Use absolute llama-server path in entrypoint

entrypoint.sh

@@ -9,7 +9,7 @@ THREADS="${THREADS:-2}"
 PARALLEL="${PARALLEL:-1}"
 REASONING_MODE="${REASONING_MODE:-off}"
 
-exec llama-server \
+exec /app/llama-server \
   -hf "$MODEL_SPEC" \
   --host "$HOST" \
   --port "$PORT" \