modaic datasets
upload
Upload a file and wait for validation.
modaic datasets upload acme train.jsonl --kind dataset --name support-labels --repo support-model
| Flag / argument | What it does |
|---|---|
WORKSPACE | Required workspace slug. |
PATH | Required local file path. |
--kind | Required: dataset, eval-only-dataset, or artifact. |
--name | Required unique dataset name. |
--description | Optional dataset description. |
--repo | Repository name to link; repeat to link multiple repositories. |
--timeout | Validation wait in seconds. |
-h, --help | Show usage and options. |
Check the returned status before using the dataset. For other files, use --kind artifact. See dataset format.
list
List uploaded datasets.
modaic datasets list acme --kind dataset --glob 'support-*'
| Flag / argument | What it does |
|---|---|
WORKSPACE | Required workspace slug. |
--commit-sha | Read a specific dataset revision. |
--kind | Filter by dataset, eval-only-dataset, or artifact. |
--glob | Filter names with * and ? wildcards. |
--cursor | Next-page cursor. Pass the previous response commit with --commit-sha. |
-h, --help | Show usage and options. |
get
Read dataset metadata.
modaic datasets get acme support-labels
| Flag / argument | What it does |
|---|---|
WORKSPACE | Required workspace slug. |
NAME | Required uploaded dataset name. |
--commit-sha | Read a specific dataset revision. |
-h, --help | Show usage and options. |
download
Save dataset bytes to a local file.
modaic datasets download acme support-labels --output train-copy.jsonl
| Flag / argument | What it does |
|---|---|
WORKSPACE | Required workspace slug. |
NAME | Required uploaded dataset name. |
--commit-sha | Read a specific dataset revision. |
--output | Output path. Defaults to the dataset name. |
-h, --help | Show usage and options. |
Dataset identity and pagination
Names are identifiers, not paths: 1–128 lowercase letters, digits, underscores, or hyphens, starting with a letter or digit. Names are unique across datasets eval-only datasets, and artifacts. Use a new name for each upload.
Uploads stream file parts and wait for asynchronous validation. Check completed status and a non-null result before using the dataset name. File size limit is 8 GiB. A successful transfer is not proof of successful dataset validation. For resumable part-level operations use the SDK; the CLI upload command starts a new session.
For another list page, pass both nextCursor and commitSha from the previous response:
modaic datasets list acme --cursor NEXT_CURSOR --commit-sha COMMIT_SHA
Downloads write to --output, or the dataset name if omitted; choose an output path that will not overwrite a file the user needs.
delete
Permanently delete a dataset, including its retained versions and large-file parts. Content still used by another dataset is retained for that dataset. Queued, running, and paused jobs using this dataset block deletion with HTTP 409.
modaic datasets delete acme support-labels
The command waits for completion by default. --no-wait returns the deletion
request immediately; --timeout SECONDS limits the wait without cancelling the
cleanup. Reads are revoked immediately and physical cleanup follows after
existing transfers have finished or expired. This is irreversible.