The web dashboard is in a pretty good state. Some of the colors need tweaked a bit still, and I want to modify some of the pages. I want to integrate semantic search on the Add Thought page, replacing the "Writing Tips" and "Platform Limits" boxes.
EverythingSings
EverythingSings@primal.net
npub1cdm4...er54
Formless art for the future
Making good progress on the web dashboard, trying to clean up some of the CSS. It might be conflicting with the templating engine.
Testing from the web dashboard
We'll try FastAPI for the web server. Jinja2 will be the templating engine.
I bet I could have the server running while the GUI is open, so it could just be another tab / pane in the terminal, streaming info about the web frontend.
I'm building an interface for decentralized protocols. It has basic command line functionality, a moderately advanced GUI in the terminal. I might have taken the terminal GUI as far as it could go, it's getting pretty complex. But I'm considering integrating a web server / local frontend, mainly for the flexibility.
If a vector embedding would be truncated, should that always be avoided? Or is it better to allow the full length of a note, and hope that there's enough meaning captured before the truncation begins?
I'm pretty excited about semantic search in my custom interface. It opens up a lot of possibilities to explore.
I lost my reading streak again. I seem to struggle getting past 4 days. Yesterday was pretty busy though so I don't blame myself. I'll start the new streak today after I work on my terminal interface a bit.
I'm using all-MiniLM-L6-v2 for generating embeddings. It looks like it has a text limit, so I need to account for that, creating transparency on when a post would be truncated.
I've scheduled over 100 notes using my custom terminal interface. I was calling it ThoughtQ, then CastQ, but that's probably not the name I'll settle on. It looks like CastQ is used by something else already. So consider it a codename, I suppose.
And I need to improve the drafting page. The GUI shifts drastically while I'm typing, since it's automatically searching for semantically relevant notes in the past.
I'm going to tweak the search page of my custom interface. Instead of toggling between text / semantic search, I think we should just prioritize semantic and then default to text as a lower priority.
I've added semantic search to the post drafting section. So now when I draft a post, it searches for relevant posts in my database. Not just by matchnig text but semantic meaning. This is a major enhancement to the #UX.
It looks like semantic search is functional. Now to refine the user experience a bit.
I'm going to use the sqlite-vss extension. This should allow me to search my past thoughts by meaning instead of just SQL queries.
I think I'm going to implement vector embeddings on CastQ for local semantic search
Researching the various #nostr NIPs
Grateful for the insects in my yard.
I ordered a new #LifeStraw water filter. Grateful for access to clean water.