Temporal has released Workflow Streams in Public Preview for Python and TypeScript, a durable streams abstraction built on top of Temporal's core primitives (Workflows, Signals, Updates, and Queries). It enables rich interactive UIs for long-running workflows by streaming status updates, agent reasoning, tool calls, and generated text in real time. The library uses batching and long-polling to balance throughput and responsiveness, provides exactly-once delivery semantics, and handles edge cases like Continue-as-New and Activity retries. It integrates out of the box with OpenAI Agents SDK, Google ADK, and LangGraph. Unlike Redis pub/sub, it requires no additional infrastructure and keeps the UI as a pure function of the durable event log. Go and Java support are planned next.

9m read timeFrom temporal.io
Post cover image
Table of contents
Overview #Background #Using Workflow Streams #The Workflow Streams implementation #What about…? #Integrations #Complete example #Conclusion #Further information #
27 Impressions