fix: Update output instructions in prompts.py to enforce case sensitivity and structure, inference.py made DRY, PROJECT works

refactor: Reward clamping in graders

refactor: Update ActionType to include costs and modified grader for task 1

refactor: Update grading logic and submission handling across tasks for improved accuracy and consistency

fix: Update file paths and ensure model loading in PropertyRetriever

task1, task2 evaluated

task3 reviewed

task2 reviewed, semanticmatcher implemented

separated actions, prompts

New matching logic for grader

Inital Commit