gibbrdev/gibsgraph
Natural language to Neo4j — query and build knowledge graphs automatically
What's novel
Natural language to Neo4j — query and build knowledge graphs automatically
Code Analysis
12 files read · 4 roundsA Python library that connects to Neo4j, auto-discovers graph schema, and uses LLM-powered NL-to-Cypher generation (with vector search fallback) to answer natural language questions about knowledge graphs, plus text-to-graph ingestion via neo4j-graphrag and a 4-stage schema validation pipeline for t
Strengths
Genuinely novel PCST subgraph pruning for vector search neighborhoods, clean dual-strategy retrieval (vector + text-to-Cypher with self-healing retry), and a well-designed 4-stage validation pipeline with enterprise severity levels. Security is thoughtfully handled with parameterized Cypher, injection pattern detection, and read-only enforcement throughout.
Weaknesses
The KG builder is primarily a thin wrapper around neo4j-graphrag's SimpleKGPipeline rather than a novel implementation, and the agent pipeline is a simple linear LangGraph chain without complex routing or iterative refinement. The GNN retriever appears to be a stub (training data ready, model pending per README), and the bundled expert store uses basic keyword-overlap scoring rather than true semantic search.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
51
Contributors
2
Files
94
Active weeks
4
Repository
Language
Python
Stars
1
Forks
3
License
MIT