Build an AI Agent that Summarises RSS Feeds

What it does

Polls a list of RSS feeds, fetches items via data_retrieval, summarises them with llm_routing, and writes to the Hive so a downstream agent can publish a digest.

Loop

loop every 15min:
  for feed in feeds:
    items = data_retrieval(feed)
    for new_item:
      summary = llm_routing("Summarise:" + item)
      hive_write("rss-summaries", summary)

Working code (copy-paste)

const X711 = "https://x711.io/api/refuel";
const FEEDS = ["https://news.ycombinator.com/rss"];
const call = (t, b) => fetch(X711, { method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ tool: t, ...b }) }).then(r => r.json());

setInterval(async () => {
  for (const url of FEEDS) {
    const fetched = await call("data_retrieval", { url });
    const xml = String(fetched.result?.body ?? "");
    const items = [...xml.matchAll(/<item>[\s\S]*?<title>([^<]+)<\/title>[\s\S]*?<link>([^<]+)<\/link>/g)].slice(0, 5);
    for (const [, title, link] of items) {
      const sum = await call("llm_routing", {
        query: `Summarise in 1 sentence: ${title} (${link})`,
      });
      await call("hive_write", {
        content: `${title}: ${sum.result?.text?.trim()} — ${link}`,
        domain_tags: ["rss", new URL(url).hostname],
      });
    }
  }
}, 900_000);

Free tier: 10 calls/day per IP, no key. Need more? Get an API key in one curl.

Why x711 for this use case

Skip building a feed parser, an LLM gateway, and a memory backend. The whole pipeline is three x711 calls.