chore(evaluator): added evaluator name and evaluator_type for report

[poshinchen]

Description

Two fixes that make EvaluationReport rows uniquely identifiable when the same evaluator class is instantiated multiple times in one experiment.

1. Evaluator instance names:

• Evaluator.__init__ now accepts an optional name= kwarg.
• get_name() returns the instance name when set, otherwise falls back to the class name.
• The per-row evaluator tag in EvaluationReport.cases (and the gen_ai.evaluation.name OTel attribute) keys on get_name() so two Contains(value="x") / Contains(value="y") instances no longer collide.
• Experiment._validate_evaluator_names rejects duplicates up front with a clear message.
• Store and retrieved through Evaluator.to_dict / from_dict.

2. evaluator_type row tag:

• Every report row now also carries evaluator_type (= Evaluator.get_type_name(), the class name) next to the existing evaluator instance tag.
  • This lets downstream consumers group/aggregate across all instances of the same class (e.g., "average score across every Contains instance") without re-deriving the type from the instance name.
  • EvaluationReport.flatten already uses dict(case), so the field flows through to JSON via to_dict / to_file automatically.

Surface area is intentionally narrow: no changes to the Rich display, the CLI summary line, or OTel span attributes — those still key on the existing gen_ai.evaluation.name.

Results

Related Issues

Follow-up to the non-blocking nit on #241: #241 (review)

Documentation PR

N/A

Type of Change

Other: small enhancement to evaluator identification + report row schema. Backwards-compatible — existing consumers that read cases[i]["evaluator"] see the same values; evaluator_type is purely additive.

Testing

• New test_run_evaluations_two_same_class_evaluators_with_distinct_names covers the doubling scenario from the PR-241 review and now also asserts evaluator_type == "Contains" across both instances.

• New test_run_evaluations_rejects_duplicate_evaluator_names covers the validation path.

• New test_evaluator_name_round_trips_through_to_dict_from_dict and test_evaluator_name_default_omitted_from_to_dict cover serialization.

• All 111 tests in tests/strands_evals/test_experiment.py pass locally.

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.