feat: enable evaluators after resume by using consistent thread_id and Command API

[Chibionos]

Summary

Enables evaluators to execute and produce scores after agent resumes from suspended state.

Changes

1. Consistent thread_id for suspend/resume

   • Changed from execution_id (new UUID each run) to eval_item.id (consistent identifier)
   • Ensures checkpoints can be found when resuming

2. Command API integration

   • Pass Command(resume=data) when --resume flag is set
   • Provides resume data to waiting interrupt() call

3. Enhanced logging

   • Added logging to track resume mode execution
   • Shows which eval_item.id is used for checkpoint loading

Testing

Tested with suspend/resume evaluation samples in uipath-langchain-python:

• ✅ Suspend phase creates checkpoints with eval_item.id as thread_id
• ✅ Resume mode detected and logged
• ✅ Checkpoint successfully loaded using consistent thread_id
• 🚧 Resume execution encounters Pydantic validation - needs investigation

Built on top of

• PR feat: add suspend/resume support for evaluations with RPA invocations #1126 - suspend/resume support for evaluations with RPA invocations

🤖 Generated with Claude Code

Development Package

• Use uipath pack --nolock to get the latest dev build from this PR (requires version range).
• Add this package as a dependency in your pyproject.toml:

[project]
dependencies = [
  # Exact version:
  "uipath==2.5.14.dev1011324039",

  # Any version from PR
  "uipath>=2.5.14.dev1011320000,<2.5.14.dev1011330000"
]

[[tool.uv.index]]
name = "testpypi"
url = "https://test.pypi.org/simple/"
publish-url = "https://test.pypi.org/legacy/"
explicit = true

[tool.uv.sources]
uipath = { index = "testpypi" }

[tool.uv]
override-dependencies = [
    "uipath>=2.5.14.dev1011320000,<2.5.14.dev1011330000",
]