Fixed now! Supermemory server runs in under 1GB ram (still self contained 🤩)
Just shipped the version 0.0.3.
Some more cool features:
- You can (finally) run it with Ollama / any other OpenAI compatible server
- It's Code-signed now so it can run in secure mac environments
- Fixed other bugs with migrations etc.
there was a small bug related to wasm embedding models locally - we optimized it to run on server environments where there's a bunch of resources available / a significant portion of server can be used.
(it was also documented & people can always change the settings, but the defaults were wrong)
There's tradeoffs, like only 2 concurrent adds at any time, but you can increase it via SUPERMEMORY_RAM environment variable for hosting it for more production use!
(also, every time someone reports a bug why is it a batsignal for haters 😄 funny to see people and competitors always doing this)
the supermemory-server is using more RAM idle (7.8GB) than my entire browser
installed it last night. it has ~5MB of actual data. five megabytes → 7.8GB of RAM.
not usable on a 24GB Mac. I was wondering why my system kept hanging, it was this. had to remove it :(