Risk-Weighted Assets

risk-weighted-assets is an institution-neutral, auditable Python reference engine for CRR III risk-weighted assets, regulatory capital, IRRBB and ICAAP. It combines a stable Python API, a command-line interface, a local browser application, canonical Excel contracts, realistic synthetic portfolios and reproducible reference calculations.

The distribution is intended for research, education, prototyping and independent model validation. It is not regulatory, legal, accounting or investment advice and is not a certified regulatory reporting system. Read the disclaimer before use.

Functional scope

• CRR III credit risk under SA and IRB, credit-risk mitigation and output-floor views
• counterparty credit risk, SFT, CCP, securitisation, CVA and crypto exposures
• settlement, large exposures, market risk with parallel FRTB views, and operational risk
• own funds, buffers, leverage, MREL/TLAC and capital headroom
• IRRBB/CSRBB and ICAAP economic and normative perspectives
• deterministic fingerprints, lineage, reconciliations and calculation controls
• 16 canonical input and six output workbooks per persisted calculation

Installation

Python 3.10 or newer is required. Install a locally built wheel with:

python -m pip install dist/risk_weighted_assets-1.0.0-py3-none-any.whl

For development from this source tree:

python -m pip install -e '.[test,build]'

Once the package has completed the separate publication process, the regular PyPI command will be python -m pip install risk-weighted-assets.

First calculation

Create an isolated, writable workspace from the immutable package resources and run the supplied universal-bank dataset:

rwa init ./rwa-workspace
rwa run --dataset ./rwa-workspace/daten/rechenlaeufe/2026-08-31/v1.0.0

Or generate a deterministic synthetic dataset and calculate it in one step:

rwa all \
  --data-root ./rwa-workspace/daten/rechenlaeufe \
  --as-of-date 2026-08-31 \
  --version v1.0.1 \
  --seed 5752026 \
  --bank-profile MID_SIZE_UNIVERSAL

Set RWA_WORKSPACE=/absolute/path to change the default workspace used by the CLI and web application. Package resources are never modified in place.

Python API

from pathlib import Path

from rwa_engine import calculate_dataset, create_workspace, validate_dataset

workspace = create_workspace(Path("rwa-workspace"))
dataset = workspace.runs_root / "2026-08-31" / "v1.0.0"

validation = validate_dataset(dataset)
if validation.valid:
    result = calculate_dataset(dataset)
    print(result.status, result.run_id, result.metrics)

For integration without Excel I/O, use calculate_tables(tables) with the canonical dict[str, pandas.DataFrame] contract. It returns the same structured CalculationResult without persisting output workbooks.

CLI and local app

rwa --help
rwa doctor
rwa data list
rwa sources
rwa-web --data-root ./rwa-workspace/daten/rechenlaeufe

The web server binds to 127.0.0.1:8080 by default and invokes the same calculation pipeline as the API and CLI. It has no authentication or TLS and is intended for a trusted local workstation only.

Bundled data and source boundary

The wheel includes both complete synthetic profiles (MID_SIZE_UNIVERSAL and KSA_BANK), their two calculation-ready datasets, retained reference outputs, configuration and the machine-readable inventory of official sources. Use rwa data export PATH or create_workspace(PATH) to obtain editable copies.

No downloaded regulation, standard, PDF or other third-party publication is redistributed. Official links and archival checksums are available through rwa sources and the regulatory source catalogue. The original construction scaffold is not part of this distribution.

Documentation and verification

• documentation map
• public API
• installation and troubleshooting
• architecture
• methodology
• governance, controls and acceptance
• package provenance

Run the complete local verification suite with:

python -m pytest
python -m ruff check src tests tools
python -m build
python -m twine check dist/*
python tools/validate_distribution.py dist

Legal and privacy

This project is published by RiskDataScience GmbH. The legally binding company information is available in the imprint, and information about the processing of personal data is provided in the privacy policy.

The installed library contains no telemetry, analytics or tracking and does not transmit portfolio, calculation or usage data to RiskDataScience GmbH. The optional browser application communicates only with the locally started RWA server. Interactions performed on GitHub or PyPI are additionally subject to the terms and privacy practices of those platforms.

License

Copyright © 2026 RiskDataScience GmbH. Original content is licensed under the GNU General Public License, version 3 only. External publications remain subject to their respective rights and are not included. See third-party notices.