?
Sample a source
Draws a random sample of a source's documents into a new source, optionally stratified by metadata.
admin.dataSourceEvals.sample({ slug, size, as?, filter?, stratify?, placement?, wait? })
Copies a random draw of a source's documents into a new, smaller corpus, so you can develop and evaluate against a representative slice instead of the whole thing. --stratify keeps the draw proportional across a metadata key, and --filter restricts which documents are eligible.
Parameters
--source
string
RequiredCorpus to draw from.
--size
number
RequiredHow many documents to draw.
--as
slug
Slug for the new sampled corpus.
--filter
k=v,...|json
Restrict which documents are eligible.
--stratify
metadata-key
Keep the draw proportional across this metadata key.
CLI
remy-admin datasources sample --source news --size 200 --as news-sample --stratify section