[pandego]

Related issue

Closes #5102

Summary

• treat an empty BaseAgentState placeholder as "no resumable state" when an agent expects a specialized state model
• let SequentialAgent resume from the beginning instead of failing on BaseAgentState() -> SequentialAgentState validation
• add focused regression coverage for the generic loader guard and the sequential resume path

Problem / solution

InvocationContext.populate_invocation_agent_states() can synthesize an empty base placeholder when an agent emitted content without an explicit state payload. That is fine for agents that really use BaseAgentState, but workflow agents like SequentialAgent later validate against a specialized state model and crash on resume.

This change keeps the placeholder behavior but teaches _load_agent_state() to treat an empty base placeholder as absent state when the caller expects a custom state type. In that case, the workflow resumes cleanly from its normal initial path instead of raising a ValidationError.

Testing plan

• uv run pytest tests/unittests/agents/test_base_agent.py tests/unittests/agents/test_sequential_agent.py

Checklist

• Added or updated unit tests
• Added manual E2E evidence if needed
• Kept the change focused to one bug

[rohityan]

Hi @pandego ,Thank you for your contribution! We appreciate you taking the time to submit this pull request. Your PR has been received by the team and is currently under review. We will provide feedback as soon as we have an update to share.

[rohityan]

Hi @wukath , can you please review this.

[GWeale]

Thanks for the fast turnaround and the clean repro/tests! This was already resolved on main: populate_invocation_agent_states() now stores the placeholder as BaseAgentState().model_dump(mode="json") (a dict) instead of a raw BaseAgentState() object, so a SequentialAgent reading it back no longer raises the ValidationError (#5102 is closed). Closing as already resolved.