actra 0.7.4

Actra - Deterministic admission control engine for automated system actions.

Actra Python SDK

Action Admission Control for Automated Systems

Deterministic policy engine that decides whether automated actions are allowed before they execute.

The Actra Python SDK provides a simple interface for loading policies and evaluating decisions using the Actra engine written in Rust.

---

Runtime Admission Control

The SDK also provides a runtime layer for protecting functions using Actra policies.

If a policy blocks the operation, the function will not execute and a PermissionError will be raised.

---

Installation

Install from PyPI:

pip install actra

The package includes a compiled Rust engine, so no Rust toolchain is required during installation.

---

Quick Start

@actra.admit()
def refund(amount):
    ...

The rule lives in policy:

rules:
  - id: block_large_refund
    when:
      subject:
        domain: action
        field: amount
      operator: greater_than
      value:
        literal: 1000
    effect: block

Result:

refund(200)   > allowed  
refund(1500)  > blocked by policy

Actra evaluates the policy before the function executes and blocks refunds greater than 1000.

---

Design Goals

The Python SDK focuses on:

• simple developer ergonomics
• deterministic policy evaluation
• minimal runtime overhead
• seamless integration with the Rust engine

The heavy lifting is performed by the core engine, ensuring fast and consistent evaluations.

---

Architecture

flowchart LR
    subgraph User Code
        U[User Code]
    end
    
    subgraph SDK Layer
        S[Python SDK]
    end

    subgraph Native Bindings
        N[Rust bindings<br/>actra]
    end

    subgraph Core Engine
        C[Actra Engine]
    end

    U --> S --> N --> C

The SDK provides a Python-friendly interface while the core engine handles compilation and evaluation.

---

License

Apache License 2.0

---

Project

Actra is designed for systems requiring explicit, reproducible control over state-changing operations in automated environments.