Is this tree — after selected (by me) for barking-up — the wrong one?

[marclar]

I'm using a Personal Knowledge Manager (a "PKM" named Logseq) and I'm starting to think of interesting ways to visualize my notes.

Is Trinity something I might be able to run using data from my "knowledge graph" used to generate the output?

[Birdasaur]

Hello @marclar !
I think the question would be how are you managing the locations and links in your Knowledge graph?
Right now Trinity has some code to manage nodes on a graph but it was a one off experiment for loading data from a similar problem at work. I never got around to implementing a 3D render of the links themselves. (ie... lines connecting nodes)
If you absolutely need the links then we can talk about implementing an upgrade to do so.

Trinity can currently plot your nodes as a scstterplot if you maintain coordinates for your nodes. (it will not do an auto layout)
So if you have coordinates already it would be as simple as exporting them to a FeatureCollection object (json) which is a list of FeatureVector objects. Each FeatureVector would be one of your nodes in your graph. The data field in each FeatureVector would contain your coordinates.
If your nodes are actually vectors (like in a vector database) then it would still work just fine if you used the vectors as the coordinates. (ie... the vectors would be the data field)
While initially this would only be a 3D scatterplot it would immediately enable you to do all the cool projection stuff like running UMAP and various built in clustering algorithms.
From there you could do further analysis by associating content with each node. For example what does each of your nodes represent? if it is text then you can add the text of each node to its respective FeatureVector object inside the json. (there is a dedicated text field) There are similar fields for images and audio. (no video support yet, working on it)

Also most importantly you can also add a label which is a string type category. this allows Trinity to autocolorize and provide filtering.

Does this jive? I can help you map your data to an appropriate json format if you like.

[marclar]

Oh, that's awesome -- thanks so much.

It happens to be my Theory of Everything and I imagine different parts will require different representations and different coordinates to make sense. I'm not sure it can be communicated (to most consumers) without ways to visualize change over time, which I believe Trinity would handle.

As for how I structure things, there's a very loosely-enforced structure around a couple of main projects, but otherwise it's mostly notes dashed off with #whatever so they can be queried later.

I like the whiteboard feature Logseq has but I think this needs to be done in 3D.

So if you have coordinates already

I don't but I was thinking they could be generated until either

1. the query used to assemble data on the PKM side outputs perfectly-formatted JSON with coordinates that make sense, AND / OR
2. I rearrange / re-tag some things to conform to a standard that makes sense for Trinity, then make a query that only returns pages that comply

There are similar fields for images and audio. (no video support yet, working on it)

Love this. idk what this would entail, but an HTML (and ideally, JS) renderer would be awesome, since some of my notes are just hyperlinks. But I can imagine that (the JS execution especially) being kind of difficult.

Anyway, lmk what you think 🫡