> ## Documentation Index
> Fetch the complete documentation index at: https://docs.veri.studio/llms.txt
> Use this file to discover all available pages before exploring further.

# veri volumes

> Create, upload, list, and delete persistent storage volumes.

The `veri volumes` group manages persistent named storage that mounts into training workers at `/mnt/<volume-name>`. Conceptual overview lives in the [Volumes](/volumes) guide.

## Commands

| Command                              | Purpose                                                    |
| ------------------------------------ | ---------------------------------------------------------- |
| `create <name>`                      | Create a volume (idempotent).                              |
| `list`                               | List your volumes.                                         |
| `get <name>`                         | Show one volume's metadata.                                |
| `delete <name> [--yes]`              | Delete a volume + queue its S3 prefix for cleanup.         |
| `upload <vol> <path> [--to /remote]` | Upload one file. Auto-multipart for files > 5 GiB.         |
| `upload-dir <vol> <dir> [--to /]`    | Recursively upload a directory (up to 8 files concurrent). |
| `ls <vol> [path]`                    | List files inside the volume.                              |
| `rm <vol> <path>`                    | Delete one file.                                           |

## Create a volume

```bash theme={null}
veri volumes create my-corpus
```

Names must match `^[a-zA-Z0-9][a-zA-Z0-9_-]{0,127}$`. Re-running with the same name returns the existing volume — no error.

## Upload files

For a single file:

```bash theme={null}
veri volumes upload my-corpus ./train.jsonl
```

The remote path defaults to `/<filename>`. Override with `--to`:

```bash theme={null}
veri volumes upload my-corpus ./train.jsonl --to /data/2026-05/train.jsonl
```

Files larger than 5 GiB are uploaded with S3 multipart automatically. A progress bar shows transfer rate, ETA, and total transferred.

For a directory:

```bash theme={null}
veri volumes upload-dir my-corpus ./training-data --to /data
```

Up to 8 files transfer concurrently. The progress bar reports cumulative bytes across all files. A per-file failure aborts the batch — you can safely re-run since S3 PUT is idempotent on the same key.

## List + inspect

```bash theme={null}
$ veri volumes list
┃ name       ┃ id      ┃ size      ┃ files ┃ updated_at           ┃
┃ my-corpus  ┃ vol_abc ┃ 14.2 GiB  ┃ 1023  ┃ 2026-05-14T18:22:09Z ┃
```

```bash theme={null}
$ veri volumes get my-corpus --format json
{
  "object": "volume",
  "id": "vol_abc",
  "name": "my-corpus",
  "size": "14.2 GiB",
  "files": 1023,
  "updated_at": "2026-05-14T18:22:09Z"
}
```

List files inside a volume:

```bash theme={null}
veri volumes ls my-corpus            # all files
veri volumes ls my-corpus /data/2026 # filter by prefix
```

## Delete

```bash theme={null}
veri volumes delete my-corpus
```

Prompts for the volume name as confirmation and shows the file count + total bytes that will be deleted. Pass `--yes` to skip the prompt in scripts:

```bash theme={null}
veri volumes delete my-corpus --yes
```

The DB row is dropped immediately; the S3 prefix is reaped by a background reconciler within \~60 seconds. There is no undo.

Delete a single file inside a volume without removing the volume itself:

```bash theme={null}
veri volumes rm my-corpus /data/old.jsonl
```

## Pipe-friendly output

`--quiet` strips formatting and emits just the relevant id:

```bash theme={null}
VOL_ID=$(veri volumes create my-corpus --quiet)
echo $VOL_ID    # vol_abc123
```

`--format json|jsonl|csv` swaps the table renderer for machine-readable output, and the progress bar on uploads is suppressed.

```bash theme={null}
veri volumes list --format json > volumes.json
```

## Common workflows

### Re-use a volume across many training jobs

```bash theme={null}
veri volumes create gsm8k
veri volumes upload gsm8k ./gsm8k-train.jsonl --to /train.jsonl
```

Then in your script:

```python theme={null}
dataset = client.datasets.from_volume(volume="gsm8k", path="/train.jsonl")
client.training_jobs.create(..., dataset_id=dataset.id)
```

Every subsequent training run skips the upload and reuses the same S3 prefix.

### Migrate from a local directory

```bash theme={null}
veri volumes create research-corpus
veri volumes upload-dir research-corpus ./local-corpus --to /
veri volumes ls research-corpus    # confirm the file tree mirrors local
```

### Inspect failed uploads

If `volumes upload-dir` aborts mid-batch, files that completed are already on the volume — you can resume by re-running the same command. S3 PUT is idempotent on the same key.

```bash theme={null}
veri volumes ls my-corpus                 # see what made it
veri volumes upload-dir my-corpus ./data  # re-run; existing files are skipped
```

## Next steps

<CardGroup cols={2}>
  <Card title="Volumes overview" icon="hard-drive" href="/volumes">
    What volumes are, pricing, when to use them.
  </Card>

  <Card title="REST API" icon="globe" href="/api-reference/introduction">
    `/v1/volumes` endpoints + multipart upload protocol.
  </Card>
</CardGroup>