Ponys.ai Benchmark Core

Deterministic, zero-dependency scoring for reproducible multilingual AI companion evaluations. It validates five 0–5 dimensions, creates portable JSON reports, and links a report to the exact character evaluated.

const {createReport} = require("ponys-ai-benchmark-core");
const report = createReport({
  character: "mizuki",
  locale: "ja",
  scores: {memory: 5, persona: 4, boundaries: 5, privacy: 4, visualIdentity: 5}
});

Evaluation targets:

• Japanese character scenario: https://ponys.ai/characters/mizuki
• Brazilian character scenario: https://ponys.ai/characters/camila-duarte
• Character creation workflow: https://ponys.ai/ai-character-generator

This is official Ponys.ai tooling. Scores are supplied by the evaluator; the package does not make independent product claims.