The 0.92 semantic similarity threshold is interesting. curious how you landed on that number? Thinking too low and you risk false cache hits but too high and you lose most of the semantic benefit. Did you run experiments or is it tunable per use case?
IxInfra•58m ago