[Abhinav Maurya (Fire162)]

Motivation & Context

When resuming a workflow from a checkpoint, passing function_invocation_kwargs or client_kwargs did not take effect because checkpoint restoration ran after _run_workflow_with_tracing committed the caller kwargs, wiping state with checkpoint.state. The contract specifies that checkpoint-restored kwargs should be preserved unless the caller explicitly supplies new kwargs on continuation.

Description & Review Guide

• What are the major changes?
  • In Workflow._run_core, restore the checkpoint (restore_from_checkpoint) before _run_workflow_with_tracing begins (matching the pattern in cancel_pending_requests).
  • Separated _execute_with_message to seed the entry message, and updated _resolve_execution_mode to handle message dispatch and responses cleanly.
  • Added regression test test_caller_kwargs_override_checkpoint_kwargs_on_restore in test_workflow_kwargs.py covering standalone checkpoint_id resumption, subsequent response resumption, and combined checkpoint_id + responses resumption for both function_invocation_kwargs and client_kwargs.
• What is the impact of these changes?
  • Callers can override run kwargs when resuming workflows from a checkpoint without losing new parameters. Existing behavior without kwargs continues to preserve checkpoint-stored kwargs.
• What do you want reviewers to focus on?
  • Execution order in _run_core and backward compatibility of _resolve_execution_mode.

Related Issue

Fixes #9850

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution. I reviewed, understood, and verified all submitted content and accept responsibility for it.

AI assistance details: Analysis of checkpoint restoration ordering, implementation in _workflow.py, and regression test in test_workflow_kwargs.py.

Contribution Checklist

• The code builds clean without any errors or warnings
• All unit tests pass, and I have added new tests where possible
• The PR follows the Contribution Guidelines
• This PR links to an agreed issue with no competing open PR, or the Related Issue section documents a trivial-change or repository-automation exception.
• This is not a breaking change.

[Copilot]

🟡 Changes recommended

Checkpoint restoration occurs outside the tracing/error boundary, so restore failures bypass failure events and error tracing and leave workflow status unchanged.

1 open issue

• Keep checkpoint restoration inside workflow error tracing boundary · New

What changed in this PR

This PR addresses #9850 by restoring Python workflow checkpoints before applying caller-supplied run kwargs, so continuation overrides take precedence over checkpoint values.

Changes:

• Moves checkpoint restoration ahead of run-kwargs application.
• Separates initial-message setup from checkpoint and response dispatch.
• Adds regression coverage for both kwargs channels across checkpoint-only, subsequent-response, and combined resumptions.

File	Description
python/​packages/​core/​tests/​workflow/​test_workflow_kwargs.py	Tests caller-kwargs precedence when resuming checkpoints.
python/​packages/​core/​agent_framework/​_workflows/​_workflow.py	Reorders restoration and refactors execution-mode setup.

_{Review effort: Balanced}