Persistent knowledge graphs with Zettelkasten / Digital Garden / Second Brain structures

[TomLucidor]

Thanks for creating this repo! I wonder if there are ways of making this more than just a single-instance notebook. Maybe there are ways for it to cross-link across multiple papers and web articles, and slowly aggregate new knowledge?

[lfnovo]

Hey Tom, this is definitely something I have in mind for this project. I am actually experimenting with it for a while to see if I can find the right way to introduce it. If you'd like to contributes with ideas, it'll be welcome.

[TomLucidor]

Give me some time, I managed to get a master list of 240+ MCPs, and I am planning to feed it into an LLM to catalog all their features.
journal.txt

[lfnovo]

This direction remains relevant, so we are keeping this open in Ideas as the product conversation for persistent knowledge, relationships between sources, and experiences that cross notebook boundaries.

The related issue #662 currently captures possible implementation concepts such as collections, tags, hierarchy, and cross-notebook actions. Please use this Discussion for concrete workflows, constraints, and examples of how connected knowledge should behave. The issue should only become executable once those product decisions are clear.

[FlorianHeigl]

Hi,

look at Mindforger for quite good workflows. Note the crucial bit is having good ways to assign tags and automactially link between notes in the notebook. I linked the project website instead of github since the showcases can explain better than me.
Unfortunately I myelf do not use the graphs because I had badly imported my original pile of markdown/text files and I can't solve it. I really dug a hole there and so far it's been easier to keep digging than climb out.
Ignoring that, use it since approx 7 years and it's worked quite well for writing 100s of pages of documentation, collecting links etc.
And I know that without such "me problems" their graphs are good. The author focusses on knowledge management and the program is the 3rd or 4th iteration of getting it done 'well'. I think it's the ideal approach so far. It's overlap with Open Notebook is otherwise limited because Mindforger doesn't really deal with external/imported pdfs etc.

[lfnovo]

Status refresh: #1315 lays out the graph direction this thread has been asking for — in the order we can trust: first the relations we already know (belongs to, derived from, mentions, generated from), then semantic extraction (entities, claims, contradictions) as optional enrichment, never as a requirement for basic search or small local installs. If your Zettelkasten use case needs something the first phase doesn't cover, that's the input we need.

[TomLucidor]

Conceptual adjacency and breakdown/decomposition would be useful as well, and also there are ideas on "seen this before" when problems/topics overlap frameworks/models sometimes by chance