The goal here is to define a schema standard that would be returned for a given graph by running a query like SHOW SCHEMA INFO;. The first few versions of the standard aim to focus on getting the statistical data about the schemaless label-property graphs. Over the long run, the goal is to incorporate the full definition of any schema the user defines.
NOTE: .json examples are formatted using https://jsonformatter.curiousconcept.com/ (2 Space Tab).
The standard should be the basis for building many different applications leveraging the specification. Examples of such applications are:
- What's the right data model in the first place (how should people model, and what's the right approach), helping in communication
- data validity checks
- semantic rules and derived knowledge (AKA inference)
- graph schema visualizers
- graph query language code completion
- tools to assist with schema changes by estimating the implications
- tools to generate mock graphs for testing and staging (AKA fakers)
- Graph coloring based on the schema
- LLM and, in general, AI applications:
- providing LLMs the right context to generate relevant knowledge retrieval queries
- implementing a generator of a diverse set of queries to, e.g., fine-tune an LLM.
- memgraph outputs v1 under
SHOW SCHEMA INFO;query - memgraph lab uses v1 under graph schema and auto completion features
- Data modeling under Lab helps you build the model/schema/ontology
- ai-toolkit uses v1 under the automatic SQL to graph agent.