mirror of
https://github.com/hanzoai/semantic.git
synced 2026-10-30 15:04:19 +00:00
A reply with no confidence was given 0.9, and an endpoint no entity pass found was given 0.8: numbers nobody measured, which a floor then kept as if they had been. An item the model did not score now has Score zero, which is documented to mean unscored, and the UNKNOWN placeholder carries none. The temporal confidence was already read that way; the fallback helper that existed for the other two is gone. Co-authored-by: Hanzo Dev <[email protected]>