axiom/dist/docs/evals/functions/Eval.md

Axiom AI SDK provides - an API to wrap your AI calls with observability instrumentation. - offline evals - online evals

[**axiom v0.51.1**](../../README.md)

***

[axiom](../../README.md) / [evals](../README.md) / Eval

# Function: Eval()

> **Eval**\<`TInput`, `TExpected`, `TOutput`, `Name`, `Capability`, `Step`\>(`name`, `params`): `void`

Creates and registers an evaluation suite with the given name and parameters.

This function sets up a complete evaluation pipeline that will run your [EvalTask](../type-aliases/EvalTask.md)
against a collection, score the results, and provide detailed EvalCaseReport reporting.

## Type Parameters

### TInput

`TInput`

### TExpected

`TExpected`

### TOutput

`TOutput`

### Name

`Name` *extends* `string` = `string`

### Capability

`Capability` *extends* `string` = `string`

### Step

`Step` *extends* `string` = `string`

## Parameters

### name

`ValidateName`\<`Name`\>

Human-readable name for the evaluation suite

### params

`Omit`\<[`EvalParams`](../type-aliases/EvalParams.md)\<`TInput`, `TExpected`, `TOutput`\>, `"capability"` \| `"step"` \| `"scorers"`\> & `object`

[EvalParams](../type-aliases/EvalParams.md) configuration parameters for the evaluation

## Returns

`void`

## Example

```typescript
import { Eval } from 'axiom/ai/evals';

Eval('Text Generation Quality', {
  capability: 'capability-name',
  data: async () => [
    { input: 'Explain photosynthesis', expected: 'Plants convert light to energy...' },
    { input: 'What is gravity?', expected: 'Gravity is a fundamental force...' }
  ],
  task: async ({ input }) => {
    const result = await generateText({
      model: yourModel,
      prompt: input
    });
    return result.text;
  },
  scorers: [similarityScorer, factualAccuracyScorer],
});
```