Text and tabular modalities

Text and tabular projects sit alongside computer vision in the same product.

Project types

Type	Modality	Annotate
text_classification	text	Class buttons
ner	text	Character span list
tabular_classification	tabular	Class buttons on rows
tabular_regression	tabular	Numeric target value

Create these from New Project. Document/OCR remains planned.

Import

• Text: POST /api/projects/{id}/import/text — JSONL ({"text","label"?,"spans"?}) or CSV

• Tabular: POST /api/projects/{id}/import/tabular — CSV with header + optional label / numeric column

UI: project → Import.

Samples API

Method	Path
GET/POST	/api/projects/{id}/samples
GET/PATCH	/api/samples/{id}
POST	/api/projects/{id}/samples/splits

PATCH body can set class_id / class_name, spans, or value.

Export

POST /api/projects/{id}/export with format: jsonl (text) or csv (tabular).

Canonical export (GET /api/projects/{id}/canonical) returns modality-aware samples.

Training adapters

Optional installs:

pip install "oneopen-ml-studio[tabular]"   # scikit-learn
pip install "oneopen-ml-studio[nlp]"       # transformers + datasets

• POST /api/projects/{id}/adapters/prepare — write prepared CSV / HF JSONL

• POST /api/projects/{id}/adapters/train — sklearn LogisticRegression / Ridge; Hugging Face returns prepared artifacts (full Trainer not in-process)

YOLO /train and YOLO/COCO export return 400 on text/tabular projects.