How to Use Local LLM in Cursor

[@Jake-ky2ed]

Thank you. very valuable video for me.

[@amrsalem6655]

Nice videos and very useful, is there ability to get an AI assistance similar to one in GPT openAI and ability to use its API

[@cheyannehutson2412]

Thank you for the video! But some questions how do you change the context length of your model when running it this way? I know usually you would set the context length when using “ollama run [model]” but it seems you don’t get that chance with this configuration. Any help with this would be appreciated, thank you!

[@loganhallucinates]

You can set the context length manually using something like `/set parameter num_ctx 32768`, don't forget to `/save ` as well

[@0057beast]

Man i needed this for coding thanks​@@loganhallucinates

[@DevJonny]

If it's local why do I need ngrok? I was looking for something offline

[@loganhallucinates]

It’s for exposing your local API to the internet so Cursor server can access it. Their logic is in their server.

[@MondoBoricua]

still working ?

[@boiserunner]

Thanks for the video. Why would anyone want to do this?

[@loganhallucinates]

Some local models are quite capable now

[@moresignal]

This is ridiculously dangerous advice given Ollama has no authentication and you are suggesting to make it open to the entire Internet

[@loganhallucinates]

ngrok has authentication you can setup

[@sneedtube]

Doesn't work anymore, to me it looks like the Cursor devs are actively sabotaging every effort from the open source community to democratize that IDE, shameless merchants

[@nielsdebont270]

ollama_origins=* command is not recognized for me

[@AndrewQuardex]

SET OLLAMA_ORIGiNS=* if windows