Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
The VoltOps Console provides a web interface for dataset management at `https://console.voltagent.dev/evals/datasets`.
### Creating Datasets
1. Click **Create Dataset**
2. Enter dataset name and description
3. Add tags for organization
4. Submit to create an empty dataset
### Adding Items
**Single Item:**
1. Open a dataset
2. Click **Add Item**
3. Enter JSON for input and expected fields
4. Optionally add labels and metadata
5. Save the item
**Bulk Import:**
1. Click **Import Items**
2. Paste JSON array of items:
```js
[
{
label: "test-1",
input: "What is 2+2?",
expected: "4",
extra: { category: "math" },
},
{
label: "test-2",
input: "What is the capital of France?",
expected: "Paris",
extra: { category: "geography" },
},
];
```
### Version Management
Datasets automatically version when items change:
1. Each modification creates a new version
2. Versions are numbered sequentially (v1, v2, v3)
3. Previous versions remain immutable
4. Experiments reference specific versions
## Working with Dataset Items
### Input Formats
Input fields accept any JSON-serializable value:
```typescript
// String input
{ input: "Translate 'hello' to Spanish" }
// Object input
{ input: { text: "Hello", targetLang: "es" } }
// Array input
{ input: ["item1", "item2", "item3"] }
// Complex nested structure
{
input: {
messages: [
{ role: "user", content: "Hi" },
{ role: "assistant", content: "Hello" }
],
context: { userId: "123", sessionId: "abc" }
}
}
```
### Expected Output Patterns
Expected values are compared by scorers:
```typescript
// Exact string match
{ expected: "The answer is 42" }
// Numeric comparison
{ expected: 3.14159 }
// Structured data
{ expected: { status: "success", result: 100 } }
// Partial matching with extra metadata
{
expected: "Paris",
extra: {
acceptableAnswers: ["Paris", "Paris, France"],
scoreThreshold: 0.8
}