​
Overview

The Reporting system provides basic evaluation results in a readable format. For each evaluation run, it shows pass/fail status, scores, and any relevant error messages or reasons. Results are displayed both individually and as a summary, making it easier to track how your model performed across different metrics.

​
Features

​
Individual Test Results

Each evaluation provides:

  • A pass/fail status
  • A numerical score (0-1)
  • Detailed reasons for the evaluation result

​
Cumulative Summaries

The report includes:

  • Total number of evaluations
  • Number of passed and failed tests
  • Overall duration of evaluations
  • Average scores across metrics

​
Metric-specific Breakdowns

For each metric type:

  • Individual scores and statuses
  • Detailed failure reasons when applicable
  • Performance trends across multiple evaluations

​
Example

import { evaluate, RelevancyMetric, HallucinationMetric } from '@evalkit/core';

const results = await evaluate({
  input: "What is the weather like today?",
  output: "The temperature is 72°F with partly cloudy skies.",
  context: "Current conditions: 72°F, partly cloudy"
}, [RelevancyMetric, HallucinationMetric]);

// outputs
📊 EvalKit Report Summary
========================
🕒 Duration: 0.81s
📝 Total Evaluations: 2
✅ Passed: 2
❌ Failed: 0

📈 Metrics Breakdown
------------------
Relevancy Evaluation:
  Score: 0.90
  Passed: 1 Failed: 0

Hallucination Evaluation:
  Score: 1.00
  Passed: 1 Failed: 0

The report provides a clear overview of how well your model performed across different evaluation metrics, making it easy to identify areas that need improvement.

Semantic Similarity