MarieCP

Lightweight runtime for querying The World Avatar (TWA) knowledge graphs via MCP (Model Context Protocol) tools.

The mini_marie Python package covers three domains:

Domain	Folder	MCP servers
Zaha (buildings & cities)	mini_marie/zaha/	twa-city, sg-old
Marie (chemistry)	mini_marie/marie/	chemistry-* (7 servers)
MOP/MOF (frameworks)	mini_marie/mop_mof/	mof-twa, twa-mops

Cross-domain tooling (kg_catalog/, kgqa/, cross_kg_competency/) lives under mini_marie/.

Project layout

MarieCP/
├── mini_marie/          # Python package, docs, scripts
├── configs/             # MCP JSON configs (add before deploy)
├── data/                # Runtime SQLite caches (gitignored)
├── evaluation/data/     # Local MOP RDF corpus (gitignored)
├── notebooks/           # Jupyter notebooks
└── requirements-mini-marie.txt

See mini_marie/README.md for full documentation, quick start, and deployment.

Quick start

pip install -r requirements-mini-marie.txt   # when available
export PYTHONPATH="$(pwd)"
python -m mini_marie.mop_mof.mof.main

Runtime caches are written to data/mini_marie_cache/ (override with MINI_MARIE_DATA_DIR).