> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentium.in/llms.txt
> Use this file to discover all available pages before exploring further.

# Reliability Evaluation

> Verify tool usage, error handling, and response completeness

## Overview

`ReliabilityEval` asserts that agents call expected tools, handle errors correctly, and produce non-empty responses.

## Quick Start

```typescript theme={null}
import { ReliabilityEval } from "@agentium/eval";
import { Agent, openai } from "@agentium/core";

const agent = new Agent({
  name: "tool-agent",
  model: openai("gpt-4o"),
  tools: [searchTool, calcTool],
});

const eval = new ReliabilityEval({
  name: "tool-reliability",
  agent,
  cases: [
    { name: "uses-search", input: "Search for latest news", expectedTools: ["search"] },
    { name: "handles-error", input: "Divide by zero", shouldError: true },
  ],
});

const result = await eval.run();
```

## Case Options

| Field           | Type       | Description                            |
| --------------- | ---------- | -------------------------------------- |
| `expectedTools` | `string[]` | Tool names that should be called       |
| `shouldError`   | `boolean`  | Whether the case should throw an error |

## Tool Call Match Scorer

Use `toolCallMatch` as a standalone scorer:

```typescript theme={null}
import { EvalSuite, toolCallMatch } from "@agentium/eval";

const suite = new EvalSuite({
  name: "tools-test",
  agent,
  scorers: [toolCallMatch(["search", "calculate"])],
  cases: [{ name: "test", input: "Search and calculate" }],
});
```