Yes. Please allow us to use local models, even if it’s a paid feature
3 Likes
In the meantime Continue published a post about using Mistral AI’s Codestral 22B, DeepSeek Coder 6.7B, and Llama 3 8B.
3 Likes
if i buy a new laptop i’d definitely be looking for something with enough ram to run some of these locally, pretty incredible!
Managed to connect a local Coder-V2 by using ollama and cloudflared tunnel. 16b model is super slow on my Mac M1 Pro, takes a minute to process any request, even the simplest ones.
But the main issue it breaks the code, injects some weird markdowns, cuts lines, and so on. Seems like it needs some additional instructions and prompts tuning on the Cursor side.
So guys, technically it’s possible but as of now it’s far from ready-to-use state.
1 Like
I am using LM Studio. I get this base url http://localhost:1234/v1 and I used it on cursor but not working. Here is the curl
curl http://localhost:1234/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer lm-studio" -d "{ \"messages\": [ { \"role\": \"system\", \"content\": \"You are a test assistant.\" }, { \"role\": \"user\", \"content\": \"Testing. Just say hi and nothing else.\" } ], \"model\": \"bartowski/DeepSeek-Coder-V2-Lite-Instruct-GGUF\" }"
Please someone tell me how to use local LLM on Cursor.
1 Like
You must have an external address. You can configure this through various proxying services or find something ready-made. Like this one works, but I’m not sure GitHub - kcolemangt/llm-router: Access models from OpenAI, Groq, local Ollama, and others by setting llm-router as Cursor's Base URL
1 Like
This needs to happen. Let me experiment using Cursor with Llama 3.1 I have on my own machine.