Method. Sources were trusted in this order: primary repository contents; project
documentation; GitHub API metadata for maturity and governance only, never adoption; published papers
for design description only; vendor marketing to locate candidates, discarded at assessment.
Licences, activity and capabilities are as verified in the August 2026 sweep.
Repository activity — last push, primary language, archive flag — was read from
the GitHub API on 16 August 2026 for the 133 rows that name a GitHub repository, and drives the
freshness dot and the maintained/stalled split. It measures the repository, not the project: a push
includes documentation and bot commits, and for datasets, ontologies and specifications a quiet
repository is not the same as an unmaintained resource. The survey's own Activity assessment is kept
beside it on every card rather than replaced by it. Three organisation-level URLs are reported from
their newest repository, and one URL (ontoportal-lirmm) no longer resolves.
What to take from this band is interpretation, not measurement: it is derived from the
survey's findings against one specific decision — upgrading a Zotero → DuckDB → MCP claim base
for agronomy and soil-microbiome literature. Read it as an argument, and the cards as the evidence.
“Not documented” means the project's own repository and documentation do not establish it
— not that it is absent. Self-reported benchmark numbers are named as self-reported and are not
repeated as fact here. Abandoned repositories were kept deliberately: their abandonment is evidence
about the field.
Soilytix internal research artifact · generated from the vendor note of 9 August 2026.