Policies

A policy is a document for the definitions and rules your whole project should agree on: the valid values for a field, the shape of your core entities, and the derived facts other rules build on. Write them once in a policy, then import the policy wherever it’s needed. Unlike a graph, a policy has no nodes or edges. It reads like a document - text and rule blocks mixed together - and the engine works out the evaluation order automatically from what each block reads and writes.

​

Writing a policy

Type / anywhere in the document to insert a block. Text blocks (headings, paragraphs, lists) document your rules; rule blocks define them: Block order doesn’t matter. If one block computes applicant.debtToIncome and another reads it, the engine runs them in the right order - even across imported policies.

​

Dictionaries

A dictionary is a named set of values with human-readable labels: The value is what rules compare against and what evaluation sees; the label is what business users read. In business mode, rules render with labels - “applicant.employment is one of Employed, Self-employed”. Use a dictionary’s name as a type: a data model property typed employmentStatus only accepts those four values, and static analysis flags anything else at edit time. Graphs that import the policy can do the same in their input schema - a field defined as {"$dictionary": "employmentStatus"} accepts only those values at evaluation time. Dictionaries also support CSV import and export for longer value lists.

​

Data models

A data model declares an entity - applicant, customer, order - and its properties with types: string, number, boolean, date, or any dictionary. Properties can be arrays or optional. Once declared, every rule that touches the entity is type-checked against it.

​

Importing

Graphs import policies with the Import button on the canvas; policies import other policies through the imports in their header. An import brings the whole chain: importing a policy also pulls in everything it imports. Imported policies evaluate together as one unit in a shared namespace. That has a practical consequence: two blocks writing the same property is a conflict (DUPLICATE_WRITER) even when they live in different files - the analyzer checks the whole import group together.

​

Evaluating policies

A policy isn’t just a library - it’s directly executable. Send it an input object and it returns the object enriched with everything its blocks computed. Every policy gets its own HTTP endpoint alongside your graphs; see Developer Tools.

​

The side panel

The Entities tab lists every entity the policy defines or touches: each property’s type, whether it’s declared or computed, which blocks write it, and where it’s used. The Problems tab shows static analysis findings for the policy.

​

Related pages