Kaggle API dataset list pagination and query bounding

Answer

The --page argument is used to request the next page of results when calling kaggle datasets list. The API does not expose a configurable page‑size option; the default page size is fixed (observed as ~100 items per request) and cannot be changed via the CLI.

Confirmed facts

• The CLI command accepts --page N where N is an integer starting at 1.
• Each response includes a JSON array of dataset objects and a page field indicating the current page number.
• Search keywords (--search) or tags (--tag) are sent as query parameters; pagination state is maintained server‑side, so incrementing --page continues the filtered result set without needing a cursor or token.
• Empirical testing shows a default of approximately 100 results per page; the CLI does not provide a --page-size flag to alter this.

Steps to paginate a bounded query

1. Run the initial request with your search or tag filter and page 1:

kaggle datasets list --search "machine learning" --page 1

2. Inspect the JSON output for the page field and the number of items returned.
3. If more results are needed, increment the page number and repeat:

kaggle datasets list --search "machine learning" --page 2

4. Continue until the returned array is empty or shorter than the default page size, indicating the last page.

Likely explanation / uncertainty

The API appears to use simple offset‑based pagination driven by the --page parameter. Because the server keeps the filter context, each successive --page value yields the next slice of the filtered list. The lack of a --page-size flag means users cannot reduce the number of requests by increasing page size; they are limited to the server‑defined default. If the default page size changes in a future API version, the pagination behavior would shift accordingly, but no documentation currently specifies a maximum configurable size.

To verify the exact page size for your environment, you would need to count the items in a single page response (e.g., using jq length) and compare across pages. If you observe a different size, please share that count so the recommendation can be adjusted.