Datasets:
roanbrasil
/
devops-gitops-corpus

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
59.1k
- `--concurrency <count>` - Maximum concurrent items (default: 1)
- `--dataset <name>` - Dataset name override applied at runtime
- `--experiment-name <name>` - VoltOps experiment name override
- `--tag <trigger>` - VoltOps trigger source tag (default: "cli-experiment")
- `--dry-run` - Skip VoltOps submission (local scoring only)
Example with concurrency:
```bash
pnpm volt eval run \
--experiment ./src/experiments/offline.experiment.ts \
--concurrency 4
```
The CLI automatically resolves TypeScript imports, streams progress to stdout, and links the run to VoltOps when credentials are present in environment variables (`VOLTAGENT_PUBLIC_KEY` and `VOLTAGENT_SECRET_KEY`).
## Step 6: Use a Custom Dataset Resolver
Instead of inline items, you can provide a custom `resolve` function to fetch data from external sources, databases, or APIs.
```ts
import { createExperiment } from "@voltagent/evals";
export default createExperiment({
id: "custom-dataset-example",
dataset: {
name: "custom-source",
resolve: async ({ limit, signal }) => {
// Fetch from your API, database, or any source
const response = await fetch("https://api.example.com/test-data", { signal });
const data = await response.json();
// Apply limit if provided
const items = limit ? data.slice(0, limit) : data;
return {
items, // Can be an array or async iterable
total: data.length, // Optional: total count for progress tracking
dataset: {
name: "custom-source",
metadata: { source: "api", fetchedAt: new Date().toISOString() },
},
};
},
},
runner: async ({ item }) => {
// Your runner logic
return { output: processItem(item.input) };
},
});
```
**Resolver function:**
- Receives `{ limit?, signal? }` as arguments
- Can return an iterable, async iterable, or object with `{ items, total?, dataset? }`
- Supports streaming large datasets with async iterables
- `signal` enables cancellation handling
Example with async iterable for streaming:
```ts
dataset: {
resolve: async function* ({ signal }) {
for await (const item of fetchItemsStream()) {
if (signal?.aborted) break;
yield item;
}
},
}
```
## Step 7: Use Named Datasets from VoltOps
For production workflows, store datasets in VoltOps and reference them by name. This enables version control, collaboration, and reusable test suites.
```ts
import { createExperiment } from "@voltagent/evals";
export default createExperiment({
id: "voltops-dataset-example",
dataset: {
name: "support-qa-v1",
versionId: "abc123", // Optional - defaults to latest version
limit: 100, // Optional - limit items processed
},
runner: async ({ item }) => {
// Your runner logic
return { output: processItem(item.input) };
},
});
```
**Dataset configuration:**
- `name` - Dataset name in VoltOps (required)
- `versionId` - Specific version ID (optional, defaults to latest)
- `limit` - Maximum number of items to process (optional)
When you run the experiment with a `voltOpsClient`, the dataset is automatically fetched from VoltOps. If no client is provided, the experiment fails with a clear error message.