If you have both ollama and llama.cpp installed on your MBP, and are interested in utilizing your local model as a cost-saving preprocessor for your frontier models, consider using https://github.com/Standard-Pentest/kultivait!
"kultivait init --setup" to evaluate your machine's hardware, access to frontier cli-tools, download models, and create a proxy for ollama to get started. works great with opencode!
I’ve been using OMLX on 48gb Mac. It’s a lot smoother setup than Ollama and has a built in model sizer and all that in a nice ui. Also will pre configure and launch opencode, pi, Hermes, codex and Claude out of the box with no fuss. I’ve been really happy with it.
Comments
If you have both ollama and llama.cpp installed on your MBP, and are interested in utilizing your local model as a cost-saving preprocessor for your frontier models, consider using https://github.com/Standard-Pentest/kultivait!
"kultivait init --setup" to evaluate your machine's hardware, access to frontier cli-tools, download models, and create a proxy for ollama to get started. works great with opencode!
Note: I am replying to my own post to say that Kultivait llama.cpp functionality is currently broken but under active development.
Ollama proxy and model selection does work. If you run into issues or have any questions please feel free to reach out.
Friends don't let friends use Ollama
Can you elaborate for your unaware Ollama-using friends?
What the parent comment is referencing: https://sleepingrobots.com/dreams/stop-using-ollama/
I’ve been using OMLX on 48gb Mac. It’s a lot smoother setup than Ollama and has a built in model sizer and all that in a nice ui. Also will pre configure and launch opencode, pi, Hermes, codex and Claude out of the box with no fuss. I’ve been really happy with it.
I did try oMLX, mlx and LM studio (Now Bionic). I returned back to Ollama because of stability issues.
Ollama is the best for backends, I prefer using LM mini as a UI though, cause it lets me sync chats between my phone and mac.
More ethical alternatives have caught up including llama.cpp
https://sleepingrobots.com/dreams/stop-using-ollama/#what-to...
Very ironic how that article looks like it was LLM generated.
Dogfooding it!
At least it looks like the quality control wasn't too bad!
- Stop using OLLAMA
- https://sleepingrobots.com/dreams/stop-using-ollama/
- also can we get a post on how to do this with llama.cpp
Lol why bother with bloat ollama