Fix all reward=0.0 cases - clamp everything to 0.001 minimum

Clamp reward in step endpoint strictly between 0 and 1

Clamp reward to strictly between 0 and 1 as required

Fix reset - handle empty body, task_id defaults to 1

Fix task 8 answer query to match agent output exactly

Fix floating point precision in task 8

Fix normalize function

Fix normalize function for floating point comparison

Fix task 6 month format and task 8 rounding

Add expert tasks 6-8

Restore clean files , removed gradio, added task 4 and 5 :)

agent setup

Initial submission