Load Pools CLI

A command-line tool for uploading HuggingFace datasets and GitHub repositories to Load S3 and tagging them for use with Load's data pools.

Features

• GitHub repository upload: Clone and upload entire repos with file-by-file tagging
• HuggingFace integration: Upload datasets by path
• Automatic tagging: Data-Protocol, Path, Filename, and Content-Type to make data discoverable
• Query support: All uploads are queryable via s3-agent's tag query API

Installation

Prerequisites

A Load Network account API key from cloud.load.network. HuggingFace calls may ask for a HuggingFace API key, freely available for registered users here

Usage

Upload a GitHub Repository

Upload all files from a GitHub repository with proper folder structure tagging:

load-pools create --github https://github.com/owner/repo --auth YOUR_LOAD_API_KEY

What happens:

1. Repository is cloned to a temporary directory
2. Each file is uploaded individually to s3-agent
3. Files are tagged with:
   • Data-Protocol: "owner/repo"
   • Path: "folder/subfolder" (relative path from repo root)
   • Filename: "file.ext"
   • Content-Type: "mime/type"

Example tags for file images/grayscale/9582.png:

[
  {"key": "Data-Protocol", "value": "owner/repo"},
  {"key": "Path", "value": "images/grayscale"},
  {"key": "Filename", "value": "9582.png"},
  {"key": "Content-Type", "value": "image/png"}
]

Upload a HuggingFace Dataset

Upload a HuggingFace dataset. Tables are uploaded row by row, with proper tagging.

load-pools create --hugging-face username/dataset-name --auth YOUR_LOAD_API_KEY 

For private datasets or to bypass anonymous rate limits, pass --hf-auth <YOUR_TOKEN>.

What happens:

1. HuggingFace dataset is downloaded
2. Table rows are extracted into individual dataitems
3. Rows are uploaded to s3-agent with metadata tags

Command Options

load-pools create [OPTIONS]

Options:
  --github TEXT          GitHub repository URL
  --hugging-face TEXT    HuggingFace repository slug (user/repo)
  --auth TEXT            Load account API key [required]
  -v, --verbose          Show detailed upload progress
  --help                 Show help message

Querying Uploaded Data

After uploading, you can query your data using the s3-agent tags API.

Query all files from a GitHub repository:

curl -X POST https://load-s3-agent.load.network/tags/query \
  -H "Content-Type: application/json" \
  -d '{
    "filters": [
      {"key": "Data-Protocol", "value": "owner/repo"}
    ]
  }'

Query files in a specific folder:

curl -X POST https://load-s3-agent.load.network/tags/query \
  -H "Content-Type: application/json" \
  -d '{
    "filters": [
      {"key": "Data-Protocol", "value": "owner/repo"},
      {"key": "Path", "value": "images/grayscale"}
    ]
  }'

Query by filename:

curl -X POST https://load-s3-agent.load.network/tags/query \
  -H "Content-Type: application/json" \
  -d '{
    "filters": [
      {"key": "Data-Protocol", "value": "owner/repo"},
      {"key": "Filename", "value": "9582.png"}
    ]
  }'

Query by content type:

curl -X POST https://load-s3-agent.load.network/tags/query \
  -H "Content-Type: application/json" \
  -d '{
    "filters": [
      {"key": "Data-Protocol", "value": "owner/repo"},
      {"key": "Content-Type", "value": "image/png"}
    ]
  }'

Examples

Upload a dataset repository:

load-pools create \
  --github https://github.com/username/my-dataset \
  --auth load_acc_xxxxxxxxxxxxx

Upload with verbose output:

load-pools create \
  --github https://github.com/ml-datasets/images \
  --auth load_acc_xxxxxxxxxxxxx \
  --verbose

Upload a HuggingFace dataset:

load-pools create \
  --hugging-face openai/graphwalks \
  --auth load_acc_xxxxxxxxxxxxx

License

MIT License - see LICENSE file for details.

Related Projects

• xans104 - HuggingFace model uploader with ANS-104
• Load Network - Decentralized data storage network
• s3-agent - Load S3 Agent API