Fix UnboundLocalError in prompty response formatting when inputs are omitted

[Copilot]

format_llm_response could raise UnboundLocalError by referencing sample_input before assignment when inputs is not provided. This surfaced in async evaluator flows that call prompty without explicit input payloads.

• Runtime fix in prompty formatting path

  • Initialize sample_input to an empty string before conditional input serialization.
  • Preserve existing behavior when inputs is present (still emits serialized sample input payload).

• Regression coverage

  • Added a focused unit test for format_llm_response(..., inputs=None) to ensure the function returns successfully and emits sample_input == "".

• Changelog

  • Added an unreleased bug-fix entry documenting the UnboundLocalError fix.

input_str = f"{json.dumps(inputs)}" if inputs else ""
sample_input = ""
if inputs and len(inputs) > 0:
    sample_input_json = []
    msg = ChatCompletionUserMessageParam(role="user", content=input_str)
    sample_input_json.append(msg)
    sample_input = json.dumps(sample_input_json)

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Fixes an UnboundLocalError in prompty response formatting when format_llm_response is invoked without inputs (seen in async evaluator flows), and adds regression coverage plus a changelog entry.

Changes:

• Initialize sample_input to "" so format_llm_response(..., inputs=None) can’t reference it before assignment.
• Add a unit test validating the inputs=None path returns successfully and sets sample_input to an empty string.
• Document the fix in the unreleased changelog.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_legacy/prompty/_utils.py	Initializes sample_input defensively to prevent UnboundLocalError when inputs is omitted.
sdk/evaluation/azure-ai-evaluation/tests/unittests/test_prompty_format_llm_response.py	Adds regression test covering format_llm_response(..., inputs=None) behavior.
sdk/evaluation/azure-ai-evaluation/CHANGELOG.md	Notes the bug fix in the unreleased section.

[github-actions]

Hi @copilot. We've noticed that new changes have been pushed to this pull request. Because it is set to automatically merge, we've reset the approvals to allow the opportunity to review the updates.