Build a daily tech digest with n8n, RSS and OpenAI

What you'll build: an automated workflow that aggregates 16 RSS sources every morning (Reddit, Hacker News, newsletters, FR/EN blogs), summarizes the news in your language with OpenAI, and sends it to your phone via Gotify.

Level: intermediate — you should already know how to create a basic n8n workflow.

Prerequisites

• A self-hosted n8n instance (v2.30+)
• An OpenAI API key
• A Gotify server (or replace with Email/Slack/Telegram)

Architecture

Schedule Trigger (7:00 AM)
  → 16 HTTP Requests in parallel
  → Merge Sources
  → Code node "Parse & Filter"
  → Aggregate
  → OpenAI
  → Gotify

Step 1 — Schedule Trigger

Set the trigger to cron expression 0 7 * * * to run every day at 7:00 AM. Make sure to configure the timezone in the workflow Settings.

Step 2 — HTTP Request nodes

Create one HTTP Request node per source, all connected to the Schedule Trigger — they run in parallel. Enable continueOnFail: true on each so the workflow keeps going if a source is unavailable.

Reddit natively supports multi-subreddit feeds using + in the URL — one HTTP Request covers all of them. Add a User-Agent: Mozilla/5.0 (compatible; n8n-digest/1.0) header to avoid rate limits.

TLDR has no official RSS feed — the open-source project bullrich/tldr-rss generates separate feeds per newsletter.

Step 3 — Merge Sources

Add a Merge node in Append mode and connect all HTTP Request nodes to it.

Enable waitForAllInputs: true — without this, the Merge fires as soon as the first 2 branches complete and ignores the rest.

⚠️ When testing manually, n8n doesn't always trigger all branches in parallel. This option works correctly on scheduled runs.

Step 4 — Code node "Parse & Filter"

This node receives the raw XML from each source and extracts relevant articles. It:

• Automatically detects the source (Reddit, HN, TLDR, etc.)
• Supports both RSS and Atom formats
• Extracts title, URL, date and description from each entry
• Filters out old articles — 48h window on weekdays, 72h on Mondays to cover the weekend
• Deduplicates by URL
• Limits to 5 articles per source
• Returns an empty array if nothing to report — the workflow stops cleanly without calling the LLM

Step 5 — Aggregate

Add an Aggregate node set to Aggregate All Item Data with items as the destination field. This bundles all articles into a single object for prompt construction.

Step 6 — OpenAI

Add an OpenAI node. The prompt is built dynamically in the User Message field via an n8n expression. It:

• Separates French sources (JDG, Numériques, Korben) from English ones
• Includes the title and description of each article for context
• Requests a digest structured into thematic sections: AI, Ruby/Rails, iOS/Android, Self-hosted, DevOps, Healthtech, Tech/Startups, FR News
• Enforces pure Markdown output, no HTML

Step 7 — Gotify

Add a Gotify node:
- Message: ={{ $json.output[0].content[0].text }}
- Title: 🗞️ Tech Digest — Yesterday
- Priority: 7

⚠️ Do not add contentType: text/markdown in the options — this parameter doesn't exist in Gotify node v1 and silently crashes the workflow.

Common pitfalls

Going further

• Add a source: one more HTTP Request node connected to the Schedule Trigger and Merge
• Change digest sections: edit the section list in the OpenAI prompt
• Avoid duplicate articles across runs: use $workflow.staticData to persist seen URLs
• Replace Gotify: Email, Slack, Telegram — the last node is interchangeable