AICORE

Documentation RAG Evaluation Baseline (R0)

Give documentation RAG a deterministic, per-module, retrieval-only evaluation baseline (dataset + scoring harness + test:ai:evaluate-documentation command + regression gate), starting with the Core/user index.

State
Done
Progress
100%
Steps
36 / 36
Started
2026-09-15
Finished
2026-10-30
Modules
LaraAI, LaraCore

Tasks.

  • Task 1: DocumentationEvaluationCase value object5/5
  • Task 2: DocumentationEvaluationDataset value object5/5
  • Task 3: Deterministic corpus fixtures (stub embedder + fake search)5/5
  • Task 4: DocumentationEvaluationService5/5
  • Task 5: test:ai:evaluate-documentation command5/5
  • Task 6: Core/user dataset + fixture corpus + regression gate5/5
  • Task 7: RAG documentation note3/3
  • Final verification3/3