[BUG] Streaming chunks does not return `complete` boolean to signify end of stream

[jer96]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

latest

Python Version

all

Operating System

mac/linux

Installation Method

pip

Steps to Reproduce

1. Invoke a strands agent with stream_async
2. Observe that no event is returned with complete == True

Expected Behavior

1. It is expected that the agent will return a complete boolean to signify the end of the stream

Actual Behavior

1. 2. Observe that no event is returned with complete == True

Additional Context

Strands docs explicitly states that this is the correct API contract: https://strandsagents.com/latest/user-guide/concepts/streaming/async-iterators/?h=complete#event-types

However no events have this boolean attached to them.

Here is a sample of the final events of a request for the customer who encountered this bug:

{'data': '', 'delta': {'text': ''}, 'event_loop_metrics': EventLoopMetrics(cycle_count=2, tool_metrics={'get_skills': ToolMetrics(tool={'toolUseId': 'tooluse_VRlHZpAsRf60cvNz4lZHkg', 'name': 'get_skills', 'input': {}}, call_count=1, success_count=1, error_count=0, total_time=0.01149606704711914)}, cycle_durations=[0.9196999073028564], traces=[<strands.telemetry.metrics.Trace object at 0x7fb75e9aaa50>, <strands.telemetry.metrics.Trace object at 0x7fb75d9f6650>], accumulated_usage={'inputTokens': 1221, 'outputTokens': 236, 'totalTokens': 1457}, accumulated_metrics={'latencyMs': 1320}), 'agent': <strands.agent.agent.Agent object at 0x7fb75e9aae40>, 'event_loop_parent_span': None, 'event_loop_cycle_id': UUID('9bb554c7-d9c8-4bc0-8c3b-29add2f18b5c'), 'request_state': {}, 'event_loop_cycle_trace': <strands.telemetry.metrics.Trace object at 0x7fb75d9f6650>, 'event_loop_cycle_span': None, 'event_loop_parent_cycle_id': UUID('dfb845c6-6d8c-4d24-acd2-5c4156527ffe')}
{'event': {'contentBlockStop': {'contentBlockIndex': 0}}}
{'event': {'messageStop': {'stopReason': 'end_turn'}}}
{'event': {'metadata': {'usage': {'inputTokens': 704, 'outputTokens': 169, 'totalTokens': 873}, 'metrics': {'latencyMs': 873}}}}
{'message': {'role': 'assistant', 'content': [{'text': '<thinking> \nI have retrieved the list of possible employee skills using the "get_skills" tool. Now I will present this list to the user.\n</thinking>\n\nHere is the list of possible employee skills:\n\n1. Kotlin\n2. AWS\n3. Spring Boot\n4. Angular\n5. Docker\n6. React\n7. Git\n8. CI/CD\n9. REST API\n10. Machine Learning\n11. JavaScript\n12. DevOps\n13. GraphQL\n14. Node.js\n15. MongoDB\n16. TypeScript\n17. Java\n18. Python\n19. Kubernetes\n20. SQL\n\nIf you need more detailed information about any specific skill or a list of employees with particular skills, please let me know!'}]}}

Possible Solution

1. The revised support for async iterators and events (driven by @pgrayy) may mitigate this.

Related Issues

No response