Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
## Combining Offline and Live Evaluations
Use live evals for real-time monitoring and offline evals for regression testing:
- **Live**: Sample 5-10% of production traffic with fast scorers (moderation, keyword match)
- **Offline**: Run comprehensive LLM judges on curated datasets nightly
Both share the same scorer definitions. Move scorers between eval types as needed.
## Next Steps
- [Offline Evaluations](/evaluation-docs/offline-evaluations) - Regression testing and CI integration
- [Prebuilt Scorers](/evaluation-docs/prebuilt-scorers) - Full catalog of prebuilt scorers
- [Building Custom Scorers](/evaluation-docs/building-custom-scorers) - Create your own evaluation scorers
<|endoftext|>
# source: VoltAgent__voltagent/website/evaluation-docs/datasets.md type: docs
---
title: Evaluation Datasets
sidebar_position: 2
---
# Evaluation Datasets
Datasets are collections of test cases used to evaluate agent performance. Each dataset item contains an input prompt, an optional expected output, and metadata to help organize and analyze results.
## Dataset Structure
Datasets follow a consistent JSON structure whether stored locally or in VoltOps:
```js
{
"name": "customer-support-qa",
"description": "Customer support question-answer pairs",
"tags": ["support", "qa", "production"],
"metadata": {
"version": "1.0.0",
"created": "2025-01-10"
},
"data": [
{
"name": "refund-policy",
"input": "What is your refund policy?",
"expected": "We offer a 30-day money-back guarantee...",
"extra": {
"category": "policies",
"difficulty": "easy"
}
}
]
}
```
### Field Descriptions
| Field | Type | Required | Description |
| ------------- | -------- | -------- | ------------------------------------- |
| `name` | string | Yes | Unique identifier for the dataset |
| `description` | string | No | Human-readable description |
| `tags` | string[] | No | Labels for filtering and organization |
| `metadata` | object | No | Additional structured data |
| `data` | array | Yes | Collection of dataset items |
### Dataset Item Structure
Each item in the `data` array contains:
| Field | Type | Required | Description |
| ---------- | ------ | -------- | --------------------------------------------- |
| `name` | string | No | Item identifier for tracking |
| `input` | any | Yes | Input to the agent (string, object, or array) |
| `expected` | any | No | Expected output for comparison |
| `extra` | object | No | Additional context or metadata |
## Creating Datasets
### JSON Files
Store datasets as JSON files in `.voltagent/datasets/`:
```js
{
"name": "math-problems",
"description": "Basic arithmetic problems",
"data": [
{
"input": "What is 15 + 27?",
"expected": "42"
},
{
"input": {
"operation": "multiply",
"a": 7,
"b": 8
},
"expected": 56
}
]
}