Scheduled digest

Runs locally

Needs Ollama with a model pulled, such as gemma4:e4b. No key. The fetch goes to the public Hacker News feed; point it anywhere else you like.

Every morning at 9, the graph fetches the Hacker News front page, asks the model for the five stories most worth reading, and writes the answer to a file named after the day.

Download the graph and save it in user/graphs/, or build it below. The graph is MIT-0: copy it and license what you build from it however you like. It has no model picked: pick one on the LLM node, or the LLM writes a mock reply instead of a digest.

1. The nodes

Node	Settings
Schedule, named Every morning	cron 0 9 * * *, timezone Local
HTTP Request, named Feed	method GET, url https://news.ycombinator.com/rss, response text
Template, named Prompt	the text below
LLM	model: pick Gemma 4 e4b (Ollama), or the model you pulled
Preview, named Digest	none
Time, named Today	format 2026-06-08
Template, named File name	digests/{today}.md
Write File, named Save	path empty: it comes from the wire

The Prompt:

Here is the Hacker News front page as RSS. Pick the five stories most worth reading and give each one line: the title, then why it matters.

{feed}

2. The wires

From	To
Every morning trigger	Feed trigger
Feed trigger	Prompt trigger
Feed body	Prompt feed
Prompt trigger	LLM trigger
Prompt out	LLM prompt
LLM trigger	Digest trigger
LLM response	Digest in
Digest trigger	File name trigger
Today out	File name today
File name trigger	Save trigger
File name out	Save path
Digest out	Save content

The trigger runs down the whole line, one node at a time. The Feed fires the Prompt only when the page has arrived, so the Prompt always has the fresh feed, and the Prompt then fires the LLM. The Digest shows the answer and passes it on with one trigger, which fires the File name and then Save, so each morning writes one file. A tag is named after the node its wire comes from: Feed body becomes {feed}, Today out becomes {today}. The file name is filled in at the moment of the run, so it carries the day.

3. Turn it On and leave it

Press On. The Schedule checks the clock every 15 seconds and fires once in the minute that matches, so at 09:00 the digest lands in user/data/files/digests/, one file per day. It skips the minute the graph turned On in, which keeps a Restart from firing twice.

To try it without waiting, set the cron to * * * * * (every minute) and turn it On (or Save & Restart if it is already On): it fires early in the next minute. Put the cron back when you are done.

Cron in one line

Five fields: minute, hour, day of month, month, day of week (0 or 7 is Sunday).

Cron	Fires
0 9 * * *	at 09:00 every day
0 9 * * 1-5	at 09:00 on weekdays
*/15 * * * *	every quarter hour
0 18 * * 5	at 18:00 on Fridays
0 8 1 * *	at 08:00 on the first of the month

Change it

• Another source. Any URL works. For JSON APIs, set the response to json and pick fields out with Get before the Prompt.
• Only when something changed. Poll with an Interval and put a Changed node after the fetch: the model only runs when the page is different.
• Send it somewhere. Swap the Write File for an HTTP Request that posts to a chat service, or an Append File that builds one long log.
• Reminders from data. For times that live in a table instead of a cron line, use Agenda.