Skip to content

Datasets CLI

Use provon datasets to manage project datasets and their examples.

View as Markdown Open the plain-text version of this page.

These commands require a project-scoped credential.

List Datasets#

bash
provon datasets list
provon datasets list --limit 50 --cursor <cursor>

Create A Dataset#

bash
provon datasets create "My Dataset" \
  --description "Successful conversations for fine-tuning" \
  --objective preserve_successful_behavior

Available collection objectives: preserve_successful_behavior, capture_failures_for_review, improve_specific_skills.

Use --collection-enabled or --no-collection-enabled to control automatic collection.

Inspect And Update A Dataset#

bash
provon datasets get <dataset-id>
provon datasets update <dataset-id> --name "New Name"
provon datasets update <dataset-id> --description "New description"
provon datasets update <dataset-id> --objective capture_failures_for_review
provon datasets update <dataset-id> --collection-enabled false

Delete A Dataset#

bash
provon datasets delete <dataset-id>

Manage Examples#

List examples:

bash
provon datasets examples list <dataset-id>
provon datasets examples list <dataset-id> --limit 100 --cursor <cursor>

Create an example from a JSON payload file:

bash
provon datasets examples create <dataset-id> \
  --payload-file ./example.json \
  --source-kind manual \
  --tags review,priority

Source kinds: manual, conversation, import, synthetic.

Inspect and update an example:

bash
provon datasets examples get <dataset-id> <example-id>
provon datasets examples update <dataset-id> <example-id> \
  --payload-file ./updated-example.json \
  --tags reviewed

Delete an example:

bash
provon datasets examples delete <dataset-id> <example-id>