01-cache.diff 576 Bytes
Newer Older
1
diff --git a/examples/server/server.cpp b/examples/server/server.cpp
2
index f255ad76..914ecfdd 100644
3
4
--- a/examples/server/server.cpp
+++ b/examples/server/server.cpp
5
@@ -1101,12 +1101,13 @@ struct server_context {
6
7
                 // add the token to slot queue and cache
             }
8
 
9
-            slot.add_token_string(result);
10
             if (slot.params.stream) {
11
12
13
14
15
                 send_partial_response(slot, result);
             }
         }
 
+        slot.add_token_string(result);
16
+
17
         if (incomplete) {
18
             slot.has_next_token = true;
19
         }