Workflow Streams: Live interactivity for agents and other applications

Workflow Streams is a new feature for Temporal's Durable Execution that allows for the creation of rich interactive user interfaces by providing a durable streams abstraction, facilitating the seamless updating of UIs through incremental updates. This feature is beneficial for applications such as AI agents, payment processing, and order status updates by allowing monitors to intervene or confirm correct operations. The library uses Temporal's core primitives, like Workflows, Signals, Updates, and Queries, to maintain an ordered sequence of typed events that can be published and subscribed to by various components including Activities, Child Workflows, and external clients. The system ensures exactly-once semantics, enabling reliable event tracking and minimal overhead for publishing within Workflows. Available in Python and TypeScript during its Public Preview, Workflow Streams integrates with frameworks like OpenAI Agents SDK and Google ADK, allowing for responsive UI updates with adjustable batch intervals. It contrasts with traditional messaging systems like Redis by providing inherent exactly-once delivery without additional infrastructure, making it suitable for scenarios involving multiple independent streams tied to single Workflows rather than high-throughput streaming needs.