This is the Guanaco Extended dataset derived from OpenAssistant/oasst1.
Guanaco only uses the first (highest rank) response from the assistant as their dataset, but, we went ahead and collapsed the message tree to the highest ranks and coverted this to a format similar to sharegpt datasets found on HF. The python script is included as well.