Support & Assistance
Welcome to the LocalModel Bridge support page. This app allows you to run generative AI models fully offline. If you have questions about loading models, configuring shortcuts, or setting up the local API server, we are here to assist.
Assistance
Open the app's Settings panel and select the Models tab. Click "+" to create a profile. You can enter a Hugging Face repository ID (e.g. mlx-community/Qwen2.5-1.5B-Instruct-4bit) or select a local weights folder on your Mac. Click "Create Profile", then load it from the status menu popover.
LocalModel Bridge operates 100% offline. Client network access is only used to query and download model weights from the Hugging Face Hub. Server network access is used solely to host the local HTTP OpenAI-compatible server on a chosen port (e.g. 8080) on your own machine (localhost).
LocalModel Bridge exposes native actions to macOS's built-in Shortcuts app. Open Shortcuts, create a new workflow, and search for "LocalModelBridge" in the action library. The "Query Local Model" action accepts a prompt and returns the text response generated by the running model.
In the General tab of Settings, you can toggle "Enable local HTTP API server". This exposes an OpenAI-compatible completion API at http://localhost:8080/v1/chat/completions, enabling dev tools (like Cursor editor) to query your local models.
Generative AI models occupy significant RAM. To prevent system slowdowns, LocalModel Bridge automatically unloads models from Apple Silicon Unified Memory after a configurable inactivity delay and purges the system Metal cache.
If the FAQ does not answer your question, feel free to send us an email. Please include the following details: