Contribute

1. Cite everything. Every record needs at least one source URL and the date you read it. A pull request without sources fails the build automatically — this is not a style preference, it is enforced in CI.
2. Omit what you cannot confirm. Never fill a field from memory or inference. A blank field renders as unknown, which is useful. A guessed field is worse than nothing, because it looks like knowledge.

1. 1. Create data/<category>/<id>.yaml.
2. 2. Fill it in using the template below. Leave verified: null unless you personally checked every field against a source.
3. 3. Run npm test. Validation errors name the exact file and field.
4. 4. Open a pull request.

id: openmrs                 # kebab-case, must equal the filename
name: OpenMRS
category: software          # therapeutics | devices | software | protocols | data
subcategory: ehr            # optional
summary: Modular EHR platform for low-resource settings.   # <= 200 chars

license: [MPL-2.0]          # SPDX ids; [NOASSERTION] if genuinely unclear
license_class: osi          # osi | open-hardware | open-data | cc | mixed | unknown
homepage: https://openmrs.org
repo: https://github.com/openmrs/openmrs-core   # optional

status: active              # active | maintained | dormant | archived | unknown
maturity: deployed          # concept | prototype | deployed | regulatory-cleared

tags: [ehr, global-health]
related: [openemr]          # ids of other entries

sources:                    # REQUIRED, at least one
  - url: https://openmrs.org/about
    retrieved: 2026-08-04
    via: manual
    note: What you actually confirmed from this source.
verified: null              # a date only if YOU confirmed the record

Telling us what is missing. A registry that claims to be complete and is not is worse than one that knows its own gaps. If a whole class of projects is absent — a field, a country, a kind of hardware that does not live in a Git repository — that is a flaw in how we gather data, not a missing row. Open an issue and say so.