| # Core | |
| torch>=2.0 | |
| transformers | |
| datasets | |
| huggingface-hub | |
| # MCP + orchestration client stack | |
| mcp[cli]==1.27.0 | |
| openai-agents==0.14.5 | |
| litellm==1.81.13 | |
| # WatsonX.ai (LLM-as-Judge, scaling comparison) | |
| ibm-watsonx-ai | |
| # Profiling and tracking | |
| wandb | |
| tensorboard | |
| # Data | |
| pandas | |
| numpy | |
| scipy # statistical-fidelity validation: KS, AD, Wasserstein, chi-squared | |
| openpyxl # IEEE DataPort DGA dataset is published as .xlsx; pandas needs this engine to read it | |
| matplotlib | |
| seaborn | |
| # Reproducible notebook execution | |
| nbconvert | |
| nbclient | |
| ipykernel | |
| # Utilities | |
| tqdm | |
| pyyaml | |
| # LLM-as-Judge scoring (scripts/judge_trajectory.py) | |
| litellm | |
| python-dotenv | |