Loading the page.
In development
The current graph is the first computational substrate for human biology. A bulk export will make its entities and relationships available as a file rather than one request at a time: 1,385 findings and the 117,642 entities they connect, with every provenance column intact.
The columns that make a finding checkable, not a flattened summary of it. An export that dropped the provenance would be the thing this platform exists to be distinguishable from.
This is the part that makes an export slower to build than an API, and it is worth stating before anyone downloads anything rather than after.
ADR-0039. A source whose terms do not permit redistribution is not ingested into a table that implies it can be — which means an export does not need a licence audit at download time, because the audit already happened.
Several reference databases behind the graph are CC BY and their attribution is a licence condition rather than a courtesy. Curated findings are this project’s own work. An export separates them rather than merging them into one file with one notice.
A bounded release that does not say it is bounded reads as the whole graph. Where a source is excluded the manifest names it and says why, in the same way every figure on this site names what it left out.
Every upstream source and its terms are listed on the attribution page, which is the authority rather than this summary of it.
The same rows, one question at a time, for an agent or an application.
What the `evidence_strength` column means, and every rule that can move it.
What the platform refuses to do, and how far along each goal actually is.