thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

server : support preserving reasoning_content in assistant message (#18994)

Browse source 
* support reasoning_content input

* report template caps to webui

* add docs

* rm commented code
This commit is contained in:
Xuan-Son Nguyen 2026-01-22 21:30:06 +01:00 • committed by GitHub
parent a5eaa1d6a3
commit 51fa458a92
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
10 changed files with 165 additions and 131 deletions
Download patch file Download diff file Expand all files Collapse all files

16
tests/test-chat.cpp
View file

@ -462,9 +462,9 @@ static void test_parser_with_streaming(const common_chat_msg & expected, const s
for (size_t i = 1; i <= raw_message.size(); ++i) {
auto curr_msg = parse_msg(std::string(utf8_truncate_safe_view(std::string_view(raw_message).substr(0, i))));
if (curr_msg == simple_assist_msg("")) continue;
LOG_INF("Streaming msg: %s\n", common_chat_msgs_to_json_oaicompat<json>({curr_msg}).dump().c_str());
LOG_INF("Streaming msg: %s\n", common_chat_msgs_to_json_oaicompat({curr_msg}).dump().c_str());
for (auto diff: common_chat_msg_diff::compute_diffs(last_msg, curr_msg)) {
LOG_INF("Streaming diff: %s\n", common_chat_msg_diff_to_json_oaicompat<json>(diff).dump().c_str());
LOG_INF("Streaming diff: %s\n", common_chat_msg_diff_to_json_oaicompat(diff).dump().c_str());
if (!diff.reasoning_content_delta.empty()) {
merged.reasoning_content += diff.reasoning_content_delta;
}
@ -480,7 +480,7 @@ static void test_parser_with_streaming(const common_chat_msg & expected, const s
merged.tool_calls.back().arguments += diff.tool_call_delta.arguments;
}
}
LOG_INF("Streaming merged: %s\n", common_chat_msgs_to_json_oaicompat<json>({merged}).dump().c_str());
LOG_INF("Streaming merged: %s\n", common_chat_msgs_to_json_oaicompat({merged}).dump().c_str());
}
assert_msg_equals(curr_msg, merged, true);
last_msg = curr_msg;
@ -622,7 +622,7 @@ static void test_msgs_oaicompat_json_conversion() {
message_assist_call_code_interpreter,
};
for (const auto & msg : msgs) {
auto oai_json = common_chat_msgs_to_json_oaicompat<json>({msg});
auto oai_json = common_chat_msgs_to_json_oaicompat({msg});
auto msgs2 = common_chat_msgs_parse_oaicompat(oai_json);
assert_equals((size_t) 1, msgs2.size());
auto msg2 = msgs2[0];
@ -646,7 +646,7 @@ static void test_msgs_oaicompat_json_conversion() {
" }\n"
"]"
),
common_chat_msgs_to_json_oaicompat<json>({message_user_parts}).dump(2));
common_chat_msgs_to_json_oaicompat({message_user_parts}).dump(2));
assert_equals(
std::string(
@ -666,7 +666,7 @@ static void test_msgs_oaicompat_json_conversion() {
" }\n"
"]"
),
common_chat_msgs_to_json_oaicompat<json>({message_assist_call_python}).dump(2));
common_chat_msgs_to_json_oaicompat({message_assist_call_python}).dump(2));
auto res = common_chat_msgs_parse_oaicompat(json::parse("[{\"role\": \"assistant\", \"tool_calls\": []}]"));
assert_equals<size_t>(1, res.size());
@ -693,7 +693,7 @@ static void test_tools_oaicompat_json_conversion() {
};
for (const auto & tool : tools) {
auto oai_json = common_chat_tools_to_json_oaicompat<json>({tool});
auto oai_json = common_chat_tools_to_json_oaicompat({tool});
auto tools2 = common_chat_tools_parse_oaicompat(oai_json);
assert_equals((size_t) 1, tools2.size());
auto tool2 = tools2[0];
@ -726,7 +726,7 @@ static void test_tools_oaicompat_json_conversion() {
" }\n"
"]"
),
common_chat_tools_to_json_oaicompat<json>({special_function_tool}).dump(2));
common_chat_tools_to_json_oaicompat({special_function_tool}).dump(2));
{
auto tools_no_params = common_chat_tools_parse_oaicompat(json::parse(