AI Content PipelineAI Workflow

“Run `npx radzor@latest add rss-feed ai-classifier prompt-template llm-completion guardrails` to install 5 Radzor components. Then read components/radzor/rss-feed/radzor.manifest.json, components/radzor/ai-classifier/radzor.manifest.json, components/radzor/prompt-template/radzor.manifest.json, components/radzor/llm-completion/radzor.manifest.json, components/radzor/guardrails/radzor.manifest.json and each component's llm/integration.md. Wire them together to ingest RSS feeds, classify articles by topic with AI, render prompts from templates, generate summaries with an LLM, and validate outputs with guardrails. A full content processing chain. Use the manifest's inputs (check envVar for required environment variables), outputs (check fields for object shapes), composability (check mapField for field extraction), and actions — don't invent custom interfaces.”

Paste this into Claude Code, Cursor, Windsurf, or any AI coding agent.

// npx radzor@latest add rss-feed ai-classifier prompt-template llm-completion guardrails
import { RssFeed }        from "./components/radzor/rss-feed"
import { AiClassifier }   from "./components/radzor/ai-classifier"
import { PromptTemplate } from "./components/radzor/prompt-template"
import { LLMCompletion }  from "./components/radzor/llm-completion"
import { Guardrails }     from "./components/radzor/guardrails"

const feed = new RssFeed({ defaultPollIntervalMs: 300_000 })
const classifier = new AiClassifier({ apiKey: process.env.OPENAI_KEY!, mode: "llm" })
const prompts = new PromptTemplate({})
const llm = new LLMCompletion({ provider: "openai", apiKey: process.env.OPENAI_KEY!, model: "gpt-4o", maxTokens: 512 })
const guard = new Guardrails({ enableBuiltinRules: true, maxOutputLength: 2000 })

// Define categories
classifier.defineCategory("tech", "Technology and software articles")
classifier.defineCategory("business", "Business and finance articles")
classifier.defineCategory("science", "Science and research articles")

// Register prompt template
prompts.register("summarize", `Summarize the following {{category}} article in 3 bullet points.

Title: {{title}}
Content: {{content}}

Be concise and factual.`)

// Process new RSS items
feed.on("onNewItem", async (item) => {
  // 1. Classify the article
  const { category } = await classifier.classify(item.title + " " + (item.description ?? ""))

  // 2. Render the prompt
  const { text: prompt } = prompts.render("summarize", {
    category,
    title: item.title,
    content: item.description ?? "",
  })

  // 3. Generate summary
  const { content: summary } = await llm.complete(prompt)

  // 4. Validate output
  const validation = guard.validateOutput(summary)
  if (!validation.passed) {
    console.warn("Guardrails blocked:", validation.violations)
    return
  }

  console.log(`[${category}] ${item.title}\n${summary}`)
})

// Start watching
await feed.watch("https://hnrss.org/newest?points=100")