Hybrid search, graph traversal, and AI workflow orchestration — in one installation. No separate services. No glue code.
pip install purple8-graph# With DiskANN on-disk vector index (larger-than-RAM indexes)
pip install "purple8-graph[diskann]"
# With production monitoring (Prometheus + OTEL)
pip install "purple8-graph[prod]"
# With SAML/SSO
pip install "purple8-graph[saml]"docker run -d \
--name purple8-graph \
-p 8000:8000 \
-e JWT_SECRET=your-secret-min-32-chars \
-v purple8-data:/data \
purple8/purple8-graph:latestfrom purple8_graph import GraphEngine
import numpy as np
engine = GraphEngine("./my_graph")
# Add nodes — no schema required
engine.add_node("doc1", labels=["Document"], properties={
"title": "Understanding transformers",
"region": "APAC",
"embedding": np.random.rand(384).tolist(), # your real embedding
})
engine.add_node("alice", labels=["Person"], properties={"name": "Alice Chen"})
engine.add_edge("doc1", "alice", "AUTHORED_BY")
# Hybrid Cypher query — vector search + graph traversal in one round-trip
results = engine.execute_cypher("""
CALL db.vector.search('Document', $vec, 10) YIELD node, score
WHERE node.region = 'APAC' AND score > 0.80
MATCH (node)-[:AUTHORED_BY]->(author:Person)
RETURN node.title, author.name, score
ORDER BY score DESC
""", {"vec": query_embedding})Full quickstart → · Hybrid search guide → · Cypher reference →
| Purple8 | Neo4j + Vector | Kùzu + HNSW | FalkorDB | |
|---|---|---|---|---|
| Vector + graph in one query | ✅ | ❌ two phases | ❌ manual wiring | ❌ |
| Durable disk storage | ✅ RocksDB | ✅ | ✅ | ⚠️ in-memory |
| Data-first (no DDL) | ✅ | ✅ | ✅ | ✅ |
| AI workflow orchestration | ✅ Journey | ❌ | ❌ | ❌ |
| One installation — no server | ✅ | ❌ JVM | ✅ | ❌ Redis |
| p50 at 100k docs | 3.53 ms | 60–150 ms | 30–80 ms | <10 ms (RAM only) |