.. |SWR| image:: https://img.shields.io/badge/SWR--25--62_-blue?label=NLR :alt: Static Badge

.. |Zenodo| image:: https://zenodo.org/badge/DOI/10.5281/zenodo.17173409.svg :target: https://doi.org/10.5281/zenodo.17173409

.. inclusion-intro

What is COMPASS?

COMPASS (formally the Infrastructure Continuous Ordinance Mapping for Planning and Siting Systems, or "INFRA-COMPASS") is an innovative software tool that harnesses the power of Large Language Models (LLMs) to automate the compilation and continued maintenance of an inventory of state and local codes and ordinances pertaining to energy infrastructure.

At a high level, COMPASS does two things: it retrieves the right ordinance documents for each jurisdiction you ask about, and then extracts structured data from those documents into a versioned database that downstream users can query as a CSV, Excel workbook, or GeoPackage.

.. image:: https://raw.githubusercontent.com/NatLabRockies/COMPASS/main/docs/source/_static/overview.png :alt: High-level overview of the COMPASS pipeline :align: center :width: 100%

What makes COMPASS different from simply asking ChatGPT for ordinance data is the architecture around the LLM call:

• Structured, downstream-ready output — consistent CSV rows with stable column names, units, and feature labels that drop straight into siting and capacity-modeling tools like reV <https://github.com/NREL/reV>_, GIS workflows, or any pipeline that needs setbacks, height limits, and noise thresholds as numbers rather than prose.
• Hallucination guardrails — cleaned text is checked against the source and dropped if it drifts too far, so fabricated values never reach the database.
• Source-URL traceability — every record carries a URL back to the original ordinance document, so any value can be audited or spot-checked.
• Cost control — cheap heuristic filters reject obviously irrelevant text before any LLM call runs, making it tractable to extract data across hundreds of jurisdictions.

Read more about the tool in the documentation <https://natlabrockies.github.io/COMPASS/misc/about.html>, check out summaries of the validation results in the validation report <https://natlabrockies.github.io/COMPASS/val/validation.html>, or explore the data on the COMPASS website <https://compass.nlr.gov/>_.

Where is the extracted ordinance data?

The National Laboratories of the Rockies (NLR) typically runs the COMPASS pipeline annually and publishes refreshed datasets to OpenEI. The latest published ordinance datasets are available here:

• Geothermal Electricity: https://data.openei.org/submissions/8817
• Geothermal Heat Pumps: https://data.openei.org/submissions/8816
• Solar: https://data.openei.org/submissions/8602
• Wind: https://data.openei.org/submissions/8519

Installing COMPASS

The quickest way to install COMPASS for users is from PyPI:

.. code-block:: bash

pip install infra-compass

If you would like to install and run COMPASS from source, we recommend using pixi <https://pixi.sh/latest/>_:

.. code-block:: bash

git clone [email protected]:NatLabRockies/COMPASS.git; cd COMPASS
pixi run compass

For detailed instructions and troubleshooting, see the installation documentation <https://natlabrockies.github.io/COMPASS/misc/installation.html>_.

Quickstart

To run a quick COMPASS demo, set up a personal OpenAI API key and run:

.. code-block:: shell

pixi run openai-solar-demo <your API key>

This will run a full extraction pipeline for two counties using gpt-4o-mini (costs ~$0.45). For more information on configuring a COMPASS run, see the execution basics example <https://natlabrockies.github.io/COMPASS/examples/execution_basics/README.html>_.

Development

Please see the Development Guidelines <https://natlabrockies.github.io/COMPASS/dev/index.html>_ if you wish to contribute code to this repository.