Description

Hook providers need to observe individual streaming chunks during model response generation for use cases like real-time logging, token metrics, and content filtering. Currently, BeforeModelCallEvent and AfterModelCallEvent bookend the streaming call, but nothing fires while the stream is active.

ModelStreamChunkEvent is now available for hook providers to observe each raw streaming chunk:

from strands.hooks import HookProvider, HookRegistry
from strands.hooks.events import ModelStreamChunkEvent

class StreamingLogger(HookProvider):
    def register_hooks(self, registry: HookRegistry) -> None:
        registry.add_callback(ModelStreamChunkEvent, self.on_chunk)

    def on_chunk(self, event: ModelStreamChunkEvent) -> None:
        if "contentBlockDelta" in event.chunk:
            print(f"Received delta: {event.chunk}")

agent = Agent(hooks=[StreamingLogger()])

The event fires for all streaming chunks (including metadata, message start/stop) after yielding the internal typed event, preserving existing consumer behavior. This achieves parity with the TypeScript SDK's ModelStreamUpdateEvent.

Related Issues

#1819

Documentation PR

N/A

Type of Change

New feature

Testing

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.