NEWVectors or files. Pick a path.Start →

    Random Sampling

    Randomly sample results for testing, analysis, or reducing result set size

    Why do anything?

    Large result sets need sampling for analysis or display. Random sampling provides representative subsets.

    Why now?

    Testing and analytics workflows need sampling capabilities.

    Why this feature?

    Random sampling with configurable sample size and optional seed for reproducibility.

    How It Works

    Sampling provides random document selection from result sets.

    1

    Pool Formation

    Collect candidate documents

    2

    Random Selection

    Randomly select sample_size documents

    3

    Stratification

    Optional stratified sampling by field

    Why This Approach

    Random sampling ensures representative subsets. Seed enables reproducibility.

    Integration

    retriever = client.retrievers.create(stages=[{"type": "sample", "sample_size": 100}])