straylight: disable laguna, default to ornith

configs/opencode.nix

@@ -17,7 +17,7 @@ let
   cfg = {
     "$schema" = "https://opencode.ai/config.json";
     enabled_providers = [ "llama.cpp" "zai-coding-plan" ];
-    model = "llama.cpp/laguna-s-2.1";
+    model = "llama.cpp/ornith-1.5-35b-a3b";
     small_model = "llama.cpp/ornith-1.5-9b-uncensored";
     provider = {
       "zai-coding-plan" = {
@@ -52,7 +52,6 @@ let
           chunkTimeout = 7200000;
         };
         models = {
-          "laguna-s-2.1" = text "Laguna S 2.1";
           "qwen3.6-35b-a3b-abliterated" = text "Qwen3.6 35B A3B";
           "gemma4-26b-a4b-abliterated" = text "Gemma 4 26B A4B";
           "ornith-1.0-35b" = text "Ornith 1.0 35B";

hw/straylight/default.nix

@@ -42,14 +42,17 @@ let
     ; DFlash GGUF is on disk (laguna-s-2.1-DFlash-BF16.gguf). nixpkgs llama.cpp 0.4.0
     ; rejects it at load ("expected 76, got 69" tensors); Poolside's file needs their
     ; laguna-branch loader. Leave unwired so Laguna still starts.
-    [laguna-s-2.1]
-    model = ${llamaModelsDir}/laguna-s-2.1-UD-Q4_K_S/Laguna-S-2.1-UD-Q4_K_S-00001-of-00003.gguf
+    ; DISABLED 2026-09-19: resident beside Ornith 35B it left ~15 GB free, and a third
+    ; model load then hit the 104 GiB TTM cap and OOM-killed the desktop session.
+    ; Weights stay on disk; uncomment to re-enable.
+    ; [laguna-s-2.1]
+    ; model = ${llamaModelsDir}/laguna-s-2.1-UD-Q4_K_S/Laguna-S-2.1-UD-Q4_K_S-00001-of-00003.gguf
     ; Two slots sharing one unified 131072-token KV pool: a second request (subagent,
     ; another client) runs alongside instead of queueing, and either one may still use
     ; the full 131072 as long as their combined context fits.
-    parallel = 2
-    kv-unified = true
-    ctx-size = 131072
+    ; parallel = 2
+    ; kv-unified = true
+    ; ctx-size = 131072
 
     ; Previous default model (a Qwen 3.5 derivative), kept for comparison runs. It was
     ; fetched by the old -hf flag into the HF cache; dedup hides the cache's own entry.