/** @jsxImportSource npm:react **/ import process from "node:process"; import { renderToString } from "npm:react-dom@18/server"; import { QueryVectorStore, sb, Substrate } from "npm:substrate"; type ShotResult = { id: string; distance: number; metadata: { doc: string; // base64 img data doc_id: string; }; }; async function getResults(q: string, n: number): Promise { const substrate = new Substrate({ apiKey: process.env.SUBSTRATE_API_KEY, }); const collectionName = "shotclip"; const query = new QueryVectorStore({ collection_name: collectionName, model: "clip", query_strings: [q], include_metadata: true, top_k: n, }, { cache_age: 60 * 60 }); const res = await substrate.run(query); return res.get(query).results[0] as ShotResult[]; } export default async (req: Request) => { const searchParams = new URL(req.url).searchParams; const { prompt = "flowers", n = 12 } = Object.fromEntries(searchParams); const res = await getResults(prompt, Number(n)); const pageWidth = 1200; const aspectRatio = 16 / 9; const widthNoPadding = pageWidth - 40 - 10; const columns = 2; const imgWidth = widthNoPadding / columns; const imgHeight = imgWidth / aspectRatio; const fontFamily = "Monaco, monospace"; return new Response( renderToString(

Shotclip

[made with substrate]

{res.map(s => (

))}

, ), { headers: { "Content-Type": "text/html" } }, ); };