xrtm-data 0.1.0

The Snapshot Vault for XRTM.

xrtm-data

The Snapshot Vault for XRTM.

xrtm-data provides the rigid schemas and temporal sandboxing infrastructure required for zero-leakage forecasting. It defines the "Ground Truth" data structures that the rest of the ecosystem (Forecast, Eval, Train) relies on.

Installation

uv pip install xrtm-data

Core Primitives

1. The Forecast Object Standard

Adhering to strict Governance v1, the ForecastOutput schema mandates that every prediction be accompanied by a structured causal graph (logical_trace) and a calibrated confidence interval.

from xrtm.data import ForecastOutput, CausalNode

prediction = ForecastOutput(
    question_id="q_123",
    probability=0.75,
    reasoning="Base rate analysis suggests...",
    logical_trace=[
        CausalNode(event="Inflation rises", probability=0.8),
        CausalNode(event="Fed cuts rates", probability=0.4)
    ]
)

2. Zero Leakage

The MetadataBase enforces a strict snapshot_time. This timestamp represents the "End of History" for the model. Any data point generated after this time is considered "Future Leakage" and is programmatically inaccessible during backtesting.

Development

Prerequisites:

• uv

# Install dependencies
uv sync

# Run tests
uv run pytest