What it does

Takes invoice-style Excel spreadsheets, chews them through pandas, and spits out Neo4j nodes and relationships. The original code targets py2neo v3 and Neo4j 3.x — think Python 3.6.5 and Windows 10 era. A newer additive example swaps in the official neo4j driver, vector indexes, and GraphRAG-style retrieval for anyone who actually wants to ship something in 2026.

The interesting bit

The repo is deliberately frozen in amber as a legacy educational baseline — pinned old dependencies, hard-coded paths, warts and all — while modern work happens in a parallel examples/modern_invoice_graphrag/ directory. There’s even a Codex/agent skill tucked in skills/neo4j-knowledge-graph/ for AI coding assistants that need to generate safe Cypher or pick between ancient and current tooling.

Key highlights

• Original path: Excel → pandas → py2neo v3 → Neo4j 3.x, plus matrix extraction for ML experiments
• Modern path: CSV → official neo4j driver → vector indexes + GraphRAG retrieval, with local deterministic embeddings for CI
• Includes Invoice_data_Demo.xls and a profile_table.py script for inspecting tabular data before graph modeling
• Maintenance restarted June 2026; modernization tracked separately so the legacy baseline stays intact
• MIT licensed, with issue templates that politely demand your Python/Neo4j/py2neo versions

Caveats

• The requirements.txt dependencies are intentionally legacy; running this on modern Python/pandas will likely require tweaks
• Hard-coded os.chdir('xxxx') paths and placeholder Graph(...) credentials need manual fixing before the original scripts work
• The modern example is additive, not a migration — the old and new paths coexist, not replace each other

Verdict

Grab this if you’re teaching (or learning) knowledge graph basics from first principles, or if you need a Codex skill for AI-assisted Neo4j schema design. Skip it if you want a production-ready, batteries-included GraphRAG framework — this is a tutorial with ambition, not a product.