Enterprise-grade knowledge graph service using Neo4j GraphRAG for intelligent document analysis and semantic querying.
- Neo4j GraphRAG Integration: Official
neo4j-graphrag-pythonpackage (v1.10.1) - 3 Retrieval Methods:
- Vector similarity search (chunk-based)
- Hybrid search (vector + fulltext fusion)
- Text-to-Cypher (LLM-generated graph queries)
- Document Indexing: SimpleKGPipeline with automatic entity resolution
- Multi-tenancy: Group-based data isolation
- Azure OpenAI: GPT-4o + text-embedding-3-large (3072 dimensions)
- 91% Code Reduction: Replaced 1,636 lines with ~150 lines
- Azure subscription
- Neo4j Aura Pro instance
- Azure OpenAI service (GPT-4o + text-embedding-3-large)
- Azure CLI (
az) - Azure Developer CLI (
azd) - Python 3.11+
┌─────────────────────────────────────────┐
│ FastAPI Application (Port 8000) │
├─────────────────────────────────────────┤
│ Neo4j GraphRAG Service │
│ ├─ VectorCypherRetriever │
│ ├─ HybridCypherRetriever │
│ ├─ Text2CypherRetriever │
│ └─ SimpleKGPipeline (Indexing) │
├─────────────────────────────────────────┤
│ Azure OpenAI │
│ ├─ LLM: gpt-4o │
│ └─ Embeddings: text-embedding-3-large │
├─────────────────────────────────────────┤
│ Neo4j Aura Pro (Graph Database) │
│ └─ Group-aware multi-tenancy │
└─────────────────────────────────────────┘
# Clone repository
cd /afh/projects/graphrag-orchestration
# Create virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# Install dependencies
pip install -r graphrag-orchestration/requirements.txtCreate .env file:
# Azure OpenAI
AZURE_OPENAI_ENDPOINT=https://your-openai.openai.azure.com/
AZURE_OPENAI_API_KEY=your-api-key
AZURE_OPENAI_DEPLOYMENT_NAME=gpt-4o
AZURE_OPENAI_EMBEDDING_DEPLOYMENT=text-embedding-3-large
AZURE_OPENAI_EMBEDDING_DIMENSIONS=3072
AZURE_OPENAI_API_VERSION=2024-10-21
# Neo4j
NEO4J_URI=neo4j+s://your-instance.databases.neo4j.io
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=your-password
NEO4J_DATABASE=neo4j
# Multi-tenancy
ENABLE_GROUP_ISOLATION=truecd graphrag-orchestration
python -m uvicorn app.main:app --reload --port 8000API available at: http://localhost:8000
Docs available at: http://localhost:8000/docs
# Login to Azure
az login
azd auth login
# Deploy
azd up# Provision infrastructure
azd provision
# Deploy application
azd deployLocal Search (Vector Similarity)
POST /graphrag/v2/query/local
{
"query": "Who is the CEO of Acme Corporation?",
"top_k": 10
}Hybrid Search (Vector + Fulltext)
POST /graphrag/v2/query/hybrid
{
"query": "Financial performance in 2024",
"top_k": 10
}Structured Search (Text-to-Cypher)
POST /graphrag/v2/query/structured
{
"query": "Show all relationships for Jane Smith"
}Index Text
POST /graphrag/v2/index/text
{
"text": "Your document content...",
"document_name": "annual_report_2024.txt"
}All requests must include:
X-Group-ID: your-tenant-id
Content-Type: application/json
# Run tests
pytest graphrag-orchestration/tests/
# Test specific module
pytest graphrag-orchestration/tests/services/test_neo4j_graphrag_service.py -v
# Run with coverage
pytest --cov=app graphrag-orchestration/tests/- Code Reduction: 91% (1,636 → ~150 lines)
- Document Compression: 84.5% (4,382 → 678 words)
- Query Latency: Sub-second
- Embedding Quality: 3,072 dimensions (text-embedding-3-large)
All data is isolated by group_id:
- Neo4j nodes have
group_idproperty - All Cypher queries filter by partition key
- Cross-tenant data leaks prevented at database level
See graphrag-orchestration/app/core/config.py for all available settings.
Neo4j Connection Issues
# Test connection
python -c "from neo4j import GraphDatabase; driver = GraphDatabase.driver('neo4j+s://...', auth=('neo4j', 'password')); driver.verify_connectivity(); print('OK')"Azure OpenAI API Issues
# Check API version
curl https://your-openai.openai.azure.com/openai/deployments?api-version=2024-10-21Missing Dependencies
pip install --upgrade neo4j-graphrag-python==1.10.1This is a standalone service extracted from the Content Processing Solution Accelerator.
MIT License - See LICENSE file for details