Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
// GET /api/fact-check-benchmark: public read of the fact-check accuracy benchmark.
//
// Serves the most recently published run (score, per-class table, per-difficulty
// table, confusion matrix, run date), plus the static claim-count summary from
// tests/fixtures/fact-check-benchmark.json so the /fact-check page always has
// something honest to render.
//
// Two sources, in order:
// 1. The DB row written by `scripts/fact-check-benchmark.mjs --publish` and by
// the scheduled re-run (api/cron/fact-check-benchmark.js). This wins so a
// new run reaches the page immediately instead of waiting for a deploy,
// which is what let the published number go stale before.
// 2. data/_generated/fact-check-benchmark.json, the run committed into the
// image. It is the fallback when nothing is published or the DB is
// unreachable, so a database blip degrades to the last shipped run rather
// than to an empty page.
// `source` names which one answered.
//
// Never fabricates a score: when neither source has a run, `ran` is false and
// `report` is null, and the page renders its designed "not yet run" empty state
// instead of a fake number.
import { readFile } from 'node:fs/promises';
import { cors, json, method, wrap } from './_lib/http.js';
import {
FIXTURE_PATH,
REPORT_PATH,
VERDICT_CLASSES,
readPublishedRun,
} from './_lib/fact-check-benchmark.js';
async function readJsonIfExists(path) {
try {
return JSON.parse(await readFile(path, 'utf8'));
} catch (err) {
if (err?.code === 'ENOENT') return null;
throw err;
}
}
function summarizeFixture(fixture) {
if (!fixture || !Array.isArray(fixture.claims)) return null;
const counts = Object.fromEntries(VERDICT_CLASSES.map((c) => [c, 0]));
for (const claim of fixture.claims) {
if (counts[claim.expected_verdict] != null) counts[claim.expected_verdict]++;
}
return { total: fixture.claims.length, by_class: counts, version: fixture.version || null };
}
export default wrap(async (req, res) => {
if (cors(req, res, { methods: 'GET,OPTIONS', origins: '*' })) return;
if (!method(req, res, ['GET'])) return;
const [fixture, published, shipped] = await Promise.all([
readJsonIfExists(FIXTURE_PATH),
readPublishedRun(),
readJsonIfExists(REPORT_PATH),
]);
const report = published || shipped || null;
const source = published ? 'database' : shipped ? 'image' : null;
return json(
res,
200,
{
data: {
fixture: summarizeFixture(fixture),
ran: Boolean(report),
source,
report,
claims_source:
'https://github.com/nirholas/three.ws/blob/main/tests/fixtures/fact-check-benchmark.json',
runner_source:
'https://github.com/nirholas/three.ws/blob/main/scripts/fact-check-benchmark.mjs',
},
},
// Short cache: a fresh publish (manual or scheduled) should surface within
// minutes, and the payload is small enough that 5 minutes at the edge is
// all the protection this read needs.
{ 'cache-control': 'public, max-age=300' },
);
});