## Check all the applicable boxes • I updated the documentation accordingly. • All new and existing tests passed. • All commits are signed-off. flyteorg/flytekit

#3310 feat: Add faulthandler registration for SIGUSR1 Pull request opened by pingsutw ## Why are the changes needed? No log or stack trace when the task is hanging. ## What changes were proposed in this pull request? Register a signal handler, so that we can see the stacktrace when the task gets killed ## How was this patch tested? Run a task, and kill the pod ### Setup process ### Screenshots ## Check all the applicable boxes • I updated the documentation accordingly. • All new and existing tests passed. • All commits are signed-off. ## Related PRs ## Docs link https://bito.ai#summarystart| ### Summary by Bito This pull request enhances signal handling in the Flytekit library by registering the faulthandler for SIGUSR1, enabling stack trace capture for better debugging when tasks hang or are terminated. flyteorg/flytekit

#51 improve secrets cli: add validation, provide secret value through input Pull request opened by cosmicBboy This PR aligns the actual implementation of

flyte create secret

with the docstring. Users need to supply a secret

--value

flag, and if not provided, an input prompt that hides the password value is shown to the user. flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/cebe3e5ac76f54d75b37c21498dca71f56d5b69a|cebe3e5a>

- improve secrets cli: add validation, provide secret value through input (#51) flyteorg/flyte-sdk

#52 implement json output format for flyte get commands Pull request opened by cosmicBboy Fixes: https://linear.app/unionai/issue/BB-5830/support-json-output-for-flyte-cli-commands With these changes, you could do something like: flyte get run --output-format json --limit 3 | jq '.[].action.id.run.name' | xargs -I {} flyte abort run {} Which will abort the last three runs. This adds the

--output-format

flag to the following commands: •

flyte get run

•

flyte get task

•

flyte get action

•

flyte get secret

flyteorg/flyte-sdk

#53 Reduce BlockingIOError (PollerCompletionQueue) log spew with reference tasks and reusable containers Pull request opened by kumare3 This adds a exception handler that removes the log spew for all loops flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/4d84d15ccfc01f290a36e7be3136d367ef6df6ed|4d84d15c>

- Reduce BlockingIOError (PollerCompletionQueue) log spew with reference tasks and reusable containers (#53) flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/8884ccd8f2e3f66588a772eb120c1a8dbc6883d8|8884ccd8>

- Add langgraph example (#41) flyteorg/flyte-sdk

#54 Abort & TimedOut propagation Pull request opened by kumare3 flyteorg/flyte-sdk

#3509 [Core feature] Improved error message with maptasks: using it within a task should error out Issue created by cosmicBboy ### Motivation: Why do you think this is important? Flytekit should raise a runtime error when map tasks are written in any of the following contexts: • When a user tries to write a

map_task

within a

task

function body @task def t1(): map_task(...)(...) • When a user tries to define a

map_task

at the module-level scope @map_task @task def t1(): ... or @task def t1(): ... my_map_task = map_task()(t1) ### Goal: What should the final outcome look like, ideally? The error should make it clear that `map_task`s can only be defined in workflow function bodies. ### Describe alternatives you've considered NA ### Propose: Link/Inline OR Additional context No response ### Are you sure this issue hasn't been raised already? • Yes ### Have you read the Code of Conduct? • Yes flyteorg/flyte

#3038 [Plugin] TypeTransformer for TensorFlow tf.data.Dataset Issue created by dennisobrien ### Motivation: Why do you think this is important? The

tf.data.Dataset

object encapsulates data as well as a preprocessing pipeline. It can be used in model

fit

,

predict

, and

evaluate

methods. It is widely used in Tensorflow tutorials and documentation and is considered a best practice when creating pipelines that saturate GPU resources. ### Goal: What should the final outcome look like, ideally? Flyte tasks should be able to pass

tf.data.Dataset

objects as parameters and accept them as return types. ### Describe alternatives you've considered There are caveats to passing

tf.data.Dataset

objects between tasks. Since a

tf.data.Dataset

object can have steps in the pipelines that use local Python functions (e.g., a

map

or

filter

step), there doesn't seem to be a way to serialize the object without effectively "computing" the graph pipeline. There are times this could be beneficial (doing an expensive preprocessing pipeline once can free up the CPU during training) but this could also be confusing to the Flyte end user. So while adding a type transformer for

tf.data.Dataset

is certainly possible, it's still a good question if Flyte should actually support it given all the caveats. The alternative to consider here is to not support

tf.data.Dataset

. This seems like a question for the core Flyte team. ### Propose: Link/Inline OR Additional context There are at least three main ways to serialize/deserialize

tf.data.Dataset

objects. 1. tf.data.Dataset.save and tf.data.Dataset.load. 2. tf.data.Dataset.snapshot 3. Iterator checkpointing These are probably in order of least complex to most complex. But determining the method of serialization/deserialization is an open question. Some additional links: • This tensorflow github issue is asking about ways to serialize/deserialize a

tf.data.Dataset

as a deep copy without having the side-effect of "computing" the pipeline. • I asked a similar question on the Tensorflow Forum. ### Are you sure this issue hasn't been raised already? • Yes ### Have you read the Code of Conduct? • Yes flyteorg/flyte

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by pingsutw

<https://github.com/flyteorg/flyte/commit/c1dfb9f7bcd198472c16808e29ecd5d37c252025|c1dfb9f7>

- Remove projectcontour annotations from helm charts by default (#6564) flyteorg/flyte

#55 [Question]: How to provide a tag to an image to be pushed to the registry Issue created by debasisdwivedy Hi, After a build as below, we want to push the image with a particular tag and not the generated tag like

cff584fb57776fcce397c6245f3de461

. Below is a code example:

import flyte

img = flyte.Image.from_debian_base()
img = img.clone(registry="<REGISTRY>", name="<NAME>")

We tried

img = img.clone(registry="<REGISTRY>", name="<NAME>:<TAG>")

, it did not work as name only means repo name. Is there a way to pass <TAG_NAME> as well Regards, flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by pingsutw

<https://github.com/flyteorg/flyte/commit/ba2e90643603166ae6462bb606b5d846ebe4e562|ba2e9064>

- Add option to configure service monitor for flyte scheduler (#6558) flyteorg/flyte

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by Sovietaced

<https://github.com/flyteorg/flyte/commit/e0d4fb37ee06fb0a447c9534c444a5d207a0663a|e0d4fb37>

- [charts] Flyte-Binary: Add preInitContainers value (#6559) flyteorg/flyte

#56 Making every command output json (In the future yaml too) Pull request opened by kumare3 ### PR Description #### Overview This pull request introduces significant improvements to the Flyte CLI and SDK around output formatting and JSON serialization. The goal is to make CLI output more flexible, scriptable, and user-friendly, especially for automation and integration with other tools. --- #### Key Changes 1. Output Format Option for CLI • Replaces the boolean --simple flag with a new --output-format (or -of) option. • Supports multiple formats: table (default), table-simple, and json. • All CLI commands that previously used get_table now use the new format function, allowing consistent output formatting based on user preference. 2. Common Formatting Utilities • Introduces a new format function in _common.py that handles table, simple table, and JSON output for iterable objects. • get_panel is updated to respect output format for consistency. 3. ToJSONMixin for Models • Adds a new ToJSONMixin class (in remote/_common.py) that provides to_dict and to_json methods. • All major API model classes (Project, Run, Task, Secret, Action, ActionDetails, ActionInputs, ActionOutputs, RunDetails, TaskDetails) now inherit from ToJSONMixin. • This enables easy and consistent JSON serialization for CLI output and programmatic use. 4. Improved List Commands • All list commands (e.g., flyte get projects, flyte get runs, flyte get tasks, etc.) now respect the --output-format option. • Limits for listing are more consistently handled (e.g., flyte get run/task now have a --limit option). 5. Codebase Modernization • Refactors all usages of simple argument to output_format. • Updates type annotations using Literal and typing_extensions.get_args for CLI option validation. --- #### Motivation • Provides a more scriptable CLI experience for users and automation by supporting JSON output. • Makes the CLI more flexible for different use cases (human-friendly tables, simple for scripting, JSON for parsing). • Unifies output formatting and serialization logic across the codebase for maintainability and consistency. --- #### Backward Compatibility • The --simple flag is removed and replaced by --output-format. The default remains a rich table, so most users will not notice a change unless they used --simple (who can now use --output-format table-simple). • All API models are backward compatible, now with additional methods for JSON serialization. --- #### How to Use • Human-readable table (default):

flyte get projects

• Simple table (for copy-pasting):

flyte get projects --output-format table-simple

• JSON output (for scripting/automation):

flyte get projects --output-format json

--- #### Example flyte get tasks --output-format json --limit 10 --- #### Additional Notes • All new output format logic is centralized, making future extensions (e.g., YAML, CSV) easy to add. • The ToJSONMixin is designed to work seamlessly with protobuf-based models as well as plain dataclasses. flyteorg/flyte-sdk

#57 [QUESTION]: How to set a directory path for a task that generates report Issue created by debasisdwivedy Hi, Below is a code:

@env.task(report=True)
async def task1():
    # Creating new tabs
    tab = flyte.report.get_tab("Task 1")
    tab.log("<p>Task 1 HTML log</p>")
    await flyte.report.flush.aio()

It tries to generate a report but i get an exception as below:

RuntimeUserError: [Errno 2] No such file or directory: '/tmp/9vebunncm4atxzmlfzl9chl5r/report.html'

I am running the task with local context:

flyte.with_runcontext(mode="local")

The

with_runcontext

has

raw_data_path

and

run_base_dir

but it does not pass to the internal_ctx() which is tied to a task. In the task decorator/function

@env.task(report=True)

is could not find a way to pass a directory path. Is there a way to create a report at a particular location os choice? Regards, flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte/tree/master|master>

by machichima

<https://github.com/flyteorg/flyte/commit/beda6aed865a4aaf2ea8b9bdc3768599dd784f4d|beda6aed>

- [BUG] invalid cron date schedule creates infinite loop in flytescheduler (#6555) flyteorg/flyte

#3311 [BUG] ImageSpec NoOpBuilder should always return True while calling should_build function Pull request opened by popojk ## Tracking issue closes #6561 ## Why are the changes needed? The NoOpBuilder should overrride should_build function in order to correctly trigger build_image function in an environment that docker daemon is not running. ## What changes were proposed in this pull request? Make

NoOpBuilder

override

should_build

function. ## How was this patch tested? Test workflow:

from flytekit.image_spec.noop_builder import NoOpBuilder
from flytekit import task, workflow, ImageSpec

image_spec = ImageSpec(
    name="test_image",
    builder="noop",
    base_image="localhost:30000/test_image:latest",
    registry="<http://ghcr.io/not_exist_registry|ghcr.io/not_exist_registry>"
    
)

@task(container_image=image_spec)
def task_1(message: str) -> str:
    print(message)
    return message

@workflow
def wf() -> str:
    return task_1("hello")

We define a docker registry in

image_spec

that is not existed, and we expect that the SDK will still use the image user defined without actually communicating with the docker registry. After remote running the workflow, the SDK will use the user defined image as expected. [截圖 2025-08-11 上午11 44 59](https://private-user-images.githubusercontent.com/66259759/476434361-7183965d-7c58-439b-89d1-dc6bd9c14919.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NTQ4OTM4MjQsIm5iZiI6MTc1NDg5MzUyNCwicGF0aCI6Ii82NjI1OTc1OS80NzY0MzQzNjEtNzE4Mzk2NWQtN2M1OC00MzliLTg5ZDEtZGM2YmQ5YzE0OTE5LnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNTA4MTElMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjUwODExVDA2MjUyNFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWQzNmMxNDAwYmYyMjQ3NWY3Y2YxNDE5Yjk3NjA2NzNhZmM4NWZkZWUyMjFjYzlhMmFiMjRmM2FlYWUzNjM3Y2QmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.sHhndKWpPeTrBaXIay-X3pVYZbG_Yiva--YOsmC7SRc) ### Setup process ### Screenshots ## Check all the applicable boxes • I updated the documentation accordingly. • All new and existing tests passed. • All commits are signed-off. ## Related PRs ## Docs link https://bito.ai#summarystart| ### Summary by Bito This pull request adds a new method, should_build, to the NoOpBuilder class, allowing the build_image function to be triggered even when the Docker daemon is down. It also enhances the testing suite with comprehensive tests to validate the NoOpBuilder's functionality across various environments, improving its reliability and behavior. flyteorg/flytekit

#58 Example of loading memory Pull request opened by kumare3 flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/92d93fe12fe4d547e5a8d682ee0816188fb8701f|92d93fe1>

- Abort & TimedOut propagation (#54) flyteorg/flyte-sdk

Pending review on flyteorg/flytekit-python-template - Manage reminder [#40] ImageSpec overhaul (zeryx) 2 years stale · 2 years old · Waiting on _ @broad-monitor-993_

Pending review on flyteorg/flyte-sdk - Manage reminder [#44] adding log handler for file if stream is false for audit and compliance (debasisdwivedy) 4 days stale · 4 days old · Waiting on _ @freezing-airport-6809_

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/e0c53b8b3ff9d977e7538d9f265a7c9b8feaff14|e0c53b8b>

- Making every command output json (In the future yaml too) (#56) flyteorg/flyte-sdk

1 new commit pushed to

<https://github.com/flyteorg/flyte-sdk/tree/main|main>

by kumare3

<https://github.com/flyteorg/flyte-sdk/commit/949e314a3ff426a35b6fb70230d7223145e487e6|949e314a>

- Example of loading memory (#58) flyteorg/flyte-sdk

#59 [WIP] proposal for triggers Pull request opened by kumare3 flyteorg/flyte-sdk