The DEVONtechnologies Blog
How to Use Local AI in DEVONthink To Go
When using AI on mobile devices, you typically access cloud services like Claude, Gemini or ChatGPT. You can’t run a full local model on an iPhone today, but you can run one on your Mac and connect DEVONthink To Go to it.
First, your AI application needs to make itself accessible on the network. Look in the application’s Settings or its help for terms to that effect. You also need to enable Local Network access for DEVONthink To Go in the OS’ Settings > Privacy & Security. If needed, make sure to allow the connection in your firewall or network application.
Using LM Studio as our example, open its Developer view, load a model, and ensure the server is running. You will see a http://localhost:1234 or http://127.0.0.1:1234 URL appear. It’s a self-reference that AI-enabled applications on your Mac can use to connect to LM Studio running on the same machine. But other devices are not your Mac, therefore you can’t use that URL in DEVONthink To Go. However, it is possible to connect our mobile application to an AI application on your Mac.
Back in LM Studio’s Developer view, click the Server Settings button, and enable Serve on Local Network. You will see the URL change to an proper IP address. Click the copy icon to copy the URL.
In DEVONthink To Go’s Settings > AI, choose LM Studio. It will be pre-populated with a URL. Select everything before the colon and either paste or type in the protocol and IP address, e.g., http://192.168.12.185. Make sure that the port, e.g., :1234, doesn’t appear twice, because the copied URL could already include it. It should connect to LM Studio on your Mac and display the available models you have loaded.
Note, however, that IP addresses can change, and when it does, you would have to update the URL in the settings. However, you can instead use the Bonjour name of your Mac. So instead of using the IP address, you can enter, e.g., http://leonidasx.local. Now the URL will remain valid even if the IP address changes.