server : move chat-template thinking probe inside the init try/catch (#24093)

A model whose chat template parses at init but fails parser generation
at apply time (e.g. uses {% call %}) throws std::invalid_argument from
common_chat_templates_support_enable_thinking(), which ran outside the
try/catch guarding common_chat_templates_init(). The throw was uncaught
and llama-cli aborted (SIGABRT) instead of failing to load. Moved the
probe inside that try/catch so an apply-time error fails load the same
way an init parse error does.

Signed-off-by: Jesse LaRose <redacted>

diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cpp
index eb41acc5b491620027d8dc04c58868578db549b4..98d0cca1ccda7f311325c5141d99953707fb7bfb 100644 (file)
--- a/tools/server/server-context.cpp
+++ b/tools/server/server-context.cpp
@@ -1444,6 +1444,7 @@ private:
         // populate chat template params
         {
             common_chat_templates_ptr chat_templates;
+            bool enable_thinking = false;
 
             try {
                 chat_templates = common_chat_templates_init(model_tgt, params_base.chat_template);
@@ -1451,6 +1452,12 @@ private:
                 SRV_TRC("%s: chat template, example_format: '%s'\n", __func__,
                     common_chat_format_example(chat_templates.get(), params_base.use_jinja, params_base.default_template_kwargs).c_str());
 
+                // thinking is enabled if:
+                // 1. It's not explicitly disabled via --reasoning off
+                // 2. The chat template supports it
+                const bool template_supports_thinking = params_base.use_jinja && common_chat_templates_support_enable_thinking(chat_templates.get());
+                enable_thinking = params_base.enable_reasoning != 0 && template_supports_thinking;
+                SRV_TRC("%s: chat template, thinking = %d\n", __func__, enable_thinking);
             } catch (const std::exception & e) {
                 SRV_ERR("%s: chat template parsing error: %s\n", __func__, e.what());
                 SRV_ERR("%s: please consider disabling jinja via --no-jinja, or use a custom chat template via --chat-template\n", __func__);
@@ -1458,13 +1465,6 @@ private:
                 return false;
             }
 
-            // thinking is enabled if:
-            // 1. It's not explicitly disabled via --reasoning off
-            // 2. The chat template supports it
-            const bool template_supports_thinking = params_base.use_jinja && common_chat_templates_support_enable_thinking(chat_templates.get());
-            const bool enable_thinking = params_base.enable_reasoning != 0 && template_supports_thinking;
-            SRV_TRC("%s: chat template, thinking = %d\n", __func__, enable_thinking);
-
             // IMPORTANT: chat_params is reused across sleeping / resuming states,
             //            never store llama_context/llama_model pointers in chat_params,
             //            as they may be invalidated after sleeping