Description
The Python sample appears to create or reuse a workflow instance in a way that is not safe for concurrent execution:
python/samples/05-end-to-end/hosted_agents/writer_reviewer_agents_in_workflow/main.py
When the hosted agent receives parallel requests, it can attempt to execute the same workflow instance concurrently and fails with:
RuntimeError: Workflow is already running. Concurrent executions are not allowed.
This makes the sample effectively broken for hosted scenarios where parallel requests are expected.
Link
https://github.com/microsoft/agent-framework/blob/main/python/samples/05-end-to-end/hosted_agents/writer_reviewer_agents_in_workflow/main.py#L55
The sample should be safe for parallel requests in a hosted agent scenario.
Instead of reusing a single workflow instance across requests, it should create a fresh workflow per execution, for example:
async def main() -> None:
client = AzureOpenAIChatClient(credential=AzureCliCredential())
await from_agent_framework(lambda: create_workflow(client)).run_async()
Code Sample
Error Messages / Stack Traces
Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/azure/ai/agentserver/agentframework/_agent_framework.py", line 249, in stream_updates
async for event in streaming_converter.convert(updates):
File "/usr/local/lib/python3.12/site-packages/azure/ai/agentserver/agentframework/models/agent_framework_output_streaming_converter.py", line 379, in convert
async for group in chunk_on_change(updates, is_changed):
File "/usr/local/lib/python3.12/site-packages/azure/ai/agentserver/agentframework/models/utils/async_iter.py", line 39, in chunk_on_change
async for group in chunk_by_key(source, lambda x: x, key_equal=key_equal):
File "/usr/local/lib/python3.12/site-packages/azure/ai/agentserver/agentframework/models/utils/async_iter.py", line 69, in chunk_by_key
pending = await it.__anext__()
^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/agent_framework/_workflows/_agent.py", line 203, in run_stream
async for update in self._run_stream_impl(
File "/usr/local/lib/python3.12/site-packages/agent_framework/_workflows/_agent.py", line 282, in _run_stream_impl
async for event in event_stream:
File "/usr/local/lib/python3.12/site-packages/agent_framework/_workflows/_workflow.py", line 529, in run_stream
self._ensure_not_running()
File "/usr/local/lib/python3.12/site-packages/agent_framework/_workflows/_workflow.py", line 234, in _ensure_not_running
raise RuntimeError("Workflow is already running. Concurrent executions are not allowed.")
RuntimeError: Workflow is already running. Concurrent executions are not allowed.Package Versions
azure-ai-agentserver-agentframework==1.0.0b14
Python Version
No response
Additional Context
No response
Description
The Python sample appears to create or reuse a workflow instance in a way that is not safe for concurrent execution:
python/samples/05-end-to-end/hosted_agents/writer_reviewer_agents_in_workflow/main.py
When the hosted agent receives parallel requests, it can attempt to execute the same workflow instance concurrently and fails with:
This makes the sample effectively broken for hosted scenarios where parallel requests are expected.
Link
https://github.com/microsoft/agent-framework/blob/main/python/samples/05-end-to-end/hosted_agents/writer_reviewer_agents_in_workflow/main.py#L55
The sample should be safe for parallel requests in a hosted agent scenario.
Instead of reusing a single workflow instance across requests, it should create a fresh workflow per execution, for example:
Code Sample
Error Messages / Stack Traces
Package Versions
azure-ai-agentserver-agentframework==1.0.0b14
Python Version
No response
Additional Context
No response