데이터셋

데이터셋과 테이블을 관리하고 파일 업로드·다운로드와 스키마 추론을 수행하는 API입니다.

📄️ List Datasets

List all datasets. Only returns datasets the user has permission to access.

📄️ Create Dataset

Create a new dataset. Schema is required for table-type datasets.

📄️ Get Dataset Versions

List metadata/resource version history for a specific dataset.

List ontology entity and relation types that use the specified dataset. Read permission on the dataset is required; results are not additionally filtered by permissions on the returned ontology resources.

📄️ Get Dataset

Retrieve a specific dataset. If a version is specified, return that version.

📄️ Update Dataset

Update a dataset. Its name cannot be changed after creation, and it cannot be detached from its collection.

📄️ Delete Dataset

Delete a dataset. Related data cleanup runs asynchronously.

📄️ Get Cleanup Status

Retrieve the cleanup progress status after dataset deletion.

📄️ List Objects In Dataset

List files in an object-type dataset. Returns all entries in one page — `next_cursor`/`has_more` are fixed (ListResponse shape); no pagination params are accepted.

📄️ Get Object Versions

List version history for a specific object.

📄️ Get Object

Download the binary data of a specific object.

📄️ Update Object

Replace the file data of an object.

📄️ Delete Object From Dataset

Remove the specified stored object from a dataset while retaining the dataset and its other objects.

📄️ Infer Schema from Files

Infer an Arrow schema, optionally validating CSV against a dataset.

📄️ Get Table Versions

List the newest `limit` versions of a Delta Lake table, newest first. The bound is what keeps this route answerable. `history` with no limit materializes every commit: measured on hub-dev, a table written once a minute reached 31,562 commits and 37.85s for one call (0.20s at `limit=100`), returning 11.3 MB of JSON. The call also holds the GIL for its whole duration — other Python threads made 2.8% of their normal progress during `history(limit=1000)` — so a threadpool would not spare the event loop; only a smaller window does.

📄️ Get Table

Download table data in CSV, JSON, Parquet, or Arrow format. `limit` caps the number of returned rows. When `limit` is omitted, the full table is returned.

📄️ Upload To Dataset

Upload raw files to object datasets (omit ``format``); otherwise parse CSV, JSON or Parquet, falling back to binary for unrecognized extensions.

📄️ Generate Mock Data

Generate schema-conformant sample (mock) rows and write them to the dataset's Delta table. Delta-only; ``append`` or ``overwrite``.

📄️ Preview Mock Data

Generate and return sample rows without writing them to the dataset. Read permission is required. The dataset must use Delta storage, must not use change-data-capture ingestion, and must have a non-placeholder schema. `limit` must be between 1 and 200. The response includes the first `limit` rows and the seed; use the same seed in a write request to reproduce the previewed data.

📄️ Clear Dataset Table

Empty the dataset's Delta table in place. Gated on ``write`` like ``upload?mode=overwrite``, which already replaces every row. A Delta ``DELETE`` commit — schema, id, permissions, and history stay, and the prior rows remain readable via ``?version=N`` until vacuum. Delta-only.