Automation
AI & Open Source Daily
An open-source n8n workflow that gathers AI and open-source news every morning, reads the full articles, and has DeepSeek write a 3,000–4,000-character in-depth digest in Chinese, delivered to your inbox.
Overview
- Follows 6 sources including Hacker News, OpenAI, Hugging Face, Google AI and GitHub, filtered and de-duplicated over the last 24 hours
- More than a summary — a daily overview, 3–5 in-depth picks (what happened / why it matters / what it means for developers) and an editor's note
- The workflow holds no secrets — import it into n8n, add two credentials and it runs; each run costs a few cents
- Type
- Automation
- My role
- Design · development · open source
- Date
- 2026.10
- Status
- Done
Repository
42ccode/n8nAINewsEmail- Star
- 0
- Fork
- 0
- Commits
- 1
- License
- MIT
git clone https://github.com/42ccode/n8nAINewsEmail.gitLast push 2026-10-23 · fetched from GitHub when the site was built
Why I built it
I wanted to keep up with AI and open source every day, but there are too many sources and most of them are in English. So I built a Chinese digest that lands in my inbox at 8 am — it reads like an edited briefing, not a list of links.
What the digest looks like
Every email has five parts:
| Section | What’s in it |
|---|---|
| 📌 Today’s overview | Two or three paragraphs on the day’s main threads and trends |
| 🔥 In depth | 3–5 picks, each covering “what happened / why it matters / what it means for developers” |
| 🤖 AI news | The rest of the AI articles, two or three sentences each, with links to the originals |
| 🛠 Developers & open source | Updates to tools, frameworks and open-source projects, also linked |
| 💬 Editor’s note | A few closing sentences with an overall take |
Sources
- Hacker News: only posts with 100+ points
- OpenAI News, Google AI Blog, Hugging Face Blog
- Simon Willison’s Weblog: an independent blog that often writes about working with LLMs
- GitHub Blog
To add another source, edit the array in the “source list” node.
How it works
Every day at 8:00 (or run it by hand)
→ source list → read RSS
→ filter and tidy: last 24 hours only, de-duplicated, at most 25 articles
→ fetch full text → merge (extract the article body, build the prompt)
→ generate the digest (DeepSeek API)
→ check the digest: don't send if it's empty
→ convert to HTML → send the email
Running it yourself
- Download
workflow/ai-daily-digest.jsonfrom the repo and use Import from File in n8n. - In the “generate digest” node, pick or create a DeepSeek credential with your API key.
- In the “send email” node, create an SMTP credential (Gmail, QQ Mail and 163 Mail all work) and set the sender and recipient to your own address.
- Set the workflow’s time zone to your own in its settings, otherwise “every day at 8:00” follows the server’s clock.
- Click Execute workflow to run it once — the digest shows up in your inbox in a minute or two. When you’re happy, click Publish and it runs every day.
Tested on n8n 2.43, both self-hosted and n8n Cloud.
Trade-offs
- DeepSeek as the model: great value, and reachable directly from mainland China. Each run is roughly 14k input and 5k output tokens, usually a few Chinese cents.
- Calling the API directly: n8n’s built-in LLM node sometimes runs out of output budget on long texts and returns an empty answer, so the workflow calls
/chat/completionsthrough an HTTP Request node instead, still using n8n’s built-in DeepSeek credential type. - When the full text can’t be fetched: some sites sit behind protection like Cloudflare; the workflow then falls back to the RSS summary without breaking the digest.
- Different model or language: the model name and prompt live in the “merge” node, and any OpenAI-compatible service works.
Privacy
- The workflow file contains no secrets. API keys and email passwords stay in n8n’s own encrypted credentials and are never included when you export or share the workflow.
- Only the titles, links and excerpts of public articles are sent to DeepSeek — never email contents or other personal data.
- The digest retells other people’s articles and is for personal reading only.
This is my first n8n automation project, open-sourced on GitHub under the MIT licence.