Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
### Managing Datasets with the CLI
Use the CLI to push local datasets to VoltOps or pull remote datasets locally:
**Push a dataset:**
```bash
pnpm volt eval dataset push \
--name support-qa-v1 \
--file ./datasets/support-qa.json
```
**Pull a dataset:**
```bash
pnpm volt eval dataset pull \
--name support-qa-v1 \
--version abc123 \
--output ./datasets/support-qa-v1.json
```
**CLI dataset commands:**
- `push` - Upload a local dataset file to VoltOps
- `--name <datasetName>` (required) - Dataset name
- `--file <datasetFile>` - Path to dataset JSON file
- `pull` - Download a dataset version from VoltOps
- `--name <datasetName>` - Dataset name (defaults to `VOLTAGENT_DATASET_NAME`)
- `--id <datasetId>` - Dataset ID (overrides `--name`)
- `--version <versionId>` - Version ID (defaults to latest)
- `--output <filePath>` - Custom output file path
- `--overwrite` - Overwrite existing file if present
- `--page-size <size>` - Number of items to fetch per request
## Configuration Reference
### Required Fields
#### `id`
Unique identifier for the experiment. Used in logs, telemetry, and VoltOps run metadata.
```ts
id: "support-regression";
```
#### `dataset`
Specifies the evaluation inputs. Three approaches:
**Inline items:**
```ts
dataset: {
items: [
{ id: "1", input: "hello", expected: "hello" },
{ id: "2", input: "goodbye", expected: "goodbye" },
];
}
```
**Named dataset (pulled from VoltOps):**
```ts
dataset: {
name: "support-qa-v1",
versionId: "abc123", // optional - defaults to latest
limit: 100, // optional - limit items processed
}
```
**Custom resolver:**
```ts
dataset: {
name: "custom-source",
resolve: async ({ limit, signal }) => {
const items = await fetchFromAPI(limit, signal);
return {
items,
total: items.length,
dataset: { name: "custom-source", metadata: { source: "api" } },
};
},
}
```
The resolver receives `{ limit?, signal? }` and returns an iterable, async iterable, or object with `{ items, total?, dataset? }`.
#### `runner`
Function that executes your agent/workflow for each dataset item. Receives a context object and returns output.
**Context properties:**
- `item` - Current dataset item (`{ id, input, expected?, label?, extra?, ... }`)
- `index` - Zero-based position in the dataset
- `total` - Total number of items (if known)
- `signal` - AbortSignal for cancellation handling