Run model pull as background task to avoid startup timeout

ensure_model() can take minutes to pull a multi-GB model, which
exceeds Hypercorn's startup timeout. Now fires as an asyncio
background task so the app starts immediately.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/fabledassistant/app.py

@@ -24,16 +24,22 @@ def create_app() -> Quart:
 
     @app.before_serving
     async def startup():
+        import asyncio
+
         from fabledassistant.services.llm import ensure_model
 
-        try:
+        async def _pull_model():
-            await ensure_model(Config.OLLAMA_MODEL)
+            try:
-        except Exception:
+                await ensure_model(Config.OLLAMA_MODEL)
-            logger.warning(
+            except Exception:
-                "Failed to ensure model '%s' on startup — chat may not work until model is available",
+                logger.warning(
-                Config.OLLAMA_MODEL,
+                    "Failed to ensure model '%s' — chat may not work until model is available",
-                exc_info=True,
+                    Config.OLLAMA_MODEL,
-            )
+                    exc_info=True,
+                )
+
+        # Fire-and-forget so model pull doesn't block startup
+        asyncio.create_task(_pull_model())
 
     @app.route("/")
     async def serve_index():