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
// Resolve one news article by its canonical (month, id16) address — the
// lookup behind the story pages (/markets/news/<YYYY-MM>/<id>) and their
// sitemap. Live-feed cache first for recent months (richer record: image,
// author, feed-provided full text), then the archive month file, which covers
// the full corpus back to September 2017 with a single GCS fetch.
import { findArticle } from './news.js';
import { loadMonth } from './news-archive-store.js';
const MONTH_RE = /^20\d{2}-(0[1-9]|1[0-2])$/;
const ID_RE = /^[a-f0-9]{16}$/;
export function validStoryKey(month, id) {
return MONTH_RE.test(String(month || '')) && ID_RE.test(String(id || ''));
}
function recentMonths() {
// Previous, current AND next month: publishers have shipped future-dated
// stories (the aggregator now clamps those, but already-minted links carry
// the bogus month), so a next-month URL must still reach the live lookup.
const now = new Date();
const cur = now.toISOString().slice(0, 7);
const shift = (n) => {
const d = new Date(now);
d.setUTCDate(1);
d.setUTCMonth(d.getUTCMonth() + n);
return d.toISOString().slice(0, 7);
};
return [cur, shift(-1), shift(1)];
}
function adjacentMonths(month) {
const d = new Date(`${month}-01T00:00:00Z`);
const shift = (n) => {
const c = new Date(d);
c.setUTCMonth(c.getUTCMonth() + n);
return c.toISOString().slice(0, 7);
};
return [shift(-1), shift(1)];
}
/**
* @returns {Promise<{article: object, origin: 'live'|'archive'} | null>}
* The live record wins for recent months (it carries image/author and, for
* some publishers, the feed's full body); everything older comes from the
* archive month file. Ids are content-addressed (sha of the publisher link),
* so a hit is always the right story — but its month can drift by one when a
* publisher revises pub_date across a boundary or a link was minted from a
* differently-parsed date. The adjacent-month fallback recovers those; the
* story-page handler 301s to the canonical path when the months differ.
*/
export async function resolveStory(month, id) {
if (!validStoryKey(month, id)) return null;
const inMonthArchive = async () => {
const records = await loadMonth(month).catch(() => null);
return records?.find((a) => a.id === id) || null;
};
if (recentMonths().includes(month)) {
const live = await findArticle({ id }).catch(() => null);
if (live) {
if ((live.pub_date || '').slice(0, 7) === month) return { article: live, origin: 'live' };
// The live record disagrees with the requested month. If the archive
// has the story filed under the REQUESTED month, that durable record
// wins — redirecting toward a live-feed month here while the archive
// redirects back would loop the two URLs forever (seen in prod when a
// publisher's bogus future date met a hand-corrected archive record).
const archived = await inMonthArchive();
if (archived) return { article: archived, origin: 'archive' };
return { article: live, origin: 'live' };
}
}
const archived = await inMonthArchive();
if (archived) return { article: archived, origin: 'archive' };
for (const near of adjacentMonths(month)) {
const nearRecords = await loadMonth(near).catch(() => null);
const hit = nearRecords?.find((a) => a.id === id) || null;
if (hit) return { article: hit, origin: 'archive' };
}
return null;
}