[Agentic Evaluators]: Accept input string as is

[m7md7sien]

Description

Updates agentic evaluators to accept string inputs directly without requiring structured (list/dict) format. When response, query, tool_calls, or tool_definitions are passed as plain strings, evaluators now forward them to the LLM as-is, skipping parsing, reformatting, and cross-validation steps that only apply to structured conversation objects.

Changed evaluators:
Groundedness, Tool Call Accuracy, Tool Call Success, Tool Input Accuracy, Tool Output Utilization, and Tool Selection.

All SDK Contribution checklist:

• The pull request does not introduce [breaking changes]
• CHANGELOG is updated for new features, bug fixes or other significant changes.
• I have read the contribution guidelines.

General Guidelines and Best Practices

• Title of the pull request is clear and informative.
• There are a small number of commits, each of which have an informative message. This means that previously merged commits do not appear in the history of the PR. For more information on cleaning up the commits in your PR, see this page.

Testing Guidelines

• Pull request includes test coverage for the included changes.

[Copilot]

Pull request overview

This PR updates several agentic evaluators to accept plain string inputs (for query, response, tool_calls, and/or tool_definitions) and forward them to the LLM without attempting structured parsing/reformatting/cross-validation intended for message-list inputs.

Changes:

• Allow tool_calls / tool_definitions / response to be passed through as strings across multiple tool-related evaluators.
• Skip tool-definition filtering and conversation/response reformatting when inputs are already strings (notably in Tool Output Utilization).
• Add a groundedness short-circuit for string response inputs in agent-response mode.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_selection/_tool_selection.py	Accept string tool calls/definitions; skip extraction/validation steps when strings are provided.
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_output_utilization/_tool_output_utilization.py	Conditionally skip filtering/reformatting when query/response/tool_definitions are strings.
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_input_accuracy/_tool_input_accuracy.py	Allow response/tool calls to pass through as strings and bypass structured cross-validation.
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_call_success/_tool_call_success.py	Avoid tool-call/result reformatting when response is a string; pass string tool definitions through.
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_call_accuracy/_tool_call_accuracy.py	Allow string tool calls/definitions; bypass structured cross-validation.
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_groundedness/_groundedness.py	Add early return for string response in agent-response mode to avoid context extraction/parsing.

Comments suppressed due to low confidence (1)

sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluators/_tool_selection/_tool_selection.py:180

• tool_names can now be a str (when tool_calls is a string), but the name and the inline comment imply it’s a list of tool names. This is easy to misread later; consider renaming the variable / clarifying the comment to reflect that it may be a raw string forwarded to the LLM.

        # Extract only tool names from tool calls, removing parameters and results (skip for strings)
        tool_names = tool_calls if isinstance(tool_calls, str) else self._extract_tool_names_from_calls(tool_calls)

        return {
            "query": query,
            "tool_calls": tool_names,  # Only tool names, no parameters
            "tool_definitions": needed_tool_definitions,