NEExT 0.3.3

Network Embedding Experimentation Toolkit - A powerful framework for graph analysis, embedding computation, and machine learning on graph-structured data

NEExT

Network Embedding Experimentation Toolkit

An open-source Python framework for network science and graph machine learning — one coherent pipeline from graph data to research evidence.

Documentation · Website · Issues

What is NEExT?

NEExT turns graph-structured data into machine-learning-ready evidence through one transparent, reproducible pipeline:

Graphs → Features → Embeddings → Evidence

Load graphs from CSV, pandas, or NetworkX into a unified GraphCollection; compute structural node features (or your own in plain Python); generate graph-level embeddings via Wasserstein optimal transport or graph neural networks; then train classifiers/regressors and extract feature importance. It runs on trusted Python libraries (NumPy, pandas, scikit-learn, XGBoost, NetworkX, iGraph) and works the same way in a script, a notebook, or the Workbench.

There are two ways to use NEExT:

• The Library — a lightweight Python package for scripting and notebook workflows.
• The Workbench — a local, desktop-style GUI over real NEExT workflows, with versioned artifacts and job tracking. It runs entirely on your machine (127.0.0.1, no accounts, no uploads) and is MCP-native, so agents like Claude can drive it.

Installation

pip install NEExT

Optional extras:

pip install "NEExT[gnn]"            # Graph Neural Network embeddings (pure PyTorch)
pip install "NEExT[workbench-mcp]"  # local Workbench + MCP integration

See the docs for the full list of extras.

Quick start

from NEExT import NEExT

nxt = NEExT()

# Load a collection of graphs from CSV
graph_collection = nxt.read_from_csv(
    edges_path="edges.csv",
    node_graph_mapping_path="node_graph_mapping.csv",
    graph_label_path="graph_labels.csv",
)

# Features → Embeddings → Evidence
features = nxt.compute_node_features(graph_collection, feature_list=["all"])
embeddings = nxt.compute_graph_embeddings(
    graph_collection, features, embedding_algorithm="approx_wasserstein"
)
results = nxt.train_ml_model(graph_collection, embeddings, model_type="classifier")

Custom features, GNN embeddings, large-graph sampling, feature importance, and the full API are covered in the documentation.

The Workbench

The NEExT Workbench is a local, single-user FastAPI + React application that exposes real NEExT workflows — datasets, features, embeddings, models, and analysis — as a desktop-style UI. Everything stays on your machine, and it speaks MCP so you can drive it from agentic clients.

neext-workbench          # installed package
make neext-workbench     # from a development checkout

Then open http://127.0.0.1:8765. Projects are stored under ~/NEExT-Workbench by default (override with NEEXT_WORKBENCH_HOME or neext-workbench --workspace <path>). The full Workbench tour, including MCP client setup, lives in the documentation.

Learn more

• 📚 Documentation — guides, API reference, and the Workbench tour: neext.app/docs
• 🌐 Website — neext.app
• 🐛 Issues & support — github.com/ashdehghan/NEExT/issues

License

NEExT is released under the MIT License. Created and maintained by Ash Dehghan.