Thanks for writing this, very interesting and matches my ideals of keeping things as simple as possible. I am going to try and do something along these lines although my vault is quite heterogeneous being a mix of doc, pdf, ppt etc so I may have to do a preprocessing step to md? I do have a question on what you think the upper limit of entries in the index is before issues might arise? And when moving that to a more db implementation would pay off? And is deduplication of source generally handled by the user ie not replicating files in the raw vault? Thanks.
Awesome. I was hoping we’d see some great posts on adoption or using this. Its a cool idea.
Thanks, man! Have you implemented something for your own setup?
limited set up still with obsidian. it works great but i haven’t had the free time to get it where i want it yet.
It’s totally worth it. I’ve been using it for the past year, and it’s dope
This is super cool, thanks for sharing!
Glad you liked it!
This is awesome; I do think the re-ranking step could improve your LLM costs and even accuracies here!
Yes, agree! I have to implement it.
Do you have a particular model in mind? Something run locally would be best
Thanks for writing this, very interesting and matches my ideals of keeping things as simple as possible. I am going to try and do something along these lines although my vault is quite heterogeneous being a mix of doc, pdf, ppt etc so I may have to do a preprocessing step to md? I do have a question on what you think the upper limit of entries in the index is before issues might arise? And when moving that to a more db implementation would pay off? And is deduplication of source generally handled by the user ie not replicating files in the raw vault? Thanks.
yes! they seem so overkill