tool_trajectory_avg_score class doesn't accept string values for `match_type`

[alexhermida]

Describe the bug
The new implementation in the ToolTrajectoryCriterion class doesn't accept string values for match_type (e.g., "ANY_ORDER") causing a Pydantic validation error. The enum expects integer values (0, 1, or 2) but in the documentation shows string values. It works with integers.

To Reproduce
Create an evaluation configuration with a string value for match_type:

from google.adk.evaluation.eval_metrics import EvalMetric, ToolTrajectoryCriterion, PrebuiltMetrics

eval_metric = EvalMetric(
    threshold=0.5,
    metric_name=PrebuiltMetrics.TOOL_TRAJECTORY_AVG_SCORE.value,
    criterion={
        "threshold": 0.1,
        "match_type": "ANY_ORDER"  # Using string as documented
    }
)

Error:

  pydantic_core._pydantic_core.ValidationError: 1 validation error for ToolTrajectoryCriterion
  match_type
    Input should be 0, 1 or 2 [type=enum, input_value='ANY_ORDER', input_type=str]
      For further information visit https://errors.pydantic.dev/2.12/v/enum

Expected behavior
The criterion should accept string values ("EXACT", "IN_ORDER", "ANY_ORDER") as shown in the ADK documentation.

Desktop:

• OS: macOS
• Python version: 3.13.0
• ADK version: 1.19.0

Root Cause
The MatchType enum in src/google/adk/evaluation/eval_metrics.py uses integer values (0, 1, 2), but Pydantic v2
doesn't automatically convert string enum names to enum values without an explicit validator.

Proposed Solution
Add a Pydantic field_validator to the ToolTrajectoryCriterion.match_type field that converts string enum names to their corresponding enum values if we want to keep compatibility with integer values.

I've already tested a fix locally and can open a PR if that would be helpful, with comprehensive test coverage of course. Let me know if you'd like me to contribute the fix.

Thanks for all hard work!

[surajksharma07]

@alexhermida You're right - Pydantic v2 doesn't auto-convert string enum names without a field_validator.

Until we add that to ToolTrajectoryCriterion (eval_metrics.py:161-225), use integer values (0=EXACT, 1=IN_ORDER, 2=ANY_ORDER) or enum members like ToolTrajectoryCriterion.MatchType.ANY_ORDER.

Your proposed fix with @field_validator is correct and we'd be happy to review your PR if you'd like to contribute it with test coverage.

Thanks!

[alexhermida]

thanks @surajksharma07 for the reply and to @sreeharsha1902 for the #3979 PR