small change

src/backends/llamacpp/rac_backend_llamacpp_register.cpp

@@ -12,6 +12,8 @@
 #include <cstring>
 #include <mutex>
 
+#include <nlohmann/json.hpp>
+
 #include "rac/core/rac_core.h"
 #include "rac/core/rac_error.h"
 #include "rac/core/rac_logger.h"
@@ -66,10 +68,28 @@ static rac_result_t llamacpp_vtable_generate_stream(void* impl, const char* prom
 static rac_result_t llamacpp_vtable_get_info(void* impl, rac_llm_info_t* out_info) {
     if (!out_info)
         return RAC_ERROR_NULL_POINTER;
+
     out_info->is_ready = rac_llm_llamacpp_is_model_loaded(impl);
     out_info->supports_streaming = RAC_TRUE;
     out_info->current_model = nullptr;
-    out_info->context_length = 0;
+    out_info->context_length = 0;  // Default if model not loaded or info unavailable
+
+    // Get actual context_length from model info JSON when model is loaded
+    if (out_info->is_ready) {
+        char* json_str = nullptr;
+        if (rac_llm_llamacpp_get_model_info(impl, &json_str) == RAC_SUCCESS && json_str) {
+            try {
+                auto json = nlohmann::json::parse(json_str);
+                if (json.contains("context_size") && json["context_size"].is_number()) {
+                    out_info->context_length = json["context_size"].get<int32_t>();
+                }
+            } catch (...) {
+                // JSON parse error - context_length remains 0
+            }
+            free(json_str);
+        }
+    }
+
     return RAC_SUCCESS;
 }