Caio Nogueira

This commit **updates the documentation for Workflows** to reflect the new capability of **JavaScript Workflows steps supporting streamed output**. It **enhances the 'Get Started' guide** (`src/content/docs/workflows/get-started/guide.mdx`) with details on returning readable streams and clarifies relevant code examples. Additionally, the **Workflows limits documentation** (`src/content/docs/workflows/reference/limits.mdx`) is updated to specify constraints for stream outputs and non-stream results, providing crucial context for developers. A **release note entry** (`src/content/release-notes/workflows.yaml`) is also added, ensuring users are informed about this significant **new feature** for Workflows. This **documentation update** provides essential guidance for users leveraging streamed data in their workflow steps.

This commit **fixes a memory tracking bug** within the **Workerd tracing API**, specifically addressing an inaccuracy in how **tail attributes** are accounted for. It corrects the `Trace::addTailAttributes` method in `src/workerd/api/trace.c++` to ensure the proper size of these attributes is included when added to a trace. This **bug fix** improves the reliability of **memory usage reporting** for trace data, preventing potential discrepancies in resource allocation and management within the Workerd runtime.

This commit introduces a **new capability** to propagate **tail attributes** within the **Workerd tracing system**, allowing richer contextual information to be passed from a calling worker to a tail worker. It extends the `TraceItem` and `Trace` data structures to store these attributes and updates the **worker communication interface** (`worker-interface.capnp`) to include them in RPC messages. The `WorkerTracer` component is modified to accept and manage these attributes, while `Server::WorkerService` is updated to pass them during worker instantiation. This enhancement provides more comprehensive and contextual tracing across distributed worker invocations, improving observability for complex workflows.

This commit **updates the documentation for Python Workflows** to align with recent **SDK changes**, primarily focusing on the `step.do` decorator and dependency resolution. It details the **new implicit dependency resolution mechanism** where parameter names define dependencies, and introduces the use of `WorkerEntrypoint` for workflow entrypoints. The updates span examples, DAG explanations, and the `python-workers-api` reference, also removing the 'experimental' compatibility flag. This ensures users have accurate and up-to-date guidance for developing Python Workflows with the latest SDK features and patterns.

This commit introduces a new capability for Python Workflows within the `workerd` environment, enabling implicit dependency resolution for Python steps. This significantly simplifies the development and execution of Python workflows by automatically handling dependencies.

This commit **updates the documentation** for **Python Workflows**, specifically modifying a configuration example to ensure compatibility. It **bumps the `compatibility_date`** within the `src/content/docs/workflows/python/bindings.mdx` file from `2024-10-22` to `2025-10-24`. This **maintenance update** ensures that the example configuration aligns with the current default entrypoint class, providing users with accurate and functional guidance for their Python Workflow setups. The change helps users avoid potential compatibility issues when implementing new workflows based on the documentation.

This commit provides an isolated fix to Miniflare's workflow binding mechanism, preventing incorrect resource disposal or behavior when dealing with void-returning methods. This change resolves a specific issue without broader impact on other systems.

This commit introduces new workflow and instance binding wrappers to the Pyodide Python workers module, enhancing its capabilities for managing worker interactions and enabling more complex worker-based workflows.

This change introduces a new capability for manually disposing workflow engine stubs, primarily affecting Miniflare's workflow plugin and shared workflow binding logic. It enhances resource management for developers by providing more control over the lifecycle of workflow engine instances.

This change promotes the `python_workflows` feature by removing its experimental flag, signaling its readiness for general availability and broader adoption.

This commit **updates the documentation** for the `wrangler workflows instances describe` command within the **Wrangler CLI**. It is a **documentation maintenance** task that clarifies a recent change in the command's behavior. Specifically, the `ID` parameter for `wrangler workflows instances describe` is now **optional and defaults to 'latest'**, improving user understanding and usability of the command. This ensures the `src/content/docs/workers/wrangler/commands.mdx` file accurately reflects the current functionality.

This commit refactors the core workflow handling mechanism within Pyodide, making `run` the primary workflow handler, which may alter how Python workflows are initiated and managed. It also includes fixes for internal worker operations and `sleep_until` functionality.

This commit introduces a new workflows module and event waiting mechanisms within Pyodide's internal worker system. This foundational change adds new capabilities for managing asynchronous operations and multi-step processes, likely enabling future workflow-driven features without immediate external impact.

This commit optimizes Python workflow execution within the `workerd` environment by memoizing steps for DAG dependency resolution. This is an internal performance improvement for the workflow engine.

This commit delivers a **critical bug fix** to the **`pyodide` internal worker system**, resolving issues with **error propagation** in **async/await chains**. It ensures that exceptions from **asynchronous Python operations** are correctly caught and handled by adjusting the **internal worker logic** in `_workers.py` and `workers.py`, and the **Python entrypoint helper** in `python-entrypoint-helper.ts`. Supporting changes include updated **metadata definitions** in `metadata.ts`, and the fix is validated by new tests for the **`workerd` server's Python workflow entrypoint**. This significantly enhances the reliability of **async workflows** by preventing silent failures and ensuring proper error reporting.

This commit introduces a **new compatibility flag**, `python_workflows`, to control the availability of Python workflow handling within the system. It **conditionally enables** Python workflow support in the `makeEntrypointProxyHandler` within the **Pyodide integration** based on this flag, affecting how Python entrypoints are processed. The internal `_do_call` method in `src/pyodide/internal/_workers.py` was refactored to support this change, ensuring that Python workflows are only active when explicitly opted in via the new flag. This **feature control** provides a controlled rollout mechanism for the Python workflow capability, preventing its activation by default.

This commit introduces a **new capability** by **refactoring** the `_WorkflowStepWrapper.do` method within **Pyodide's internal workflow mechanism**. It now supports both a **decorator-based syntax and an explicit callback pattern** for defining workflow steps, enhancing flexibility in workflow design. The `src/pyodide/python-entrypoint-helper.ts` module was updated to correctly handle these new workflow entrypoint patterns, ensuring seamless integration between Python and TypeScript components. This change provides developers with more versatile options for structuring **workflow entrypoints** and includes an updated example in `src/workerd/server/tests/python/workflow-entrypoint/workflow.py` to demonstrate the new usage.

This commit **implements special handling for `WorkflowEntrypoint` classes**, enabling their proper execution and interaction within the system. This **new capability** introduces `_WorkflowStepWrapper` in **Pyodide's internal worker management** (`src/pyodide/internal/_workers.py`) to facilitate JavaScript step interactions and correctly processes `on_run` for Python `WorkflowEntrypoint` subclasses. Concurrently, the **Python entrypoint helper** (`src/pyodide/python-entrypoint-helper.ts`) is updated to recognize and proxy these workflow entrypoints and their `run` method. This **feature implementation** is thoroughly validated with new **integration tests** under `src/workerd/server/tests/python/workflow-entrypoint`, ensuring robust **Python workflow execution within Workerd**.

This commit introduces a **new feature** by adding the `WorkflowRetentionDuration` field to the `WorkflowInstanceCreateOptions` interface. It updates the **type definitions** in `types/generated-snapshot/experimental/index.ts` and `types/generated-snapshot/latest/index.ts` to include new types for workflow duration and retention. This enhancement allows users to specify a **retention policy** directly when creating a workflow instance, providing greater control over the lifecycle and data persistence of workflows.

This commit performs a **text correction** within the **Workflows** module by **removing a duplicated word**. This is a **maintenance** task aimed at improving the clarity and correctness of the user interface or documentation. The change has no functional impact on the system, but it enhances the overall quality and professionalism of the application's textual content.