Fix: try pre-built wheel first, fall back to source build

app.py

@@ -426,10 +426,18 @@ def _run_benchmark():
     except ImportError:
         _log("Benchmark: installing llama-cpp-python...")
         import subprocess
-        subprocess.check_call(
-            [sys.executable, "-m", "pip", "install", "llama-cpp-python>=0.4.0", "--only-binary=:all:"],
-            stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
-        )
+        # Try pre-built wheels first, fall back to source build
+        try:
+            subprocess.check_call(
+                [sys.executable, "-m", "pip", "install", "llama-cpp-python>=0.4.0", "--only-binary=:all:"],
+                stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+            )
+        except subprocess.CalledProcessError:
+            _log("Benchmark: no pre-built wheel, building from source...")
+            subprocess.check_call(
+                [sys.executable, "-m", "pip", "install", "llama-cpp-python>=0.4.0"],
+                stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+            )
         from llama_cpp import Llama
 
     os.makedirs(_BENCHMARK_CACHE, exist_ok=True)