Clamp rewards strictly inside (0,1) with EPS=0.01

fix: clamp reward formatting to keep scores strictly in (0,1)

Clamp reset reward to (0,1) interval as well

Clamp task scores to open interval (0, 1)

Add inference.py at repo root

Add friendly root route

Deploy AccessibilityAudit-Env

Deploy AccessibilityAudit-Env

initial commit