Architecture

XRTM is a 4-layer forecasting stack plus a product shell.

The Layers

Layer 4: xrtm-train      Backtesting & optimization
         ↓
Layer 3: xrtm-forecast    Runtime engine (agents, providers, topologies)
         ↓
Layer 2: xrtm-eval        Scoring (Brier, ECE, log score)
         ↓
Layer 1: xrtm-data        Schemas & question sources

Dependencies flow downward. A layer can import from layers below it, never from above.

Framework vs Product

	Framework	Product
Packages	data, eval, forecast, train	xrtm
Audience	Developers building custom systems	Users running forecasts
Interface	Python API	CLI
Install	pip install xrtm-forecast	pip install xrtm

The framework provides composable building blocks. The product composes them into a turnkey CLI.

Repository Map

Repo	PyPI	GitHub
xrtm-data	xrtm-data	xrtm-org/data
xrtm-eval	xrtm-eval	xrtm-org/eval
xrtm-forecast	xrtm-forecast	xrtm-org/forecast
xrtm-train	xrtm-train	xrtm-org/train
xrtm	xrtm	xrtm-org/xrtm

Governance

Architecture standards, policies, and forecast object schemas live in xrtm-org/governance.