Daily Multi-Source Tech Digest with AI Ranking + Slack Delivery (n8n)

12 nodes, 3 parallel data sources, one AI summarization step, one Slack Block Kit payload — wired and saved on n8n cloud. Import, swap the placeholder Slack webhook URL (Step 6), and schedule. Execution screenshots below show the data flowing through the AI summarizer.

⚠️ Before activating: the Post to Slack Webhook node ships with a placeholder URL (https://hooks.slack.com/services/REPLACE/WITH/YOUR_WEBHOOK). The first scheduled run will fail with a 404 if you don't replace it. See Step 6 for the exact replacement procedure.

What You'll Build

A daily-trigger n8n workflow that:

• Runs every morning at 8 AM (UTC by default, configurable)
• Fetches top 5 items in parallel from three sources: Hacker News front page, Dev.to top articles, GitHub repos created in the last 48 hours sorted by stars
• Normalizes each shape into a common { source, title, url, score, summary_seed } object
• Uses a LangChain Summarization Chain to write a one-sentence summary per item
• Builds a Slack Block Kit message with grouped sections, clickable links, and dividers
• POSTs to your Slack Incoming Webhook — no OAuth needed

Skip the Build — Import the Workflow

Prerequisites

Step 1 — Import the Workflow

Create a new workflow, paste the JSON onto the blank canvas. n8n loads 12 nodes. It looks like this:

Schedule ──┬─→ HN Fetch   ─→ Normalize HN
           ├─→ Dev.to     ─→ Normalize Dev.to    ──┐
           └─→ GitHub     ─→ Normalize GitHub    ──┼─→ Merge → Summarize Each → Combine → Format Slack → POST
                                                  ──┘

Double-click OpenAI Chat Model, confirm the n8n free credits credential and gpt-5-mini with temperature 0.3 (low — we want consistent one-sentence output).

Step 2 — Understand the Three Sources

The GitHub URL uses an n8n expression: {{ $now.minus({days: 2}).toFormat('yyyy-MM-dd') }} — this builds a date string for "two days ago" at runtime so you always see the freshest repos.

Step 3 — Normalize Each Source

Three Code nodes convert each source into the same object shape. For example, Normalize HN:

const hits = $input.first().json.hits || [];
return hits.map(h => ({
  json: {
    source: 'Hacker News',
    title: h.title,
    url: h.url || `https://news.ycombinator.com/item?id=${h.objectID}`,
    score: h.points || 0,
    summary_seed: (h.story_text || '').slice(0, 400)
  }
}));

The summary_seed gives the AI some context for items that lack a URL body (Show HN, Ask HN). The fallback URL (news.ycombinator.com/item?id=...) ensures Ask HN items are still clickable.

Normalize Dev.to reads description into summary_seed. Normalize GitHub reads the repo's one-line description.

After this step, all 15 items (5 × 3) have identical structure — no downstream node has to know which source they came from.

Step 4 — Merge and Summarize

The Merge Node

Set to mode: append. It takes 3 input streams (the three Normalize nodes) and outputs one combined list of 15 items in source order.

Summarization Chain

The chain runs once per item. The prompt is inlined into the node's combineMapPrompt:

⚠️ Important syntax difference: The Summarization Chain prompt uses LangChain's {text} template variable, NOT n8n's {{ $json.x }} expressions. The default chain feeds each input item's content as {text} into the prompt automatically. If you need access to other fields ($json.title, $json.url), the cleanest fix is to swap Summarization Chain for Basic LLM Chain (chainLlm) — that node DOES substitute n8n expressions like Guide 1 does. We hit this in our live test: with {{ $json.title }} literally in the Summarization Chain prompt, the model returned its own template-style placeholders rather than a real summary. Use Basic LLM Chain unless you specifically need map-reduce summarization over very long documents.

Source: {{ $json.source }}
Title: {{ $json.title }}
URL: {{ $json.url }}
Score/engagement: {{ $json.score }}
Context: "{{ $json.summary_seed }}"

Write ONE crisp sentence (≤ 25 words) that tells a busy developer what this
is and why it matters TODAY. Lead with the concrete thing, not with "This
story is about...". If it is a GitHub repo, mention the language or primary
domain. No hedging.

Each item gets one API call to gpt-5-mini. Total daily cost: ~$0.005 (way under the 100 free credits).

Above: the Summarize Each node opened after a real execution — left panel is the merged input from all three sources, right panel is the gpt-5-mini one-sentence summaries the chain produced.

Combine Items

A Code node groups the summarized items by source:

const bySource = {};
items.forEach(it => {
  const src = it.json.source || 'Other';
  bySource[src] = bySource[src] || [];
  bySource[src].push({
    summary: (it.json.output || it.json.response?.text || it.json.text || '').trim(),
    url: it.json.url,
    title: it.json.title,
    score: it.json.score
  });
});

The triple-fallback it.json.output || it.json.response?.text || it.json.text handles three n8n versions' summarization output shapes. Keep it — the cost is one character comparison, the benefit is forward-compatibility.

Step 5 — Slack Block Kit Formatting

Double-click Format Slack Blocks. This Code node builds Slack's official Block Kit payload — a structured JSON that renders into rich cards, not plain text.

const blocks = [
  { type: 'header', text: { type: 'plain_text', text: `🧠 Daily Tech Digest — ${date}` } },
  { type: 'context', elements: [{ type: 'mrkdwn', text: `Merged from ${Object.keys(bySource).length} sources` }] },
  { type: 'divider' }
];
Object.entries(bySource).forEach(([src, items]) => {
  blocks.push({ type: 'section', text: { type: 'mrkdwn', text: `*${src}*` } });
  items.forEach(i => {
    const scoreStr = i.score ? ` _(${i.score})_` : '';
    blocks.push({ type: 'section', text: { type: 'mrkdwn', text: `• <${i.url}|${i.title}>${scoreStr}\n${i.summary}` } });
  });
  blocks.push({ type: 'divider' });
});

The result: one bolded section header per source, each item on its own line with a clickable title and the AI summary underneath. Scores render in italics _(312)_.

Why Block Kit vs plain text? Plain text messages truncate titles at long URLs and don't render as clickable. Block Kit's <url|text> syntax gives you clickable link labels, grouping, dividers, and context footers.

Step 6 — Connect a Slack Webhook

6a. Create the Webhook in Slack

1. Open Slack in a browser → click your workspace name top-left → Settings & administration → Manage apps
2. Search "Incoming Webhooks" → Add to Slack
3. Choose the channel you want the digest posted to (e.g. #ai-digest)
4. Click Add Incoming WebHooks integration — Slack gives you a URL like https://hooks.slack.com/services/T01.../B02.../abc123
5. Copy that URL

6b. Paste Into the Workflow

Double-click Post to Slack Webhook in n8n. Replace the placeholder URL:

https://hooks.slack.com/services/REPLACE/WITH/YOUR_WEBHOOK

…with your real webhook URL. Save the workflow.

6c. Test It

Click Execute workflow on the canvas. The workflow runs through all 12 nodes and posts to your Slack channel. If everything's right, you get a rich message that looks like a curated newsletter — sources separated, links clickable, AI summaries under each item.

If the Slack message doesn't arrive:

• Open the execution log → click the Post to Slack Webhook node → check the HTTP response code. 200 = success. 400 = malformed JSON (usually a missing text fallback).
• Verify the channel the webhook was installed into — you can't redirect a webhook without installing a new one.

Extensions: Email, Teams, Discord, Personalize

Email Instead of Slack

Replace Post to Slack Webhook with a Gmail node (needs OAuth to your Google account). Map:

• To → your email
• Subject → 🧠 Daily Tech Digest — {{ $json.text }}
• Body (HTML) → convert the Slack blocks to HTML with a small Code node before this step

Microsoft Teams

Teams accepts Adaptive Cards instead of Block Kit. Create a new Code node that translates the bySource object into Adaptive Card JSON (schema here). POST to your Teams channel's incoming webhook URL.

Discord

Discord accepts webhooks but uses a different shape — embeds array with title, description, url, fields. Simple to translate from bySource.

Personalize With User Preferences

Add a Read From Database node before the fetches. Query a table of { user_id, topics[], min_score, slack_webhook } and fan-out per user. Each user gets a digest filtered to their topics with their personal webhook.

Tune the Sources Per Day

Add a Switch node after Schedule that routes to different fetch sets based on weekday. Monday = Hacker News + Dev.to; Friday = GitHub Trending + Product Hunt. Keeps the digest fresh.

What's Next

Combine this with YouTube → AI Blog Post to turn the GitHub repo of the day into a full walkthrough article, or with AI Lead Enrichment to identify hiring companies from Dev.to posts and generate outreach.