Introduction (Coming Soon)

​

What are Datasets?

Datasets in Basalt will provide a structured way to store, manage, and utilize collections of data for your AI workflows. This upcoming feature will enable you to:

• Store representative examples of your inputs and expected outputs
• Create standardized test sets for evaluating your AI systems
• Run experiments against consistent data
• Compare different approaches using the same baseline data

​

Planned Capabilities

When the Datasets feature is released, you’ll be able to:

​

Push Data to Datasets

// Add examples to your dataset
await basalt.datasets.add('customer-queries', [
  { input: 'How do I reset my password?', category: 'account' },
  { input: 'When will my order arrive?', category: 'orders' }
])

​

Use Datasets in Experiments

// Create an experiment
const experiment = await basalt.monitor.createExperiment('query-classification')

// Run your workflow against a dataset
await basalt.datasets.runExperiment({
  datasetSlug: 'customer-queries',
  experiment: experiment,
  workflow: async (item) => {
    // Your workflow code here, using the dataset item
    return classifyQuery(item.input)
  }
})

​

Retrieve Dataset Items

// Get items from your dataset
const items = await basalt.datasets.get('customer-queries', {
  limit: 10,
  shuffle: true
})

​

Benefits of Datasets

The upcoming Datasets feature will help you:

• Ensure Consistency: Test on the same data over time
• Improve Reproducibility: Get reliable comparisons between different approaches
• Streamline Testing: Automate experimentation with real-world examples
• Track Progress: Measure improvements against a consistent benchmark

​

Stay Tuned

We’re excited to bring Datasets to Basalt soon! This feature will complement our existing monitoring and experimentation capabilities, creating a more comprehensive platform for developing, testing, and optimizing your AI applications.

Follow our release announcements to be notified when Datasets become available.