carp-cli

Read Copenhagen Research Platform study data from Python.

This is the same client the carp command line uses, exposed as a module, so the two share a session: sign in once and either can use it.

pip install carp-cli
pip install 'carp-cli[pandas]'   # adds to_pandas()

Installed as carp-cli, imported as carp. PyPI's carp was claimed in 2012 by an unrelated package that has not been touched since; a distribution name has never had to match what you import.

import carp

client = carp.Client(env="production")
client.login()                       # opens a browser, once

for study in client.studies():
    print(study["studyId"], study["name"])

Getting data out

data_stream returns one dictionary per measurement, with the stream it came from folded into each — the shape a DataFrame wants:

rows = client.data_stream(
    deployment="df98d925-3ab4-4b78-8139-fea86d809dc5",
    device="Primary Phone",
    data_type="dk.cachet.carp.heartrate",
    start="7d",
)

frame = carp.to_pandas(rows)         # pip install 'carp-cli[pandas]'

Windows are written the way you would say them — "7d", "36h", "2026-08-01", "2026-08-01T09:30:00Z", or None for now.

For the bulk of a study, ask for an export and download the archive:

client.create_export(STUDY)                       # packaged in the background
[export] = [e for e in client.exports(STUDY) if e["status"] == "AVAILABLE"]
path = client.download_export(STUDY, export["id"])

Sessions

The session lives in the same file carp auth login writes, keyed by the deployment's host. So:

• carp auth login --env test in a terminal, then carp.Client(env="test") in a notebook, needs no second login
• client.login() in a notebook signs the terminal in too
• each deployment keeps its own session; moving between them signs you out of neither

Exceptions

Building it

pip install maturin
maturin develop            # into the active virtualenv
maturin build --release    # a wheel, in target/wheels/