Frasier Bingo https://frasier-bingo.replit.app/
partner at SHV, where I build prototypes and help new companies.
mostly trail running; myriad typos, no complaints
| Website | https://sambreed.dev |
| Work | https://shv.com |
| Location | Mill Valley, CA |
"sam breed"
@sambreed
- 211 Followers
- 244 Following
- 426 Posts
@tbeseda this parrot is stone dead. An ex-parrot.
my relentless link dumping continues unabated — Dec 2023 edition https://sambreed.dev/writing/2023/links-december-2023
@ceejbot I once tried to help someone in a co-working space who had deleted the wheel user group on his iMac because it seemed suspicious
I made a page on my website with all the links I've posted from turning my browser upside down and vigorously shaking out the loose tabs at the end of every month. There's even a random button. Get 'em before they link-rot! https://sambreed.dev/links
@tbeseda I'm doing something similar on a (non-sharable) work project. Postgres is great for testing this because it supports full text search and vector search with pgvector.
What we've found is that embeddings search can be made to perform well but requires smart processing of the data up front to get good. LukeW touched on that in his post today https://lukew.com/ff/entry.asp?2045
@tbeseda sick! there's a cool local-js search in the twitter archive tool that @darius made that uses elasticlunr. I had never heard of it but it works pretty good! http://elasticlunr.com/ https://hellsite.sambreed.dev/?q=hotdog
Elasticlunr.js, lightweight full-text search engine in Javascript for browser search and offline search.
Elasticlunr.js, lightweight full-text search engine in Javascript for browser search and offline search. Elasticlunr.js is developed based on Lunr.js, but more flexible than lunr.js. Elasticlunr.js provides Query-Time boosting and field search. A bit like Solr, but much smaller and not as bright, but also provide flexible configuration and query-time boosting.
@tbeseda yep, that's exactly right. There are lots of different models for making embeddings that you can run locally — calling out to OpenAI is really just for convenience. And yeah, to do the comparison you need to generate an embedding for the search term at runtime, again this is where calling their API makes things more convenient. This post on Huggingface has good info on the "what" part https://huggingface.co/blog/getting-started-with-embeddings
I wrote a post about the little toy embedding search I added to my website over the weekend. It’s pretty good (if you cherry-pick the good queries)!
https://sambreed.dev/writing/2023/lil-vector-search
https://sambreed.dev/writing/2023/lil-vector-search
I'm gonna go ahead an put a 100% probability that they ate Danny to survive on Mars.