API documentation

Load any licensed dataset in three lines of Python.

Quickstart

Install the SDK:

pip install corpora

Then load a dataset you have licensed:

import corpora
corpora.api_key = "sk_live_..."
ds = corpora.load("pit-radio-1998")
print(ds.schema)

All requests require an API key passed in the Authorization header. Keys are created in your workspace settings.

curl https://api.corpora.ai/v1/datasets \
  -H "Authorization: Bearer sk_live_..."

GET /v1/datasets?q=<query> returns datasets matching your query, including ones you have not licensed.

results = corpora.search("legal text", fmt="Text")
for d in results:
    print(d.id, d.price_chf, d.status)

Large datasets are streamed in batches. Batches are shuffled and de-duplicated, and names are removed where detected.

for batch in ds.stream(size=512, shuffle=True):
    model.train(batch)

Every response includes a licence object. Check licence.status before training.

Status	Meaning
`commercial`	Cleared for training and production.
`research_only`	Non-commercial use. Honour system.
`pending`	Seller is obtaining rights. Usable in the meantime.
`unknown`	Status unknown to all parties.

Explorer: 60 requests per minute. Team: 6,000. Enterprise: unlimited, subject to available data.

Code	Meaning
`401`	Invalid API key.
`402`	Licence required for this dataset.
`410`	Dataset withdrawn. A similar one may be relisted shortly.
`451`	Unavailable for legal reasons in your jurisdiction. Try another region.