#Visual Graph-Based Interface for Large Language Models

1 messages · Page 1 of 1 (latest)

stable current
#

Hi everyone and @turbid obsidian! Wanted to share what I've been building in the last few months happy_engineer

I've always felt that the existing text interface for Large Language Models made it difficult to learn and organize information. There is no easy way to branch off and dive deep into a topic and then back track, all of which are essential for learning.

To solve for this, I decided to build Stitch, a visual, graph-based interface for large language models.

Its still pretty early stage but I think ready for a first round of users and feedback. Really appreciate anyone that tries it out and I look forward to hearing your thoughts!

I've attached a short demo video I made and a link to the website to try it out.

Matthew

turbid obsidian
#

I played with this earlier, a bit rough but very promising! I'm keen to see some obsidian style linking features maybe? 😁

hallow sleet
#

(of course tris wants obsidian features)

stable current
# turbid obsidian I played with this earlier, a bit rough but very promising! I'm keen to see some...

For sure! I have Obsidian-like linking in the plans 🙂 - was actually thinking of a related new feature where users can select two nodes and ask for the AI what the connection between them are (and thus form a new node on the graph).

In addition, there are a few things I want to add in the near future:

  1. A chat-to-graph toggle.

This way users can use Stitch in the native text-based interface that LLMs have now if they so choose (so they aren't forced to always use the graph view).

  1. A Stitch conversion tool.

The idea is to allow users to take any form of content (blog posts, lecture notes, PDFs, YouTube videos, etc.) and upload the file/paste the link in Stitch and have it converted into a dynamic graph. I'm particularly excited about this feature as I think it has the potential to revolutionize how people synthesize new information.

  1. Share feature.

Idea is pretty simple - allow users to share graphs with each other via links. I almost have this done (but was trying to do too much and allow users to collaborate in real time in graphs... and that got way to difficult lol).

I'm thinking in the future there can be like a social media like feed where users can see graphs that other people made and view them/copy them to their canvases and explore them further.

#

So.. hopefully a lot of buffs coming soon~