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

# Croissant

> Import MLCommons Croissant dataset metadata as a queryable markdown page.

[MLCommons Croissant](https://mlcommons.org/croissant/) describes a dataset — not rows. KiwiFS imports one Croissant document as one page: the dataset description lands in frontmatter, and the column schema lands in `kiwi-data` blocks so you can query it with DQL.

## Import

<Tabs>
  <Tab title="URL">
    ```bash theme={null}
    kiwifs import --from croissant --url https://www.kaggle.com/datasets/OWNER/NAME/croissant/download --root ./knowledge
    ```
  </Tab>

  <Tab title="File">
    ```bash theme={null}
    kiwifs import --from croissant --file croissant.json --root ./knowledge
    ```
  </Tab>
</Tabs>

## Query the schema

```sql theme={null}
TABLE name, dtype, column, source-file
FROM RECORDS "dataset-schema"
WHERE dtype = "float"
```

<Note>
  Croissant is JSON-LD. The importer expands the document to absolute IRIs before reading it, so two emitters can use different property names for the same dataset. Contexts resolve offline: an inline `@context` (what Kaggle and Hugging Face emit) and the well-known MLCommons context URL both work with no network. Any other remote context is refused rather than fetched.
</Note>

## Related

<CardGroup cols={2}>
  <Card title="Import overview" icon="download" href="/import/overview">
    All supported sources.
  </Card>

  <Card title="Widgets" icon="puzzle-piece" href="/concepts/widgets">
    `kiwi-data` blocks and `FROM RECORDS`.
  </Card>
</CardGroup>