Skip to content

Locy Use Case: Fraud Risk Propagation (Rust)

Propagate account risk backward over transfer edges and isolate clean accounts.

This notebook uses schema-first mode and mirrors the Python flow using the Rust API (uni_db).

How To Read This Notebook

  • Define schema first, then load data.
  • Keep Locy rules declarative and focused.
  • Read output rows together with materialization stats.

1) Setup

Initialize an in-memory database and import DataType for schema definitions.

use uni_db::{DataType, Uni, Result};

let db = Uni::in_memory().build().await?;

Define labels, typed properties, and edge types before inserting graph facts.

db.schema()
    .label("Account")
        .property("id", DataType::String)
        .property("flagged", DataType::Boolean)
    .edge_type("TRANSFER", &["Account"], &["Account"])
    .apply()
    .await?;

println!("Schema created");

3) Seed Graph Data

Insert the minimal graph needed for the scenario.

let session = db.session();
let tx = session.tx().await?;
tx.execute("CREATE (:Account {id: 'A1', flagged: true})").await?;
tx.execute("CREATE (:Account {id: 'A2', flagged: false})").await?;
tx.execute("CREATE (:Account {id: 'A3', flagged: false})").await?;
tx.execute("CREATE (:Account {id: 'A4', flagged: false})").await?;
tx.execute("MATCH (a1:Account {id:'A1'}), (a2:Account {id:'A2'}) CREATE (a1)-[:TRANSFER]->(a2)").await?;
tx.execute("MATCH (a2:Account {id:'A2'}), (a3:Account {id:'A3'}) CREATE (a2)-[:TRANSFER]->(a3)").await?;
tx.execute("MATCH (a4:Account {id:'A4'}), (a3:Account {id:'A3'}) CREATE (a4)-[:TRANSFER]->(a3)").await?;
tx.commit().await?;
println!("Seeded graph data");

4) Locy Program

Rules derive relations, then QUERY ... WHERE ... RETURN ... projects the final answer.

let program = r#"CREATE RULE risky_seed AS\nMATCH (a:Account)\nWHERE a.flagged = true\nYIELD KEY a\n\nCREATE RULE risky AS\nMATCH (a:Account)\nWHERE a IS risky_seed\nYIELD KEY a\n\nCREATE RULE risky AS\nMATCH (a:Account)-[:TRANSFER]->(b:Account)\nWHERE b IS risky\nYIELD KEY a\n\nCREATE RULE clean AS\nMATCH (a:Account)\nWHERE a IS NOT risky\nYIELD KEY a\n\nQUERY risky WHERE a.id = a.id RETURN a.id AS risky_account\nQUERY clean WHERE a.id = a.id RETURN a.id AS clean_account"#;

5) Evaluate

Evaluate the Locy program and inspect stats/rows.

let session = db.session();
let result = session.locy(program).await?;
println!("Derived relations: {:?}", result.derived.keys().collect::<Vec<_>>());
println!("Iterations: {}", result.stats().total_iterations);
println!("Queries executed: {}", result.stats().queries_executed);
for (name, rows) in &result.derived {
    println!("{}: {} row(s)", name, rows.len());
}

if let Some(rows) = result.rows() {
    println!("Rows: {:?}", rows);
}

6) What To Expect

Use these checks to validate output after evaluation: - A1 is risky by seed; A2 and A4 become risky by backward propagation through TRANSFER. - A3 should remain in clean because it does not transfer to a risky account. - Two query result blocks should appear: one for risky, one for clean.

Notes

  • Rust notebooks are included for API parity and learning.
  • In this docs build, Rust notebooks are rendered without execution.