Manasi/removing llm judge consistency

[manasi-sharma]

Had to make the following changes to change the codebase to focus on the auto-eval aspect:

1. remove all code for PDF extraction (markdown only now)
2. evaluate_rubrics only on the markdown files saved in the agent_responses directory
3. remove calculate_metrics script for human alignment (not needed to run externally)

[socket-security]

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	huggingface-hub@​1.1.4

View full report