fix(llm-runner): set -np 1 to prevent OOM crash on large context windows

bin/llm-runner

@@ -136,6 +136,7 @@ launch_instance() {
         --host "${HOST:-0.0.0.0}"
         --port "${port}"
         --ctx-size "${ctx}"
+        -np "${PARALLEL:-1}"
         --n-gpu-layers "${ngl}"
         --load-mode "${load_mode}"
         --temp "${temp}"

bin/llm-tui

@@ -734,6 +734,7 @@ class LLMConfigApp(App):
         env["FLASH_ATTN"] = "on" if self.query_one("#check-fa", Checkbox).value else "off"
         env["CTK"] = str(self.query_one("#select-kv", Select).value or "q8_0")
         env["CTV"] = env["CTK"]
+        env["PARALLEL"] = "1"
 
         if model_val and Path(model_val).exists():
             env["MODEL"] = model_val