Shared Partitions

Partitions are named SPARQL graphs that can be shared across multiple workspaces. They provide a way to maintain common knowledge bases (ontologies, domain facts, company knowledge) independently of any single workspace.

Concepts

A partition is a named graph in the SPARQL store:

#![allow(unused)]
fn main() {
Partition {
    name: String,            // e.g., "ontology", "common-sense"
    graph_name: String,      // SPARQL IRI for the named graph
    source: PartitionSource, // Local or Shared
}
}

Partition Sources

Shared partitions live in a dedicated directory and are mounted by workspaces via configuration.

Creating Shared Partitions

Programmatic

#![allow(unused)]
fn main() {
use akh_medu::partition::PartitionManager;

let pm = PartitionManager::new(partitions_dir);

// Create a new shared partition
let partition = pm.create_shared("company-ontology")?;

// Insert triples into the partition
partition::insert_into_partition(
    &engine,
    triple,
    "company-ontology",
)?;
}

Via Workspace Config

Mount shared partitions in the workspace configuration:

# ~/.config/akh-medu/workspaces/default.toml
shared_partitions = ["company-ontology", "shared-common-sense"]

When the workspace is loaded, mounted partitions become queryable.

Querying Partitions

Query a specific partition's triples:

#![allow(unused)]
fn main() {
let results = partition::query_partition(
    &engine,
    "company-ontology",
    "?s ?p ?o",
)?;
}

This is equivalent to a SPARQL query with a FROM <partition_graph> clause, restricting results to triples within that partition.

Partition Manager

The PartitionManager handles discovery and lifecycle:

#![allow(unused)]
fn main() {
let pm = PartitionManager::new(partitions_dir);

// Discover existing partitions on disk
let count = pm.discover()?;

// Register a partition
pm.register(partition)?;

// List all partitions
let names = pm.list();

// Get a specific partition
let p = pm.get("ontology")?;

// Remove a partition
pm.remove("old-partition")?;
}

Use Cases

• Shared ontologies: A common set of relations and categories used by all workspaces in an organization.
• Domain knowledge: Medical, legal, or engineering terminology shared across project workspaces.
• Cross-workspace inference: Triples in a shared partition are visible to inference and traversal in any workspace that mounts it.

Relationship to Compartments

Partitions and compartments both use SPARQL named graphs, but serve different purposes: