ShrikeDB documentation
Store property graph relationships, RDF terms, vector embeddings, key-value payloads, and spatiotemporal observations in one database. Query them with Cypher, SPARQL, gRPC, and HTTP APIs.
This guide covers installation, data modeling, ingestion, every public query surface, API clients, query planning, administration, backup, and HA configuration. Examples use an explicit local configuration and identify the values to replace for a remote deployment.
Start with a task
Run your first query
Start a local server, check that it is ready, and send a query with a known result.
Load your data
Write application records, import a file, or register a Parquet source. Choose identifiers that let graph, vector, and payload results join correctly.
Query connected data
Match relationships with Cypher, query RDF with SPARQL, and run graph analytics.
Search embeddings
Create named vector indexes, use streaming or packed search, and join nearest neighbors to graph and KV data.
Compose modalities
Move between graph, vector, key-value, RDF, SpaceKey, and lakehouse data through shared identifiers.
Operate a server
Configure storage and access, inspect health and query traces, and back up a database before changing its deployment.
Understand execution
Follow parsing, semantic checks, physical selection, snapshot binding, profiling, and optimizer behavior.
How the data fits together
A product can have a graph node for its relationships, a vector for similarity search, and a key-value record for its full description. A shared entity identifier lets an application find similar products, filter them by brand, traverse related entities, and return their payloads.
SpaceKey stores timestamped geographic observations in a separate index for position, time-window, and entity-track searches. Lakehouse views expose reviewed mappings over external Parquet files. Semantic profiles describe the domain vocabulary used by natural-language queries.
Start with your first multimodal dataset for a complete graph, vector, and key-value example. For an existing deployment, use the connection guide and workflow overview.
Core documentation
| Area | Coverage |
|---|---|
Cypher |
Patterns, pipelines, optional matches, aggregation, collections, subqueries, paths, writes, schema, compatibility, plans, and procedures |
SPARQL and RDF |
Graph patterns, optional groups, aggregates, property paths, result graphs, named graphs, and local services |
Vector |
Named indexes, single and bulk insertion, streaming and packed search, Cypher KNN, tuning, durability, and recovery |
Multimodal |
Shared identity, graph-vector-KV composition, query-derived relationships, SpaceKey handoff, and snapshot consistency |
Optimization |
Execution lifecycle, specialized candidates, cardinality, |
Operations |
Configuration, startup, health, observability, backup, restore, security, users, and HA |
APIs |
gRPC services and messages, HTTP routes, environment settings, MCP tools, and downloadable protocol definitions |
About this version
This guide describes ShrikeDB 0.1. Feature availability depends on the server package and configuration. Each feature chapter explains its prerequisites and current limits. See feature availability when planning a deployment or migrating an application.