fix: set keep_alive to 2h on all Ollama requests

Prevents models from sitting in VRAM indefinitely. Applies to both
streaming chat calls and the non-streaming generate_completion path,
as well as the startup warm-up request.

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

src/fabledassistant/app.py

@@ -169,7 +169,7 @@ def create_app() -> Quart:
                 async with httpx.AsyncClient(timeout=300.0) as client:
                     await client.post(
                         f"{Config.OLLAMA_URL}/api/generate",
-                        json={"model": model, "prompt": "", "keep_alive": "30m"},
+                        json={"model": model, "prompt": "", "keep_alive": "2h"},
                     )
                 logger.info("Warmed model '%s' into VRAM", model)
             except Exception: