Add properties in Evaluation Result - Custom Evaluator extra fields. #46077
Merged
Add properties in Evaluation Result - Custom Evaluator extra fields. #46077
Conversation
github-actions
bot
added
the
Evaluation
Contributor
There was a problem hiding this comment.
Pull request overview
Adds support for passing through custom evaluator “extra fields” via a properties bag into the AOAI-style evaluation result objects produced by the evaluation results converter.
Changes:
- Update
_extract_metric_valuesto detect anoutputs.<criteria>.propertiesdict and propagate it onto per-metric extracted values. - Update
_create_result_objectto includepropertiesin the final AOAI result payload when present. - Add a unit test asserting
propertiesis preserved and not flattened into the top-level result object.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate.py | Propagates a per-criteria properties dict into per-metric result objects during AOAI conversion. |
| sdk/evaluation/azure-ai-evaluation/tests/unittests/test_evaluate.py | Adds coverage validating properties passthrough behavior for a custom evaluator result row. |
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate.py
Show resolved
Hide resolved
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate.py
Show resolved
Hide resolved
…esults Pass through evaluator properties dict in AOAI evaluation results. When an evaluator returns a properties dict, it is included alongside score, label, reason, threshold, and passed in the result object. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
w-javed
force-pushed
the
waqasjaved02/aoai-properties-passthrough
branch
from
2fd8ce2 to
c8f5958
Compare
Update _extract_metric_values and _create_result_object docstrings to document the new properties field and its expected dict type. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
YoYoJa
approved these changes
Apr 2, 2026
aprilk-ms
reviewed
Apr 2, 2026
sdk/evaluation/azure-ai-evaluation/azure/ai/evaluation/_evaluate/_evaluate.py
Outdated
Show resolved
Hide resolved
Address PR review: warn users when their custom evaluator returns 'properties' as a non-dict type so they can fix the output format. Also add properties to _create_result_object example input. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…thub.com/Azure/azure-sdk-for-python into waqasjaved02/aoai-properties-passthrough
slister1001
force-pushed
the
waqasjaved02/aoai-properties-passthrough
branch
from
e9b6431 to
5ec9b95
Compare
Remove erroneous space in self._eval_metric. value (two occurrences) that would cause an AttributeError at runtime when building result keys for _details and _total_tokens fields. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
nagkumar91
reviewed
Apr 3, 2026
Member
nagkumar91
left a comment
nagkumar91
left a comment
There was a problem hiding this comment.
Review — Properties passthrough
- Shared dict reference (bug-risk) — In
_extract_metric_values, the samepropertiesobject is assigned to every metric entry:
for metric_dict in result_per_metric.values():
metric_dict["properties"] = properties # same object referenceIf anything downstream mutates one entry's properties, all entries are affected. Consider metric_dict["properties"] = properties.copy() (or copy.deepcopy if nested dicts matter).
-
No test for the warning path — The
isinstance(metric_value, dict)guard logs a warning when properties isn't a dict, but no test covers this branch. A quick test passingproperties="not_a_dict"would confirm the warning fires and properties is omitted. -
Typo fixes in
_base_rai_svc_eval.py— Good catch onself._eval_metric. value→self._eval_metric.value. Cosmetic (Python allows whitespace after dot) but worth cleaning up.
- Use properties.copy() to avoid shared dict reference across metrics - Add test for non-dict properties logging and omission - Change properties type mismatch log level from warning to info Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
nagkumar91
approved these changes
Apr 3, 2026
Member
nagkumar91
left a comment
nagkumar91
left a comment
There was a problem hiding this comment.
All review items addressed — properties.copy(), non-dict test coverage, and log level adjustment. LGTM.
slister1001
added a commit
that referenced
this pull request
Apr 3, 2026
…46077) * feat(evaluation): support properties passthrough in AOAI evaluation results Pass through evaluator properties dict in AOAI evaluation results. When an evaluator returns a properties dict, it is included alongside score, label, reason, threshold, and passed in the result object. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * docs: update docstrings for properties passthrough per PR review Update _extract_metric_values and _create_result_object docstrings to document the new properties field and its expected dict type. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: log warning when properties is not a dict Address PR review: warn users when their custom evaluator returns 'properties' as a non-dict type so they can fix the output format. Also add properties to _create_result_object example input. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Release-1-16-4 * Fix stray space in _eval_metric.value attribute access Remove erroneous space in self._eval_metric. value (two occurrences) that would cause an AttributeError at runtime when building result keys for _details and _total_tokens fields. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Remove empty changelog sections to fix Build Analyze check Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Address PR feedback: copy properties dict and add non-dict test - Use properties.copy() to avoid shared dict reference across metrics - Add test for non-dict properties logging and omission - Change properties type mismatch log level from warning to info Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Sydney Lister <sydneylister@microsoft.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Evaluation Result is OpenAI complaint. It contains score, label, reason etc.
We have a need to present more eval results to UI. So, introducing this property bag, in which we can add more details. Science team is also going to use this property bag, which can show more outputs per evaluator.