ModelsProviders
LM Studio
Run local models in LM Studio and use them in Hull through its OpenAI-compatible server.
LM Studio is a desktop app for downloading and running local models. Its built-in server speaks the OpenAI API, and Hull connects to it with the built-in lmstudio provider.
Website: lmstudio.ai
Set up LM Studio
- Download and install LM Studio.
- Search for and download a model. Prefer models that support tool calling and have a large context window.
- Load the model and start the local server from LM Studio's developer view. The default address is
http://127.0.0.1:1234/v1.
Connect LM Studio
Open Settings, go to Providers, choose Add provider and select LMStudio. Leave the API key empty unless you've enabled authentication in LM Studio.
Use a model that isn't listed
If the model you loaded doesn't appear in the model picker, register it in hull.jsonc. The model key must match the identifier LM Studio serves, or use id to map a friendlier key to it:
{
"model": "lmstudio/my-llama",
"provider": {
"lmstudio": {
"models": {
"my-llama": {
"id": "meta-llama-3.1-8b-instruct",
"name": "Llama 3.1 8B (Local)",
"limit": { "context": 32768, "output": 8192 },
},
},
},
},
}See Custom models for every model field.
Tips
- The LM Studio server must be running for Hull to connect.
- If LM Studio reports "Please check the LM Studio developer logs to debug what went wrong", the context length is usually too small. Increase it in LM Studio's model settings.
- See the LM Studio documentation.