Skip to content

Evaluation: Human-readable summary generation #1093

Description

@AkhileshNegi

Is your feature request related to a problem?
Evaluation runs currently lack a single human-readable summary that reflects overall performance, forcing users to scan individual rows to gauge run quality without a weighted view of contributing dimensions.

Describe the solution you'd like

  • Create a human-readable AI summary for each evaluation run.
  • Calculate an overall score as a weighted combination of:
    • Adherence to ground truth — 50%
    • Adherence to knowledge base — 30%
    • Adherence to prompt — 20%
  • Display a breakdown per dimension with: score, weight, and delta indicator.
  • Include an overall status badge (e.g. "Good") based on the overall score.
  • Implement legend/color coding for score bands (0–0.3 needs improvement, 0.3–0.6 could improve, 0.6–1 good) and confirm the canonical scale before implementation.

Metadata

Metadata

Assignees

Type

No type

Projects

Status
Closed

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions