From 88f98c891de8b87a0ec432a5ad7a750a26e97097 Mon Sep 17 00:00:00 2001
From: Samuel Oliveira Alves
 <107287165+SamuelOliveirads@users.noreply.github.com>
Date: Tue, 17 Feb 2026 08:33:28 -0300
Subject: [PATCH] server: add string ban in speculative path (#1274)

---
 examples/server/server-context.cpp | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/examples/server/server-context.cpp b/examples/server/server-context.cpp
index 43d9c0a9..81fcf060 100644
--- a/examples/server/server-context.cpp
+++ b/examples/server/server-context.cpp
@@ -2996,13 +2996,17 @@ void server_context::speculative_decoding_accept() {
                 populate_token_probs(slot, result, slot.params.post_sampling_probs, params_base.special, i);
             }
 
-            if (!process_token(result, slot)) {
-                // release slot because of stop condition
-                send_final_response(slot);
-                slot.release();
-                slot.print_timings();
-                metrics.on_prediction(slot);
-                break;
+            if (slot.n_buffer == 0) {
+                if (!process_token(result, slot)) {
+                    // release slot because of stop condition
+                    send_final_response(slot);
+                    slot.release();
+                    slot.print_timings();
+                    metrics.on_prediction(slot);
+                    break;
+                }
+            } else {
+                buffer_and_check_string_ban(slot, result);
             }
         }
         SLT_DBG(slot, "accepted %d/%d draft tokens, new n_tokens = %d\n", (int)ids.size() - 1, (int)slot.drafted.size(), slot.n_past);