Python API Reference

RootCause SDK: causal discovery, digital twins, and ontology queries from Python.

Two modes, one object model:

import rootcause as rc
rc.login()

ws = rc.workspace("Calix Forecasting")            # platform mode
twin = ws.twin("C8 Temporal")
twin.forecast(horizon=24).to_frame()

graph = rc.discover(df, target="re78")            # direct mode — no workspace ceremony
twin = graph.train()
twin.intervene({"treat": rc.set(1)}, where={"re75": ("<", 5000)})

Conventions throughout: every long-running call blocks with a progress line and raises JobFailedError on failure; everything tabular answers to_frame(); names resolve case-insensitively, with close-match suggestions on a miss.

Every failure is a RootCauseError carrying one readable sentence and the thing to try next — arguments are checked before a request goes out, and no call raises an httpx, pyarrow or KeyError traceback of its own.

• Module functions: The module-level session and the direct-mode entry points.

• Workspaces and data: Workspaces and what lives in them: sources, data views, connectors.

• Twin: Digital twins: forecast, simulate, intervene, score, update.

• Graph: Discovered causal graphs, and the domain knowledge pinned onto them.

• Results: The result objects twin operations hand back.

• Ontology: Ontology concepts and queries over them.

• Interventions: The intervention and metric helpers.

• Notebook apps: Notebook host for the platform's interactive app bundles. Requires the jupyter extra.

• Exceptions: The exception hierarchy every call raises from.