[zix790prors] More local LLM updates

Using qwen3:30b explicitly. The default "qwen3" from ollama was pulling
a very outdated model apparently (qwen3:8b). qwen3:4b and qwen3:30b are the newest.

Also sets up some defaults for gptel that have been useful
This commit is contained in:
2025-09-11 08:39:36 -07:00
parent 6bf0a37533
commit 56b1111f54
2 changed files with 14 additions and 8 deletions

View File

@@ -83,12 +83,16 @@
(setq! gptel-api-key (my/get-rbw-password "openai-api-key-chatgpt-el") (setq! gptel-api-key (my/get-rbw-password "openai-api-key-chatgpt-el")
gptel-default-mode 'org-mode gptel-default-mode 'org-mode
gptel-use-tools t gptel-use-tools t
gptel-confirm-tool-calls 'always) gptel-confirm-tool-calls 'always
gptel-include-reasoning 'ignore
gptel-model "qwen3:30b")
(gptel-make-ollama "Ollama-Local" ;; Set default backend to be Ollama-Local
:host "localhost:11434" (setq! gptel-backend
:stream t (gptel-make-ollama "Ollama-Local"
:models '(deepseek-r1 deepseek-r1-fullctx qwen3 qwen3-coder llama3.1 qwen2.5-coder mistral-nemo gpt-oss)) :host "localhost:11434"
:stream t
:models '(deepseek-r1 deepseek-r1-fullctx qwen3:30b qwen3:4b llama3.1 qwen2.5-coder mistral-nemo gpt-oss)))
;; Define custom tools ;; Define custom tools
(gptel-make-tool (gptel-make-tool

View File

@@ -71,9 +71,11 @@ with lib;
powerManagement.finegrained = false; powerManagement.finegrained = false;
}; };
environment.systemPackages = [ services.ollama = {
(pkgs.ollama.override { acceleration = "cuda"; }) enable = true;
]; acceleration = "cuda";
loadModels = [ "gpt-oss" "deepseek-r1" "qwen3:30b" ];
};
# This option defines the first version of NixOS you have installed on this particular machine, # This option defines the first version of NixOS you have installed on this particular machine,
# and is used to maintain compatibility with application data (e.g. databases) created on older NixOS versions. # and is used to maintain compatibility with application data (e.g. databases) created on older NixOS versions.