Snapshot
All ten board documents in exactly the shape app/data/ holds them: the model board, the GPU board, and the eight meter boards.
GET /api/v1/snapshotcurl https://vitna.ai/api/v1/snapshotResponse
{
"modelMarket": { "meta": { "unit": "USD per 1M tokens" }, "data": [ ... ] },
"gpuMarket": { "meta": { "unit": "USD per GPU-hour" }, "products": [ ... ], "providers": [ ... ] },
"toolMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"retrievalMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"extractionMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"imageMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"videoMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"musicMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"translationMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"speechMarket": { "meta": { "units": [ ... ] }, "data": [ ... ] },
"meta": { "generatedAt": "...", "source": "database", "units": [ ... ] }
}What it is for
The daily mirror. scripts/export-snapshot.mjs fetches this and overwrites the JSON files under app/data/ verbatim.
That keeps two logs at two resolutions: git is the daily one, with a commit per day and a diff you can read; the database is the hourly one. Neither is derived from the other after the fact.
node scripts/export-snapshot.mjs https://vitna.aiBoard shape, not tape shape
These are the documents the site renders, so they keep board structure. The GPU market is a matrix: a products array, and a providers array whose cells line up with it positionally, with null where a provider does not rent that card.
If you want rows rather than boards, use `/markets/{class}/tape`, which flattens all of this.
meta.units is an array
A meter board can be metered several ways (the tool board is), so each one contributes all of its units rather than one. A single unit per board would have to be wrong for most of the tool board's rows.
Parameters
None.
Caching
public, max-age=60, stale-while-revalidate=3600.
