new | show | ask | jobs Github

osigurdson 6 hours ago

I built a way to chat with the documents in this post (updates live).

https://nthesis.ai/public/702445eb-f0e4-4730-b34f-f34eb06dd6...

Or you can do a basic text search: https://nthesis.ai/public/hn-working-on

▲

SamDc73 6 hours ago | parent | next [-]

Nice one.

But the graph view seems broken?

And for the data, it would be nice to have the original URL for each comment as a reference.

▲

osigurdson 6 hours ago | parent [-]

Thanks! Is it that you can't see the graph view at all? It uses DeckGL which might be a bit too fancy and may not work on all browsers possibly. It works on Chrome/Safari on my iPhone and Chrome on Arch.

I didn't realize that you actually could provide a working link back to Hacker News but it seems HN does support that. Thanks, I will give that a try!

▲

SamDc73 6 hours ago | parent [-]

https://imgur.com/a/EmtADiA

This how the graph looks like, not clustered by tag or anything ... (I was expecting a view like Logseq or Obsidian)

	▲	osigurdson 5 hours ago \| parent [-]
		Oh ok. Yeah, it is just not really complete yet. What this does is displays a UMAP of the embeddings for the selected feature and you can mouse over for labels. It isn't identifying clusters itself yet, finding exemplars, etc. Right now, points that are close to each other are just semantically close - as judged by UMAP. That being said, there isn't much data for this month yet. If you look at last month's data (https://nthesis.ai/public/e4883705-ec05-4e7a-83ac-6b878cc1e8...) , clusters are more apparent (particularly if you view the tags instead of summary).

▲

duckb 6 hours ago | parent | prev [-]

What LLM do you use for this? Self hosted or cloud?

	▲	osigurdson 6 hours ago \| parent [-]
		This uses the OpenAI API currently. I tried Gemini for embeddings for a little while but didn't seem materially better.