Skip to content

LLM inference in C/C++

C++ ggml

server : reject prefilled assistant messages with tool calls (#27626)

Kyozzz1729ed5371cd
diff --git a/tools/server/server-common.cpp b/tools/server/server-common.cpp
index 4f5b8202ac..7997d4016a 100644
--- a/tools/server/server-common.cpp
+++ b/tools/server/server-common.cpp
@@ -1280,6 +1280,12 @@ json oaicompat_chat_params_parse(
if (inputs.continue_final_message != COMMON_CHAT_CONTINUATION_NONE && inputs.add_generation_prompt) {
throw std::invalid_argument("Cannot set both add_generation_prompt and continue_final_message to true.");
}
+ if (inputs.continue_final_message != COMMON_CHAT_CONTINUATION_NONE
+ && !inputs.messages.empty()
+ && inputs.messages.back().role == "assistant"
+ && !inputs.messages.back().tool_calls.empty()) {
+ throw std::invalid_argument("Cannot continue an assistant message that contains tool calls.");
+ }
inputs.reasoning_format = opt.reasoning_format;
if (body.contains("reasoning_format")) {
inputs.reasoning_format = common_reasoning_format_from_name(body.at("reasoning_format").get<std::string>());