Overview

Warning

Pre-release / design spec. This section defines proposed non-API contracts for a framework still under design. Treat these pages as implementation blueprints until code and contract tests land.

Audience. Implementers, plugin authors, release engineers, data consumers, and anyone who needs normative contracts for config fields, schemas, algorithms, operations, or on-disk formats.

Before you start. Familiarity with Concepts is helpful. Reference pages assume you understand the six-layer pipeline.

What this section is not. This section is not tutorial content. For step-by-step guides, see How-to guides. For conceptual explanations, see Concepts.

---

The normative non-API surface of the framework. Implementations must match what’s documented here, or this page should change first. Python classes, ABCs, and method signatures live in the root-level API Reference.

Organized into six clusters:

Schemas

Normative serialized contracts: configuration, labels, storage, annotation templates, plugin metadata.

• Schemas Overview

• Config Schema

• Label Schema

• Storage Layout

• Annotation Templates

• Plugin Metadata

Data Formats

Physical data layout, tensor conventions, and external interoperability.

• Data Formats Overview

• IQ Layout Policy

• TorchSig Interop

• Scene Geometry Assets

Algorithms & Math

How computation works: synthesis, impairments, metrics, seed flow.

• Algorithms Overview

• Emitter Algorithms

• Scene Composition Algorithm

• Fingerprint Math

• Fingerprint Parameter Priors

• RF Frontend Models

• 3GPP TR 38.901 propagation tables

• Noise Floor Table

Operations

How to run it: compute, packaging, tests.

• Operations Overview

• CLI

• Compute Environment

• Cloud Backends

• Logging and Observability

• Build and CI

Quality Contracts

Reproducibility, audits, metrics, and test contracts.

• Quality Overview

• Determinism

• Metrics

• Test Contracts

Project Contracts

Repository structure and implementation readiness.

• Project Overview

• Project Layout

• Maturity Matrix

Reading paths

Different audiences read the Reference in different orders:

Audience	Path
Pre-implementation reviewer	API Reference → Schemas → Data Formats → Algorithms & Math
Implementer (writing code)	Project Layout → API Reference (module ABCs) → relevant Algorithms page → Test Contracts
Downstream data consumer	Storage Layout → Label Schema → Annotation Templates
Operator deploying v0	Compute Environment → CLI → Build and CI
Release engineer	Build and CI → Test Contracts → Metrics (release gates)
Researcher reproducing a run	Determinism → Storage Layout → CLI