use with remote llama-swap

[khimaros]

i'm running a llama-swap server with many models available on a remote machine. what is the best way to configure the vscode plugin for this use case?

[igardev]

@khimaros Good question. I don't have experience with llama-swap, but I guess it proveds an endpoint, which supports the appropriate API. If you want to use code completion - load/configure a FIM model with llama.cpp (becuase /infill endpoint is needed) in llama-swap. After that add the llama-swap endpoint as external model ("More" button in Environment page - see the video).
In a similar way you could add for example external tools model (should support tools, but could be other than llama.cpp based) to support agents.

Screencast.from.2026-01-04.18-31-35.webm