[shuaiyafei]

What does this PR do?

The DatasetClient.run_experiment vocabulary-diversity example reads result["output"], but run evaluators receive ExperimentItemResult objects. Copying it into a run evaluator raises TypeError: 'ExperimentItemResult' object is not subscriptable, so no run-level diversity score is produced. Use result.output, matching the result type and the other run-evaluator examples.

Found by checking the existing example against the SDK; no linked issue. This changes only the docstring, with no runtime API or generated-file changes.

Type of change

• Documentation update

Verification

Local verification used Python 3.12.15.

• Extracted the documented content_diversity function and executed it through the SDK's _run_evaluator with real ExperimentItemResult objects. Before: the nonempty case raises the TypeError above. After: two outputs (red blue, blue green) produce 0.75; an empty result list produces 0.0.
• uv run --frozen pytest tests/unit/test_datasets.py -q: 9 passed.
• uv run --frozen ruff format --check .: passes, 593 files already formatted.
• Configured pre-commit hooks on the changed file: Ruff check/format and full SDK mypy pass.
• Built the API reference through scripts/build_reference_docs.sh into a local output directory. The build succeeds with warnings about unresolved generated API re-exports. No live Langfuse or model-provider tests were run; validation uses local synthetic data.

Checklist

• Self-reviewed the one-line diff with code_review.md using Codex.
• Updated the affected example and executed it against the actual SDK result type.
• Did not hand-edit generated files or commit secrets.

AI assistance: OpenAI Codex (GPT-6, autonomous) prepared and validated this change.

Confidence Score: 5/5

This documentation fix appears safe to merge.

Summary

Fixes the content_diversity example in DatasetClient.run_experiment.

• The diversity example reads output from each result object.

_{Reviews (1) · Last reviewed commit: "docs(datasets): use result attributes in..." · Reviewed by Greptile}

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}