Changelog

All notable changes to structured-evaluation are documented here.

For the canonical changelog, see CHANGELOG.md in the repository.

v0.4.0 - 2026-05-23

Breaking Changes: Switch from numeric scores to categorical pass/partial/fail values.

Highlights

• Switch from numeric scores to categorical pass/partial/fail values
• New terminal and markdown renderers for evaluation reports

Changed

• Breaking: CategoryScore renamed to CategoryResult with Score field (pass/partial/fail)
• Breaking: ScoreStatus renamed to ScoreValue with values pass/partial/fail
• Breaking: Removed WeightedScore from EvaluationReport
• Added CategoryCounts to Decision (pass/partial/fail counts)
• Updated detailed renderer to display category counts

Added

• render/terminal package with ANSI-colored output and UTF8 icons
• render/markdown package for Markdown report generation
• CLI terminal and markdown render format options

---

v0.3.1 - 2026-05-19

Build

• Updated CI workflows to use shared workflows
• Renamed workflow files to standard filenames

Dependencies

• Bump github.com/invopop/jsonschema from 0.13.0 to 0.14.0

---

v0.3.0 - 2026-03-01

Changed

• Breaking: Module path changed from github.com/agentplexus/structured-evaluation to github.com/plexusone/structured-evaluation

---

v0.2.0 - 2026-01-26

Highlights

• Rubric definitions with score anchors
• Judge metadata tracking
• Pairwise comparison mode
• Multi-judge aggregation

Added

• Rubric and ScoreAnchor types
• JudgeMetadata for tracking LLM configuration
• PairwiseComparison for relative evaluation
• MultiJudgeResult for aggregating evaluations

---

v0.1.0 - 2026-01-26

Highlights

• Initial release
• LLM-as-Judge evaluation reports
• GO/NO-GO summary reports
• DAG-based aggregation

Added

• evaluation package with core types
• summary package for deterministic checks
• combine package with DAG aggregation
• render packages for terminal and box output
• sevaluation CLI tool