search

brianpetro / obsidian-smart-connections

Chat with your notes & see links to related content with AI embeddings. Use local models or 100+ via APIs like Claude, Gemini, ChatGPT & Llama 3
https://smartconnections.app
GNU General Public License v3.0
2.57k stars 174 forks source link

Best embedding models? #800

Open huachuman opened 2 days ago

huachuman commented 2 days ago

I have a lot of psychology and neuroscience content in my vault I like to use to help write articles and things of that nature. What embeddings model should I use? I'm worried about trying different ones and spending too many tokens on it.

In terms of online vs local, currently I'm using a laptop with an i7 and 16gb RAM but no GPU. Should I opt for something online or will the lack of GPU make a difference in the quality of embeddings? I'm not too worried about uploading my data. I just want whatever will work best. Only have $5 of OpenAI credits right now so trying not to go too overboard on the cost...

brianpetro commented 2 days ago

The default local model would be a good start to get a feel for things. The OpenAI models should perform better but it would make sense to get familiar with things using the BGE-micro-v2 🌴

huachuman commented 2 days ago

Interesting. Maybe something else is going on then. Maybe I should clarify why I asked then. I tried getting it to query my notes by asking "based on my notes" but it doesn't seem to be pulling up much relevant stuff, or will even just reference only the existing note. I remember in V1 it would be able to do what I'm asking very well... I must have something set up wrong here.