Match passing submission pattern - score clamped, top level client

Clamp reward to strictly between 0 and 1 as required

Final commit :

FIX 2 files

Updated inference.py 2nd time;)

Updated inference.py ;)

Fix: Single task per invocation as required :)

Fixed HF_TOKEN has no default value as required :)

Fix END log format to match sample exactly :)

added mandatory stdout format :)

Restored load_dotenv :)

fixed hardcoded api key :)

Add expert tasks 6-8

Add tasks 4 and 5, fix inference with dotenv

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

agent setup

Initial submission