Skip to content

Python: Python : checkpoint restore +pending request deserialization failure when hosting workflow as agent with ResponsesHostserver in foundry #7618

Description

@dgsear

Scenario

This occurs when a workflow is hosted as an agent via ResponseHostServer (Response protocol), aand the workflow is resumed across requests using previous_response_id

when it happened

  • During the hosted-agent execution (work-as-agent ), sfter a prior turn produced request_info and pending requests.
  • on the next /responses call with previous_response_id + function_call_output, the server restored checkpoint state and failed on deserializing a pending request type.

observed behaviour

  • checkpoint decode warning:
    checkpoint deserialization blocked for type 'agent_framework_orchestrations._handoff:HandoffAgentUserRequest'
    checkpoint deserialization blocked for type agent_framework_orchestrations._magentic:MagenticPlanReviewRequest

  • Restore proceeded from an older/other checkpoint state

  • Follow-up response was rejected

  • `ValueError :Response provided for unknown request Id :<request_id>

Expected

When hosting workflow as agent:
-pending request checkpoint state should restore reliably across turns

Package versions:
agent-framework-core==1.13.0
agent-framework-foundry-hosting==1.0.0.b260730

Metadata

Metadata

Labels

hostingUsage: [Issues, PRs], Target: all hosting related solutionspythonUsage: [Issues, PRs], Target: PythonreproducedUsage: [Issues], Target: all issues that can be reproduced by the triage workflow

Type

Projects

Status
No status

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions