diff --git a/hosts/architect/llm.nix b/hosts/architect/llm.nix index bf46d56..25e17dc 100644 --- a/hosts/architect/llm.nix +++ b/hosts/architect/llm.nix @@ -1,11 +1,9 @@ { pkgs, ... }: let - frontendDomain = "pino.giugl.io"; backendDomain = "ollama.giugl.io"; - frontendPort = 3030; + ollamaHost = "127.0.0.1"; ollamaPort = 11434; - listenAddress = "127.0.0.1:${toString ollamaPort}"; ollamaPkg = pkgs.unstablePkgs.ollama; in { @@ -14,37 +12,23 @@ in }; services.ollama = { - inherit listenAddress; - enable = true; - acceleration = "cuda"; + package = ollamaPkg; + host = ollamaHost; + port = ollamaPort; + acceleration = "cuda"; environmentVariables = { - # OLLAMA_ORIGINS = "10.0.0.0/24"; OLLAMA_FLASH_ATTENTION = "1"; OLLAMA_NUM_PARALLEL = "2"; }; }; - # architect.vhost.${frontendDomain} = { - # dnsInterfaces = [ "tailscale" "lan" ]; - - # locations."/" = { - # host = "127.0.0.1"; - # port = frontendPort; - # allowLan = true; - # allowWAN = true; - # extraConfig = '' - # proxy_read_timeout 600s; - # ''; - # }; - # }; - architect.vhost.${backendDomain} = { dnsInterfaces = [ "tailscale" "lan" ]; locations."/" = { - host = "127.0.0.1"; + host = ollamaHost; port = ollamaPort; allowLan = true; allowWAN = true; @@ -56,28 +40,4 @@ in ''; }; }; - - # virtualisation.oci-containers = { - # containers = { - # ollama-webui = { - # image = "ghcr.io/open-webui/open-webui:main"; - # autoStart = true; - - # ports = [ - # "127.0.0.1:${toString frontendPort}:8080" - # ]; - - # environment = { - # OLLAMA_BASE_URL = "https://${backendDomain}"; - # }; - - # extraOptions = [ - # "--pull=always" - # ]; - # volumes = [ - # "/var/lib/ollama-webui:/app/backend/data" - # ]; - # }; - # }; - # }; }