streaming

Streaming

Strands' model stream produces a sequence of StreamEvents (token deltas, tool-use chunks, message-stop, etc.). With TemporalAgent(streaming_topic="events"), the model activity publishes each chunk onto a workflow-hosted WorkflowStream; external subscribers read it via WorkflowStreamClient. Chunks are batched on streaming_batch_interval (default 100ms) to keep activity overhead low.

What This Sample Demonstrates

• Hosting a WorkflowStream on a workflow
• TemporalAgent(streaming_topic=...) publishing model events
• An external subscriber reading StreamEvents in real time via WorkflowStreamClient

Running the Sample

# Terminal 1
uv run strands_plugin/streaming/run_worker.py

# Terminal 2
uv run strands_plugin/streaming/run_workflow.py

The starter prints text deltas as they arrive, then the final workflow result.

Files

File	Description
workflow.py	StreamingWorkflow hosting WorkflowStream + TemporalAgent(streaming_topic="events")
run_worker.py	Registers StrandsPlugin, starts the worker
run_workflow.py	Starts the workflow and subscribes to the events topic