cvPal
cvPal
Dashboard

Manage credentials
and connection settings

Tool Reference

External Sources

Discover, evaluate, and import datasets from the three largest computer vision platforms through a unified interface. Search by keyword, inspect metadata before downloading, and ingest directly into your cvPal workspace. All three tools accept a source parameter to select the provider.

RoboflowRoboflow

Browse your workspace projects and download specific versions in any supported format.

KaggleKaggle

Access the world's largest data science community. Supports pagination, size filtering, and credential-based auth.

Hugging FaceHugging Face

Search by tags, sort by popularity, and use streaming ingest for datasets of any size.

search_external_datasets

Instant

Search or list datasets from an external source. Supports Hugging Face, Kaggle, and Roboflow. If query is omitted, lists the authenticated user's own datasets/projects on that platform.

Parameters

sourcestring
"huggingface", "kaggle", or "roboflow".
Required
querystring
Search term. Omit to list your own datasets.
Optional
limitint
Max results 1–100. Defaults to 20.
Optional
authorstring
(Hugging Face) Filter by author/org.
Optional
workspacestring
(Roboflow) Workspace slug override.
Optional
pageint
(Kaggle) Page number for pagination.
Optional
sortstring
(Hugging Face) Sort by "downloads", "likes", "updated", or "created".
Optional
filter_tagslist[string]
(Hugging Face) Filter by tags, e.g. ["image-classification"].
Optional
max_sizeint
(Kaggle) Max dataset size in bytes.
Optional

Returns

List of dataset objects with name, description, size, download count, and source-specific metadata.

Example prompt

“Search Kaggle for face mask detection datasets”

inspect_external_dataset

Instant

Inspect an external dataset without downloading it. Returns metadata, file listing, total size, and format hints. Useful for evaluating a dataset before committing to an ingest.

Parameters

sourcestring
"huggingface" or "kaggle".
Required
dataset_namestring
Dataset identifier (e.g. "rajpurkar/squad").
Required
versionstring
Version/revision. Defaults to latest.
Optional

Returns

Dataset metadata including file list, total size, format detection, license, and last update date.

Example prompt

“Inspect the "andrewmvd/face-mask-detection" dataset on Kaggle before downloading”

ingest_dataset

Async Job

Download a dataset from an external source into cvPal. Submits a serverless job that downloads, processes, hashes, and stores the dataset in content-addressable storage.

Parameters

sourcestring
"huggingface", "kaggle", or "roboflow".
Required
dataset_namestring
Dataset identifier (format depends on source).
Required
versionstring
Version override.
Optional
workspacestring
(Roboflow) Workspace slug override.
Optional
formatstring
(Roboflow) Export format. Defaults to "yolov8".
Optional
splitstring
(Hugging Face) Specific split, e.g. "train".
Optional
streamingboolean
(Hugging Face) Use streaming mode for large datasets.
Optional

Returns

Job ID for tracking via get_processing_status. When complete, the dataset is available in your cvPal workspace.

Example prompt

“Download the face-mask-detection dataset from Kaggle into my workspace”

Typical workflow

1

Search

Find datasets matching your criteria across any provider.

“Search Hugging Face for aerial drone datasets with at least 1000 downloads”

2

Inspect

Preview metadata, file structure, and size before downloading.

“Inspect that first result — show me its size and format”

3

Ingest

Download and import the dataset into your cvPal workspace.

“Download it into my workspace”