Fascinating article. It would be useful to know how you got to this architecture. Where did you start first, what prompted you to add more complexity, and which parts of the architecture added most bang for the buck.
Fellow Readwise user here - do you have a particular system for using it at all? Like tagging articles to be able to query the "golden" ones for a particular topic?
Do you export highlights to Obsidian or anything like that?
I tried using tags, but couldn't find a complete system that works so far. If you add tags such as RAG, Agents, etc. it's useless. I am trying to find something more actionable such as "personal-story", "to-create-content", etc. but that's still a work in progress. Still text, vector and graph search work really well! (Readwise has text + vector search by default)
And yes, I export my highlights to Obsidian, still, that's not super useful here, as you can query them as well through the Readwise CLI.
Fascinating article. It would be useful to know how you got to this architecture. Where did you start first, what prompted you to add more complexity, and which parts of the architecture added most bang for the buck.
Great feedback. Would you like an article just on that topic?
Yes! If it is not worthy of a full article, a section on this current article would also do
Added it to my pipe!
Sorry for asking, but what’s the current price for running this loop per article? An estimate or range is fine.
It depends on how many images you want to generate and the length of the article, but I would say on average, $2
That’s great, thanks Paul.
Fellow Readwise user here - do you have a particular system for using it at all? Like tagging articles to be able to query the "golden" ones for a particular topic?
Do you export highlights to Obsidian or anything like that?
I tried using tags, but couldn't find a complete system that works so far. If you add tags such as RAG, Agents, etc. it's useless. I am trying to find something more actionable such as "personal-story", "to-create-content", etc. but that's still a work in progress. Still text, vector and graph search work really well! (Readwise has text + vector search by default)
And yes, I export my highlights to Obsidian, still, that's not super useful here, as you can query them as well through the Readwise CLI.
What is your system at the moment?
Awesome!
Have you implemented anything similar?
Nothing quite like this for content creators!
what are you cooking atm?
A RAG voice assistant for D&D, an LLM evaluation benchmark, and a Spotify track recommendation system - all with varying degrees of success! 😅
The RAG voice assistant for D&D sounds dope! How is that working so far?