mirror of
https://github.com/wolfpld/tracy
synced 2026-01-19 04:52:09 +00:00
Request prompt caching.
This commit is contained in:
@@ -948,6 +948,7 @@ void TracyLlm::SendMessage( std::unique_lock<std::mutex>& lock )
|
|||||||
req["model"] = m_api->GetModels()[m_modelIdx].name;
|
req["model"] = m_api->GetModels()[m_modelIdx].name;
|
||||||
req["messages"] = std::move( chat );
|
req["messages"] = std::move( chat );
|
||||||
req["stream"] = true;
|
req["stream"] = true;
|
||||||
|
req["cache_prompt"] = true;
|
||||||
req["tools"] = m_toolsJson;
|
req["tools"] = m_toolsJson;
|
||||||
if( m_setTemperature ) req["temperature"] = m_temperature;
|
if( m_setTemperature ) req["temperature"] = m_temperature;
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user