# Minimal deps for evaluation scripts.
# LoCoMo scoring: bert-score, nltk, regex, numpy; run: python -c "import nltk; nltk.download('punkt')"
# Locomo-Plus judge: openai (or use Ollama via OPENAI_BASE_URL)
requests>=2.28.0
tqdm>=4.64.0
bert-score>=0.3.13
nltk>=3.8
regex
numpy
openai>=1.0
