feat: enhance LocalAGI installation and update scripts with source build and systemd service integration

frontend/public/json/localagi.json

@@ -2,8 +2,7 @@
   "name": "LocalAGI",
   "slug": "localagi",
   "categories": [
-    20,
-    3
+    20
   ],
   "date_created": "2026-03-03",
   "type": "ct",
@@ -13,7 +12,7 @@
   "documentation": "https://github.com/mudler/LocalAGI#installation-options",
   "website": "https://github.com/mudler/LocalAGI",
   "logo": "https://github.com/mudler/LocalAGI/raw/main/webui/react-ui/public/logo_1.png",
-  "config_path": "/opt/localagi/docker-compose.yaml",
+  "config_path": "/opt/localagi/.env",
   "description": "LocalAGI is a self-hostable AI agent platform with a web UI, OpenAI-compatible APIs, and local-first model orchestration.",
   "install_methods": [
     {
@@ -34,11 +33,15 @@
   },
   "notes": [
     {
-      "text": "GPU passthrough can be enabled during CT creation. Backend auto-selects (`cu128` for NVIDIA, `rocm7.2` for AMD) and falls back to CPU.",
+      "text": "This script builds LocalAGI from source (Go + Bun) and runs it as a systemd service.",
       "type": "info"
     },
     {
-      "text": "Set `var_localagi_backend=cpu|cu128|rocm7.2` (or `var_torch_backend`) to force a specific backend profile.",
+      "text": "GPU passthrough can be enabled during CT creation. `var_localagi_backend=cpu|cu128|rocm7.2` is recorded as `LOCALAGI_GPU_BACKEND` for your runtime setup.",
+      "type": "info"
+    },
+    {
+      "text": "By default, LocalAGI is configured to call an external OpenAI-compatible backend at `http://127.0.0.1:8081` via `LOCALAGI_LLM_API_URL`.",
       "type": "info"
     }
   ]