From c6c4f7c41f2ebb8db91db3193244be7cecb1515e Mon Sep 17 00:00:00 2001
From: ExtReMLapin <3909752+ExtReMLapin@users.noreply.github.com>
Date: Mon, 11 Aug 2025 18:21:59 +0200
Subject: [PATCH] Update chat.cpp to support (at least) qwen3 + tool_choice =
 required

---
 common/chat.cpp | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/common/chat.cpp b/common/chat.cpp
index b1a1218ca2..7ac3686e12 100644
--- a/common/chat.cpp
+++ b/common/chat.cpp
@@ -1586,8 +1586,9 @@ static common_chat_params common_chat_params_init_hermes_2_pro(const common_chat
     }
 
     if (!inputs.tools.is_null()) {
+        auto supports_thinking = tmpl.source().find("<think>") != std::string::npos && data.thinking_forced_open == false;
         // (content)?(<tool_call>{"name": "foo", "arguments": {"a": 1}}</tool_call>)*
-        data.grammar_lazy = inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_REQUIRED;
+        data.grammar_lazy = inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_REQUIRED || supports_thinking;
         data.grammar = build_grammar([&](const common_grammar_builder & builder) {
             std::vector<std::string> tool_rules;
             std::vector<std::string> tool_call_alts;
@@ -1639,9 +1640,19 @@ static common_chat_params common_chat_params_init_hermes_2_pro(const common_chat
             tool_call_alts.push_back(
                 "( \"```\\n\" | \"```json\\n\" | \"```xml\\n\" ) space " + wrappable_tool_call + " space \"```\" space ");
             auto tool_call = builder.add_rule("tool_call", string_join(tool_call_alts, " | "));
-            builder.add_rule("root",
+            if (supports_thinking)
+            {
+                builder.add_rule("thinking", "\"<think>\" [^\\x00]* \"</think>\" space");
+                builder.add_rule("root", 
+                "(thinking)? space " + 
+                (inputs.parallel_tool_calls ? "(" + tool_call + ")+" : tool_call));
+            }
+            else
+            {
+                builder.add_rule("root",
                 std::string(data.thinking_forced_open ? "( \"</think>\" space )? " : "") +
                 (inputs.parallel_tool_calls ? "(" + tool_call + ")+" : tool_call));
+            }
             // Trigger on some common known "good bad" outputs (only from the start and with a json that's about a specific argument name to avoid false positives)
             data.grammar_triggers.push_back({
                 COMMON_GRAMMAR_TRIGGER_TYPE_PATTERN_FULL,