DocEval

Fix Gemini model IDs, swap gpt-5.5 for gpt-4.1, add task/provider filters

Add CLAUDE.md for future Claude Code sessions

Build DocEval — LLM evaluation harness for legal docs

Initial commit