crowdsky-client

A lightweight Python client for the CrowdSky data API — discover and download crowd-sourced Seestar S50 stacked frames. Built as the shared contract between CrowdSky, the scientists writing CrowdSci analysis modules, and the CrowdSci runner on Zeus, so a pipeline behaves identically on a laptop and in production.

Deliberately light: it depends only on requests, astropy, and astropy-healpix — not the stacking engine (opencv/astroalign/sep).

Install

pip install crowdsky-client

Authenticate

Mint a personal API key at https://crowdsky.univie.ac.at/account.php (log in → Account → Create key). The secret is shown once — store it safely. Pass it directly or via the CROWDSKY_API_KEY environment variable.

from crowdsky_client import Client

c = Client(api_key="csk_…")                     # or: export CROWDSKY_API_KEY=csk_…; Client()

Use

# All frames imaging a target (cone-search over NSIDE=128 RING tiles, deduped by id):
frames = c.frames_for_target(159.6998, 53.5095, radius_deg=0.75, filter_name="IRCUT")

# Frames in a single HEALPix tile with server-side filters:
frames = c.frames_by_healpix(49152, start="2026-07-01T00:00:00Z", min_exptime=120)

# Download a stacked FITS by its unique id (bytes, or straight to disk):
fits_bytes = c.download_frame(frames[0]["id"])
c.download_frame_to(frames[0]["id"], "frame.fits")

# A frame's SEP + Gaia star table:
stars = c.star_data(frames[0]["id"])

# Public sky-coverage feed (no key needed) as an astropy Table:
coverage = c.sky_coverage()

Every data request sends Authorization: Bearer <api_key>; a revoked/expired/invalid key returns HTTP 401, surfaced as CrowdSkyError.

Develop / test

pip install -e ".[dev]"
pytest

Tests run fully offline via an injected fake session — no network, no key required.