[sippeangelo]

The biggest latency improvement I saw was switching off OpenAI's API that would have a latency anywhere between 0.3 - 6 seconds(!) for the same two word search embedding...