Change the Model Server
Each gateway publishes one model server. To point it at a different one, for example after moving from Ollama to LM Studio, or after changing the port:
- Open your gateway page and find the Model server box.
- Set the scheme, host and port as seen from the machine running the connector:
localhostif the model runs on the same machine as the connector.- A LAN address, such as
192.168.1.20, if it runs on another machine on your network.
- Click Save. The connector picks up the change within a minute, with no restart and no new key.
Model server tips
| Model server | Default port | Notes |
|---|---|---|
| Ollama | 11434 | Runs in the background once installed. On another machine, it must listen on the network, not only on localhost. |
| LM Studio | 1234 | Start its local server in LM Studio first. |
| vLLM | 8000 | Started with vllm serve. |
| llama.cpp server | 8080 | Started with llama-server. |
Two model servers?
Create a second gateway. Each one has its own address and key.