Skip to content

Automation: RSS → Daily Digest

📖 3 min read resourcesautomations
Pull new items from your RSS feeds each morning, summarize and group them with an LLM, and deliver one tidy digest to email or Slack.

Stop doom-scrolling feeds. Each morning, this pulls new items from your RSS sources, has an LLM summarize and group them by theme, and sends you one digest — 2 minutes of reading instead of 20 tabs.

How it works

flowchart LR
A[Schedule<br/>daily 7am] --> B[Read RSS feeds<br/>last 24h items]
B --> C[LLM summarizes +<br/>groups by theme]
C --> D[Send digest<br/>Email / Slack]

You’ll need

  • A few RSS feed URLs (blogs, newsletters, arXiv, Hacker News, a subreddit’s .rss)
  • An LLM key — Flash / Haiku / DeepSeek Flash (runs once a day over many short items)
  • Email (SMTP/Gmail) or Slack

Build it in n8n

  1. Schedule Trigger → daily at your time.
  2. RSS Read node (one per feed, or loop a list) → fetch items.
  3. Filter node → keep items where isoDate is within the last 24h.
  4. Aggregate node → combine titles + links + snippets into one text blob.
  5. HTTP Request → the AI transform node, input = the aggregated blob, system prompt below, max_tokens: 1200.
  6. Code node → parse the JSON.
  7. Send Email / Slack node → format themes into a readable digest (a Code node can build simple HTML/markdown).

The AI prompt

System prompt:

You build a daily reading digest from a list of articles (title, link, snippet).
Return ONLY valid JSON, no prose:
{
"tldr": "<2-sentence overview of today's themes>",
"themes": [
{
"theme": "<short theme name>",
"items": [
{"title": "<title>", "url": "<link>", "why": "<one line: why it matters>"}
]
}
]
}
Rules:
- Group related items under a shared theme; drop near-duplicates.
- Keep "why" to one concrete sentence. Don't editorialize or invent facts.

Zapier alternative

Schedule by Zapier (daily) → RSS by Zapier → New items in feedFormatter → Text (combine) → Anthropic → Send PromptFormatter → Parse JSONEmail/Slack with the formatted digest. (Zapier’s RSS trigger fires per-item; collect with a Digest step or run per feed.)

Cost

10–20 short items ≈ 3–5K input + ~800 output tokens once a day. On Gemini 3.5 Flash (1.50/1.50/9 per 1M) that’s well under **0.10/day( 0.10/day** (~2–3/month).

Variations

  • Weekly instead of daily for slower feeds.
  • Add a relevance filter: tell the model your interests and have it drop off-topic items.
  • Pipe the digest into Notebook/Notion as a dated note instead of email.