Graph Stores
Graph databases in the MATIH Platform store knowledge graphs, data lineage, and entity relationships. They enable the AI Service to understand data context, trace data provenance, and resolve business terminology.
Graph Store Options
| Technology | Use Case | Query Language |
|---|---|---|
| Neo4j 5.x | Knowledge graphs, entity relationships | Cypher |
| Dgraph | Context graphs, high-performance graph operations | GraphQL, DQL |
Use Cases
| Use Case | Graph Store | Description |
|---|---|---|
| Data lineage | Neo4j / Dgraph | Track data flow from source to dashboard |
| Context graph | Dgraph | AI context for query understanding |
| Business term relationships | Neo4j | Ontology relationships between terms |
| Impact analysis | Neo4j | Understand downstream effects of schema changes |
| Entity resolution | Dgraph | Link related entities across data sources |
Context Graph Architecture
The context graph stores relationships between data entities to help the AI Service understand the data landscape:
[Table: orders] --HAS_COLUMN--> [Column: amount]
[Table: orders] --HAS_COLUMN--> [Column: order_date]
[Table: orders] --JOINS_TO--> [Table: customers]
[Column: amount] --USED_IN--> [Metric: revenue]
[Metric: revenue] --DEFINED_BY--> [Model: sales_model]Data Lineage
Graph stores track the complete lineage of data:
| Relationship | From | To | Example |
|---|---|---|---|
SOURCE_OF | Data source | Table | S3 bucket --> raw_orders |
TRANSFORMS_TO | Pipeline step | Table | raw_orders --> clean_orders |
JOINS_WITH | Table | Table | orders JOIN customers |
FEEDS_INTO | Table | Dashboard | orders --> Revenue Dashboard |
TRAINED_ON | Dataset | ML Model | features --> churn_model |
Multi-Tenancy
| Strategy | Implementation |
|---|---|
| Neo4j | Tenant ID property on all nodes, Cypher queries filtered by tenant |
| Dgraph | Tenant-scoped predicates, query-time filtering |
// Neo4j: All queries include tenant filter
MATCH (t:Table)-[:HAS_COLUMN]->(c:Column)
WHERE t.tenant_id = 'acme-corp'
RETURN t.name, c.name, c.data_typeDeployment
| Technology | Deployment | Namespace |
|---|---|---|
| Neo4j | Helm chart (StatefulSet) | matih-data-plane |
| Dgraph | Helm chart (StatefulSet) | matih-data-plane |
Related Pages
- Vector Stores -- Embedding-based search
- Agent Flow -- Context graph in AI pipeline
- Data Infrastructure -- Technology overview