#local-ai
Measuring why local AI felt slow
A slow greeting pushed me to separate model loading, instruction processing and queued work from response generation.
A slow greeting pushed me to separate model loading, instruction processing and queued work from response generation.