Anders ThoressonMar 29, 2025

Language models make it possible to approach media monitoring in a new way. Instead of keyword searches that try to filter out what I might be interested in, the job is handled by a language model that “assesses the news value” of my most important feeds — based on a prompt where I describe both myself and my interests. #n8n

https://anders.thoresson.se/post/2025/03/media-monitoring-with-llm-assistance/

Media monitoring with LLM assistance

Language models make it possible to approach media monitoring in a new way. Instead of keyword searches that try to filter out what I might be interested in, the job is handled by a language model that “assesses the news value” of my most important feeds — based on a prompt where I describe both myself and my interests.

Show threadSamApr 2, 2025

@anders nice work Anders! I'll be trying to replicate this, seeing if I can leave this running overnight with some Ollama models crunching through the articles.

How many articles does your system crunch through daily? Just to get an idea of the cost per article…

Show threadAnders ThoressonApr 2, 2025

@warnaars Cool! Please let me know how it goes!

I don’t now for sure. Full length newsletters like 5-10 articles per day, and for the broader, more general feeds maybe 100-200.

Show threadSamApr 2, 2025

@anders I have started passing them through Gemini (depending on the model used, the limits should be fine).

I’ve to figure out how to transform the LLM response from a markdown version code block to actual JSON.

The support forums have several threads around this issue, I have yet to figure out how to solve the issue though (I’m still a novice with n8n, this is me learning the ropes)

Show threadAnders Thoresson
@warnaars I had a lot of help using ChatGPT and Claude for the development.
Apr 2, 2025 at 6:07pmWeb
Show threadSamApr 3, 2025

@anders I didn't turn to the LLMs yet, but that would be my next stop. I think for now my understanding of n8n is still limited, so I might be doing things the wrong way. I'll play more with this over time.

Once I get it going a thing I consider doing is doing a preselection on the firehose of feeds I track by Inoreader and only use the feed of those items as generated by Inoreader for the n8n workflow.

Show threadSamApr 3, 2025

@anders I track industries or technological developments, which causes the amount of feeds/articles I track to be very wide. I'd need to see if that falls within the rate limits of my Gemini API (maybe if I do an hourly run)?

Anyways, enough to explore.

Show threadAnders ThoressonApr 6, 2025
@warnaars I’ve been considering the same, connecting n8n to a feed generated by my feed reader. That way it should be possible to make the n8n flow way easier to build.