Skip to content

Introduce Reason property on EvaluationMetric - #6087

Merged
shyamnamboodiripad merged 18 commits into
dotnet:mainfrom
shyamnamboodiripad:metrics
Mar 13, 2025
Merged

Introduce Reason property on EvaluationMetric#6087
shyamnamboodiripad merged 18 commits into
dotnet:mainfrom
shyamnamboodiripad:metrics

Conversation

@shyamnamboodiripad

@shyamnamboodiripadshyamnamboodiripad commented Mar 12, 2025

Copy link
Copy Markdown
Contributor

Also includes changes to the report to display this information. This addresses #6032.

Additionally, this PR also includes numerous general improvements to the TypeScript report rendering -

  • Make the metric cards clickable and display metric details (such as diagnostics, reasons etc.) on click in a new collapsible section inline in the report (as opposed to in a hover tooltip). This addresses [AI Evaluation] Improve interaction for metric cards in the report #6037.
  • Show conversations in a friendlier chat bubble form and make the sections that display conversation history collapsible so that long conversations can be collapsed. This addresses [AI Evaluation] Improve how chat history is displayed in reports #6036 partially.
  • Introduce a global settings pane and move the per-textbox toggles for rendering markdown to the global location. This also addresses [AI Evaluation] Improve how chat history is displayed in reports #6036 partially.
  • In the scenario tree, collapse single children into their respective parent level to reduce the amount of clicking required to expand deep trees.
  • Remove the scenario-level section for failure reasons since it is now possible to view failure reasons (and diagnostics) on a per-metric basis by clicking on metric cards.
  • A bunch of other minor layout, sizing and UX improvements and fixes for the report.
Microsoft Reviewers: Open in CodeFlow

peterwald pushed a commit that referenced this pull request Mar 13, 2025
Also includes changes to the report to display this information. This addresses #6032.
Additionally, this PR also includes numerous general improvements to the TypeScript report rendering -
* Make the metric cards clickable and display metric details (such as diagnostics, reasons etc.) on click in a new collapsible section inline in the report (as opposed to in a hover tooltip). This addresses #6037.
* Show conversations in a friendlier chat bubble form and make the sections that display conversation history collapsible so that long conversations can be collapsed. This addresses #6036 partially.
* Introduce a global settings pane and move the per-textbox toggles for rendering markdown to the global location. This also addresses #6036 partially.
* In the scenario tree, collapse single children into their respective parent level to reduce the amount of clicking required to expand deep trees.
* Remove the scenario-level section for failure reasons since it is now possible to view failure reasons (and diagnostics) on a per-metric basis by clicking on metric cards.
* A bunch of other minor layout, sizing and UX improvements and fixes for the report.
@github-actionsgithub-actionsBot locked and limited conversation to collaborators Apr 12, 2025
Sign up for freeto subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

area-ai-evalMicrosoft.Extensions.AI.Evaluation and related

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

@shyamnamboodiripad@peterwald