On 2026-02-19 06:42, Petter Reinholdtsen wrote:
> When starting the llama.cpp server to provide an API and a simple web
> service and providing it on a network, the web service do not work as
> the requests are sent to the wrong location.
>
> I start my server like this:
>
> LC_ALL=C llama-server -ngl 256 -c $(( 42 * 1024)) --temp 0.7 \
> --repeat_penalty 1.1 -n -1 \
> -m ~/models-llama/Qwen3-Coder-30B-A3B-Instruct-Q5_K_S.gguf
>
> The cause is the hardcoded use of http://127.0.0.1:8080 as the API end
> point in the web service javascript.
>
> Applying the following patch fixes it locally, but it would be better if
> the baseURL was dynamically calculated from where the web page
> originated, and perhaps be overridable using some local configuration.
The server can be started with a --host parameter and I think this could
be passed on, but TBH I don't think anyone upstream is looking at the
simplechat frontend.
I made the backend changeable via the alternatives mechanism. I did
package the regular frontend, but never uploaded it (I forgot). I will
do so over the weekend. The downside is that it needs to go into
non-free because not all (node) build dependencies are in Debian.
Best,
Christian
>
> ---
> /usr/share/llama.cpp-tools/llama-server/themes/simplechat/simplechat.js.orig
> 2026-02-19 06:30:45.484328145 +0100
> +++ /usr/share/llama.cpp-tools/llama-server/themes/simplechat/simplechat.js
> 2026-02-19 06:33:44.000033528 +0100
> @@ -722,7 +722,7 @@
> class Me {
>
> constructor() {
> - this.baseURL = "http://127.0.0.1:8080";;
> + this.baseURL = "https://server.example.com:8080";;
> this.defaultChatIds = [ "Default", "Other" ];
> this.multiChat = new MultiChatUI();
> this.bStream = true;
>
