text
stringlengths
0
59.1k
```bash
npm run volt eval dataset pull -- [options]
```
**Options:**
| Option | Description | Default |
| ----------------- | --------------------------------------- | --------------------------------- |
| `--name <name>` | Dataset name to pull | Interactive prompt |
| `--id <id>` | Dataset ID (overrides name) | - |
| `--version <id>` | Specific version ID | Latest version |
| `--output <path>` | Output file path | `.voltagent/datasets/<name>.json` |
| `--overwrite` | Replace existing file without prompting | `false` |
| `--page-size <n>` | Items per API request (1-1000) | `200` |
**Interactive Mode:**
When no options provided, presents an interactive menu:
1. Select dataset from list
2. Choose version if multiple exist
3. Confirm file location
**Conflict Resolution:**
When target file exists:
- Prompts for action (overwrite/new file/cancel)
- `--overwrite` flag skips prompt
- Suggests alternative filename
**Examples:**
```bash
# Interactive mode
npm run volt eval dataset pull
# Pull specific dataset
npm run volt eval dataset pull -- --name production-qa
# Pull specific version
npm run volt eval dataset pull -- --name qa-tests --version v3
# Pull by ID with custom output
npm run volt eval dataset pull -- \
--id dataset_abc123 \
--version version_xyz789 \
--output ./test-data/qa.json
# Force overwrite
npm run volt eval dataset pull -- --name staging-data --overwrite
```
## Experiment Commands
### `eval run`
Execute an experiment definition against a dataset.
```bash
npm run volt eval run -- --experiment <path> [options]
```
**Options:**
| Option | Description | Default |
| -------------------------- | ----------------------------------------- | ---------------- |
| `--experiment <path>` | Path to experiment module file (required) | - |
| `--dataset <name>` | Override dataset name at runtime | - |
| `--experiment-name <name>` | Override experiment name for VoltOps | - |
| `--tag <trigger>` | VoltOps trigger source tag | `cli-experiment` |
| `--concurrency <n>` | Maximum concurrent items (1-100) | `1` |
| `--dry-run` | Run locally without VoltOps submission | `false` |
**Environment Variables:**
- `VOLTAGENT_PUBLIC_KEY` - Required for VoltOps integration
- `VOLTAGENT_SECRET_KEY` - Required for VoltOps integration
- `VOLTAGENT_API_URL` - VoltOps API endpoint
- `VOLTAGENT_DATASET_NAME` - Default dataset name
**Experiment File Format:**
The experiment file must export a default `createExperiment` result:
```typescript
// experiments/my-test.experiment.ts
import { createExperiment } from "@voltagent/evals";
export default createExperiment({
id: "my-test",
dataset: { name: "test-data" },
runner: async ({ item }) => ({
output: await processItem(item.input),
}),
scorers: [
/* ... */
],
passCriteria: { type: "passRate", min: 0.95 },