coreason-graph-nexus

The Graph Orchestration & Reasoning Engine

coreason-graph-nexus acts as the "Platinum Layer" Builder and Graph Logic Engine. It bridges the gap between persistent graph storage (Neo4j) and high-speed in-memory reasoning (NetworkX), ensuring robust data ingestion, ontology alignment, and advanced algorithmic analysis.

🚀 Features

• Hybrid Compute Architecture: Seamlessly moves data between Cold Storage (Neo4j) and Hot Compute (NetworkX) for on-demand analysis.
• Projection Engine (The Builder): Declarative ETL pipeline that transforms raw data into a semantic graph, enforcing schema constraints.
• Ontology Resolver (The Librarian): Standardizes entity identities (e.g., merging "Tylenol" and "APAP") with high-performance Redis caching.
• Graph Computer (The Thinker): Executes complex algorithms like PageRank, Betweenness Centrality, and Community Detection in memory.
• Link Predictor (The Analyst): Infers implicit relationships using both heuristic rules and semantic vector embeddings.
• Graph Logic Microservice (Service G): Exposes all engine capabilities via a high-performance REST API (FastAPI) for seamless integration into microservice architectures.

📦 Installation

pip install coreason-graph-nexus

🛠️ Usage

Here is a quick example of how to initialize the client and run a PageRank analysis:

from coreason_graph_nexus.adapters.neo4j_adapter import Neo4jClient
from coreason_graph_nexus.compute import GraphComputer
from coreason_graph_nexus.models import GraphAnalysisRequest, AnalysisAlgo

# 1. Initialize Connection
neo4j_auth = ("neo4j", "password")
with Neo4jClient(uri="bolt://localhost:7687", auth=neo4j_auth) as client:

    # 2. Initialize the Graph Computer
    computer = GraphComputer(client)

    # 3. Define Analysis Request
    request = GraphAnalysisRequest(
        center_node_id="RxNorm:123",
        algorithm=AnalysisAlgo.PAGERANK,
        depth=2,
        write_property="pagerank_score"
    )

    # 4. Run Analysis
    results = computer.run_analysis(request)
    print(f"Computed PageRank for {len(results)} nodes.")

Running as a Microservice

You can also deploy the engine as a standalone service:

# Start the API server
uvicorn coreason_graph_nexus.server:app --host 0.0.0.0 --port 8000

Refer to vignette.md for detailed API usage examples.

📄 License

This software is dual-licensed. It is available under the Prosperity Public License 3.0 for open-source and non-commercial use. Commercial use beyond a 30-day trial requires a separate license.

Copyright (c) 2025 CoReason, Inc.