Datasphere-API 0.1.0

Unofficial async Python client for the SAP Datasphere automation APIs.

Datasphere-API

Unofficial async Python client for the SAP Datasphere automation APIs. This library powers the SAP-Datasphere-CLI and can be used to build your own automations (e.g. MCP servers or scheduled jobs).

[!NOTE] This project is not affiliated with, endorsed by, or supported by SAP. It uses the same internal HTTP endpoints as the SAP Datasphere web UI, which may change without notice.

Features

• Views: export view analytics (persistence candidates), search views by attribute, persist/unpersist views, create/remove partitions, lock/unlock partitions.
• Remote Tables: list tables with statistics information, create/update statistics (Record Count / Simple / Histogram), refresh statistics.
• Task Chains: run task chains and wait for their completion.
• Analytical Models: export models with all their views (optionally per space), measure view persistence runtimes.
• OAuth login included: interactive authorization code flow via a real Chrome/Edge window (Playwright) with automatic token refresh and a shared token cache.

Installation

uv add datasphere-api
# or
pip install datasphere-api

The interactive login drives an installed Chrome or Edge browser via Playwright channels — a regular Chrome/Edge installation is required, no playwright install download is needed.

Quickstart

import asyncio

from datasphere_api import DatasphereClient, DatasphereConfig


async def main() -> None:
    config = DatasphereConfig(
        base_url="https://example.eu10.hcs.cloud.sap",
        authorization_url=(
            "https://example.authentication.eu10.hana.ondemand.com"
            "/oauth/authorize"
        ),
        token_url=(
            "https://example.authentication.eu10.hana.ondemand.com"
            "/oauth/token"
        ),
        client_id="...",
        client_secret="...",
    )
    client = DatasphereClient(config)
    try:
        await client.login()
        results = await client.task_chains.run(
            chains=[{"entity": "MY_CHAIN", "space": "MY_SPACE"}],
        )
        print(results)
    finally:
        await client.aclose()


asyncio.run(main())

The URLs and credentials can be found in your tenant under System > Administration (Tenant Links and App Integration). The OAuth client has to be of type "Interactive Usage" with the redirect URI http://localhost:8080.

Authentication

client.login() first tries to refresh cached tokens from the token store (session.json in the user data directory of Datasphere). If no tokens are cached or the refresh fails, a browser window opens for the interactive login. All consumers of this library share the same token cache, so a login in one tool also benefits the others.

Layered results

The library returns data on two levels:

• Curated results (recommended): high-level operations like views.persist_views() or remote_tables.create_statistics() return small, typed result structures (see datasphere_api.models). Their keys intentionally match the CSV/JSON exports of the CLI.
• Raw payloads: low-level fetchers like views.get_all_views() or remote_tables.get_all_tables() return the parsed API payload, typed with broad TypedDicts. For anything not covered, client.session is the authenticated httpx.AsyncClient — you can call any endpoint directly.

Long-running batch operations accept an on_result/on_update callback that is invoked after every finished item, so consumers can save intermediate results during runs that take hours.

Development

uv sync
uv run pytest
uv run ruff check .
uv run pyright

License

MIT