The ax datasets commands are currently in BETA. The API may change without notice. A one-time warning is emitted on first use.
ax datasets commands let you create, retrieve, and manage datasets on the Arize platform. Supported file formats: CSV, JSON, JSON Lines, Parquet.
ax datasets list
List all datasets in a space.
ax datasets list [--space <id>] [--name <filter>] [--limit <n>] [--cursor <cursor>]
| Option | Description |
|---|
--space | Filter by space name or ID |
--name | Case-insensitive substring filter on dataset name |
--limit | Maximum number of results to return (default: 15) |
--cursor | Pagination cursor for the next page |
ax datasets list --space sp_abc123
ax datasets list --space sp_abc123 --output datasets.json
ax datasets create
Create a new dataset from a local file.
ax datasets create --name <name> --space <id> (--file <path> | --json <json-array>)
| Option | Description |
|---|
--name | Name for the new dataset |
--space | Space name or ID to create the dataset in |
--file | Path to the data file (CSV, JSON, JSONL, or Parquet), or - for stdin |
--json | Inline JSON array of examples, e.g. '[{"question": "...", "answer": "..."}]' |
ax datasets create --name "my-eval-set" --space sp_abc123 --file ./examples.csv
ax datasets create --name "my-eval-set" --space my-space --json '[{"q": "What is AI?"}]'
ax datasets get
Retrieve a dataset by name or ID.
ax datasets get <name-or-id> [--space <id>]
| Option | Description |
|---|
--space | Space name or ID (required when using a dataset name instead of ID) |
ax datasets get ds_xyz789
ax datasets get my-eval-set --space my-space
ax datasets update
Rename a dataset.
ax datasets update <name-or-id> --name <new-name> [--space <id>]
| Option | Description |
|---|
--name | New name for the dataset |
--space | Space name or ID (required when using a dataset name instead of ID) |
ax datasets update ds_xyz789 --name "renamed-eval-set"
ax datasets update my-eval-set --space my-space --name "renamed-eval-set"
ax datasets delete
Delete a dataset.
ax datasets delete <name-or-id> [--space <id>] [--force]
| Option | Description |
|---|
--space | Space name or ID (required when using a dataset name instead of ID) |
--force | Skip the confirmation prompt |
ax datasets delete ds_xyz789
ax datasets delete my-eval-set --space my-space --force
ax datasets export
Export examples from a dataset to a JSON file.
ax datasets export <name-or-id> [--space <id>] [--version-id <id>] [--output-dir <path>] [--stdout] [--all]
| Option | Description |
|---|
--space | Space name or ID (required when using a dataset name instead of ID) |
--version-id | Export examples from a specific dataset version (default: latest) |
--output-dir | Directory to write the output file (default: current directory) |
--stdout | Print JSON to stdout instead of saving to a file |
--all | Use Arrow Flight for bulk export — streams all examples |
ax datasets export ds_xyz789
ax datasets export my-eval-set --space my-space --version-id v_123 --output-dir ./exports
ax datasets export ds_xyz789 --all
ax datasets export ds_xyz789 --stdout | jq 'length'
ax datasets append
Append examples to an existing dataset. Provide examples via --json (inline) or --file. Exactly one input source is required.
ax datasets append <name-or-id> (--json <json-array> | --file <path>) [--space <id>] [--version-id <id>]
| Option | Description |
|---|
--json | Inline JSON array of examples, e.g. '[{"question": "...", "answer": "..."}]' |
--file | Path to the data file (CSV, JSON, JSONL, or Parquet), or - for stdin |
--space | Space name or ID (required when using a dataset name instead of ID) |
--version-id | Dataset version to append to (default: latest version) |
# Append from a file
ax datasets append ds_xyz789 --file ./new_examples.csv
# Append inline JSON
ax datasets append ds_xyz789 --json '[{"question": "What is AI?", "answer": "..."}]'
# Append to a specific version
ax datasets append my-eval-set --space my-space --file ./more.jsonl --version-id v_123
ax datasets annotate-examples
Annotate a batch of examples in a dataset. Provide annotations via --file (JSON, JSONL, CSV, or Parquet; use - for stdin). Each record must have a record_id (the dataset example ID) and values (a list of annotation dicts with at least name, plus optionally score, label, or text). Annotations are upserted; up to 1000 examples may be annotated per request.
ax datasets annotate-examples <name-or-id> --file <path> [--space <id>]
| Option | Description |
|---|
--file | Path to a file containing annotation records (JSON, JSONL, CSV, Parquet), or - for stdin |
--space | Space name or ID (required when using a dataset name instead of ID) |
ax datasets annotate-examples ds_xyz789 --file ./annotations.jsonl
ax datasets annotate-examples my-eval-set --space my-space --file ./annotations.json
