vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge 01db429161 into d7b31a9d84

Browse source
This commit is contained in:
Olivier Chafik 2025-02-10 17:34:14 +08:00 • committed by GitHub
commit 41de60d728
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
17 changed files with 979 additions and 272 deletions
Download patch file Download diff file Expand all files Collapse all files

29
examples/server/server.cpp
View file

@ -173,6 +173,7 @@ struct slot_params {
{"grammar_trigger_words", grammar_trigger_words},
{"grammar_trigger_tokens", sampling.grammar_trigger_tokens},
{"preserved_tokens", sampling.preserved_tokens},
{"chat_format", common_chat_format_name(oaicompat_chat_format)},
{"samplers", samplers},
{"speculative.n_max", speculative.n_max},
{"speculative.n_min", speculative.n_min},
@ -724,9 +725,19 @@ struct server_task_result_cmpl_final : server_task_result {
msg.content = content;
}
json tool_calls;
json message {
{"role", "assistant"},
};
if (!msg.reasoning_content.empty()) {
message["reasoning_content"] = msg.reasoning_content;
}
if (msg.content == "" && !msg.tool_calls.empty()) {
message["content"] = json();
} else {
message["content"] = msg.content;
}
if (!msg.tool_calls.empty()) {
tool_calls = json::array();
auto tool_calls = json::array();
for (const auto & tc : msg.tool_calls) {
tool_calls.push_back({
{"type", "function"},
@ -737,15 +748,7 @@ struct server_task_result_cmpl_final : server_task_result {
{"id", tc.id},
});
}
}
json message {
{"content", msg.content},
{"tool_calls", tool_calls},
{"role", "assistant"},
};
if (!msg.tool_plan.empty()) {
message["tool_plan"] = msg.tool_plan;
message["tool_calls"] = tool_calls;
}
json choice {
@ -4056,7 +4059,7 @@ int main(int argc, char ** argv) {
}
auto body = json::parse(req.body);
json data = oaicompat_completion_params_parse(body, params.use_jinja, ctx_server.chat_templates);
json data = oaicompat_completion_params_parse(body, params.use_jinja, params.reasoning_format, ctx_server.chat_templates);
return handle_completions_impl(
SERVER_TASK_TYPE_COMPLETION,
@ -4069,7 +4072,7 @@ int main(int argc, char ** argv) {
// same with handle_chat_completions, but without inference part
const auto handle_apply_template = [&ctx_server, &params, &res_ok](const httplib::Request & req, httplib::Response & res) {
auto body = json::parse(req.body);
json data = oaicompat_completion_params_parse(body, params.use_jinja, ctx_server.chat_templates);
json data = oaicompat_completion_params_parse(body, params.use_jinja, params.reasoning_format, ctx_server.chat_templates);
res_ok(res, {{ "prompt", std::move(data.at("prompt")) }});
};