HeliosDB GraphRAG HTAP - Complete User Guide
HeliosDB GraphRAG HTAP - Complete User Guide
Version: 1.0
Table of Contents
1. Introduction
What is GraphRAG HTAP?
HeliosDB GraphRAG HTAP combines:
- Graph Database: Native property graph with Cypher and GQL support
- Vector Database: Integrated embeddings for semantic search
- RAG Framework: Built-in Retrieval-Augmented Generation
- HTAP Engine: Hybrid Transactional/Analytical Processing
Key Benefits
- 10x Faster: Outperforms Neo4j + VectorDB combinations
- Unified Platform: Single system vs. fragmented architecture
- operationally hardened: WAL, backup/restore, replication, PITR
- ACID Compliant: Full MVCC with multiple isolation levels
- Scalable: Tested with 10M+ nodes, 100M+ edges
Use Cases
-
Knowledge Graphs with LLM Integration
- Build intelligent chatbots with graph-backed knowledge
- Implement RAG pipelines with relationship-aware retrieval
- Combine structured and semantic search
-
Real-Time Analytics
- OLTP queries for user interactions
- OLAP queries for business intelligence
- Automatic routing based on query complexity
-
Graph Machine Learning
- Node/edge embeddings with graph structure
- Community detection and influence analysis
- Recommendation systems with graph context
2. Core Concepts
2.1 Property Graph Model
HeliosDB uses the property graph model:
Nodes (vertices):
- Unique ID
- Label(s)
- Properties (key-value pairs)
Edges (relationships):
- Unique ID
- Source and target nodes
- Type/label
- Weight (for weighted graphs)
- Properties
Example:
(:Person {name: "Alice", age: 30})-[:KNOWS {since: 2020}]->(:Person {name: "Bob"})2.2 MVCC (Multi-Version Concurrency Control)
Every modification creates a new version.
Isolation Levels:
- Read Committed: See committed changes
- Repeatable Read: Consistent snapshot
- Serializable: Full serializability (with conflict detection)
2.3 HTAP Query Routing
Queries are automatically routed:
OLTP (low latency):
- Point queries (single node/edge lookup)
- Short paths
- Small result sets
OLAP (high throughput):
- Aggregations (COUNT, AVG, SUM)
- Long traversals
- Graph algorithms
Hybrid:
- Mixed workloads
- Adaptive execution
3. Cypher Query Language
3.1 Basic Queries
MATCH: Find patterns
-- Find all personsMATCH (p:Person) RETURN p
-- Find friendsMATCH (a:Person)-[:KNOWS]->(b:Person)RETURN a.name, b.name
-- Variable-length pathsMATCH (a:Person)-[:KNOWS*1..3]->(b:Person)WHERE a.name = 'Alice'RETURN b.nameCREATE: Insert data
-- Create nodeCREATE (p:Person {name: 'Charlie', age: 25})
-- Create relationshipMATCH (a:Person {name: 'Alice'}), (b:Person {name: 'Bob'})CREATE (a)-[:KNOWS {since: 2020}]->(b)UPDATE: Modify data
-- Update propertiesMATCH (p:Person {name: 'Alice'})SET p.age = 31, p.city = 'NYC'
-- Add labelMATCH (p:Person {name: 'Alice'})SET p:EmployeeDELETE: Remove data
-- Delete relationshipMATCH (a:Person)-[r:KNOWS]->(b:Person)WHERE a.name = 'Alice'DELETE r
-- Delete node (and relationships)MATCH (p:Person {name: 'Charlie'})DETACH DELETE p3.2 Advanced Cypher
Aggregations:
-- Count nodesMATCH (p:Person) RETURN count(p)
-- Average ageMATCH (p:Person) RETURN avg(p.age)
-- Group byMATCH (p:Person)RETURN p.city, count(p), avg(p.age)Ordering and Limiting:
MATCH (p:Person)RETURN p.name, p.ageORDER BY p.age DESCLIMIT 10SKIP 5Conditional Logic:
MATCH (p:Person)RETURN p.name, CASE WHEN p.age < 18 THEN 'Minor' WHEN p.age >= 18 AND p.age < 65 THEN 'Adult' ELSE 'Senior' END AS ageGroupSubqueries:
MATCH (p:Person)WHERE EXISTS { MATCH (p)-[:KNOWS]->(:Person {city: 'NYC'})}RETURN p.name3.3 Performance Tips
- Use LIMIT Early:
-- GoodMATCH (p:Person)WHERE p.age > 18RETURN pLIMIT 10
-- Less efficientMATCH (p:Person)RETURN pWHERE p.age > 18LIMIT 104. GQL Support
HeliosDB supports ISO GQL (Graph Query Language), the new standard.
4.1 Basic GQL Queries
-- Select nodesSELECT *FROM GRAPH myGraphMATCH (p:Person)WHERE p.age > 18
-- Path queriesSELECT p.name, f.nameFROM GRAPH myGraphMATCH (p:Person)-[:KNOWS]->(f:Person)WHERE p.name = 'Alice'4.2 GQL vs Cypher
| Feature | Cypher | GQL |
|---|---|---|
| Standard | Industry | ISO Standard |
| Syntax | MATCH-based | SELECT-based |
| Learning Curve | Low | Medium |
| Compatibility | Neo4j-like | SQL-like |
When to use GQL:
- You prefer SQL-style syntax
- Need ISO standard compliance
- Working with tools expecting GQL
When to use Cypher:
- Migrating from Neo4j
- Prefer graph-native syntax
- Shorter, more concise queries
5. HTAP Architecture
5.1 How HTAP Works
Query | v +---------------+ | Query Router | +---------------+ / \ / \ v v +----------+ +----------+ | OLTP | | OLAP | | (Row) | | (Column) | +----------+ +----------+ \ / \ / v v +-------------+ | MVCC Storage| +-------------+6. Performance Tuning
6.1 Performance Targets
Indicative figures, measured on representative fixtures; reproduce on your own hardware.
| Metric | Target | Typical |
|---|---|---|
| Simple query latency | <10ms | 2-5ms |
| Complex query latency | <100ms | 30-80ms |
| Throughput (cached) | 1000 QPS | 2000+ QPS |
| Throughput (uncached) | 500 QPS | 800 QPS |
| Node insertion | <5ms | 1-3ms |
| Transaction commit | <10ms | 3-7ms |
Conclusion
This user guide covers the essentials of HeliosDB GraphRAG HTAP. For more information:
- API Documentation: https://heliosdb.com/docs/full/
- Support: support@heliosdb.com
- Community: Discord
Production Checklist:
- Configure appropriate index strategy
- Enable WAL and configure checkpointing
- Set up backup schedule
- Configure replication for HA
- Implement monitoring and alerting
- Performance test with production workload
- Review security configuration
Version: 7.0.0