Introduction

CoordiNode is a graph-native hybrid retrieval engine for AI and GraphRAG workloads.

It combines three retrieval modalities in one engine:

• Graph traversal — variable-length paths, pattern matching, aggregation
• Vector similarity — HNSW index, SQ8 quantization, cosine/L2/dot metrics
• Full-text search — BM25 scoring, 23+ languages, fuzzy/phrase/wildcard queries

One query language (OpenCypher-compatible), one transaction model (MVCC, Snapshot Isolation).

Is CoordiNode Right for You?

Use this today if you are building:

• GraphRAG — knowledge retrieval, relationship-aware AI
• Fraud detection — ring detection through shared-device graphs + behavioral embedding similarity
• Semantic recommendations — traverse social graphs, filter by semantic similarity
• Threat intelligence — correlate attack patterns with MITRE ATT&CK + vector + text search

Not yet ready for:

• 100% drop-in Neo4j Enterprise replacement (gRPC and REST are available now; the Bolt protocol is planned)
• APOC procedures, Neo4j Browser/Bloom, or GDS
• A cluster with years of production mileage. Raft replication, follower reads and mutual TLS ship today under an automated fault-injection suite; the operational history behind them is still short
• Datasets larger than one machine holds. The open-source edition replicates the full dataset to every node; horizontal sharding is an Enterprise feature

Next Steps

• Quick Start — Docker → seed data → first hybrid query in 5 minutes
• OpenCypher Extensions — vector_distance(), text_match(), point.distance(), EXPLAIN SUGGEST
• Python SDK — pip install coordinode
• LlamaIndex integration — PropertyGraphIndex with CoordiNode backend
• LangChain integration — GraphCypherQAChain