pysepal-api 0.2.0

UI-free HTTP client for SEPAL platform services (user files, tasks, processing recipes).

pysepal-api

UI-free HTTP client for SEPAL platform services. Powers Jupyter notebooks, Voila apps, CLI scripts, and background jobs running inside a SEPAL sandbox.

from pysepal_api import SepalClient

# sync — notebooks, CLI, jobs
with SepalClient(session_id="...") as sepal:
    listing = sepal.files.list("/")
    task = sepal.tasks.submit("image.download", params={...})
    sepal.tasks.wait(task.id)

    # escape hatch: any route the typed endpoints don't model
    image = sepal.get("/api/gee/image/json", params={"recipeId": "foo"}).json()

from pysepal_api import AsyncSepalClient

# async — Voila / Solara
async with AsyncSepalClient(session_id="...") as sepal:
    listing = await sepal.files.list("/")
    image = (await sepal.get("/api/gee/image/json", params={"recipeId": "foo"})).json()

The two clients are twins: identical surface, the only difference is await.

• Scoped use — with SepalClient(...) / async with AsyncSepalClient(...). Construction never performs network I/O; entering the context creates the module results dir (when module_name is given).
• Long-lived clients — sepal = SepalClient.create(...) / sepal = await AsyncSepalClient.create(...): same eager setup without a context manager; call close() / aclose() when done. async with AsyncSepalClient.create(...) also works.
• Typed endpoints for the common cases: files, tasks, recipes. File verbs follow pathlib: read_bytes / read_text / read_json / write / mkdir.
• Generic primitive — request() / get() / post() / put() / delete() — for any other SEPAL route, with the same typed-error mapping (NotFound, Unauthorized, …). Returns the raw httpx.Response.
• One exception root — catch PysepalError for anything this library raises.

TLS is verified by default. host.docker.internal is auto-skipped for local dev; for any other self-signed host pass verify=False or list it in PYSEPAL_INSECURE_TLS_HOSTS (comma-separated, exact-host match).