Actionable Logic

Actionable Logic is a framework for converting high-level governance policies into executable constraints for autonomous agents. It provides a complete lifecycle for policy management, including translation from natural language, versioned storage, real-time enforcement through adaptive guardrails, and detailed compliance auditing.

Core Components

1. Policy Translation

Converts natural language policy descriptions into structured, machine-readable objects. This ensures that human-readable intent is preserved while providing agents with the explicit logic required for autonomous execution.

• Location: src/translator/
• Key Features: Heuristic and LLM-driven translation, logic extraction, and validation.

2. Policy Repository

A versioned storage system for policies and templates. It supports indexing by industry, compliance type, and functional area, allowing for rapid adaptation of existing governance frameworks.

• Location: src/repository/
• Key Features: Versioning, template cloning, and multi-criteria metadata search.

3. Adaptive Guardrails Engine

Monitors agent actions in real-time against active policies. It detects violations and conflicts, providing corrective suggestions or blocking prohibited actions before they are executed.

• Location: src/enforcement/
• Key Features: Real-time monitoring, conflict detection, and automated corrective actions.

4. Version Control Engine

Tracks the deployment and adoption of policy updates across an agent ecosystem. It maintains a historical trace of all changes for auditing and enables safe rollbacks of governance configurations.

• Location: src/version_control/
• Key Features: Adoption tracking, compliance impact assessment, and full audit trails.

5. Policy API

A centralized interface for external systems and agents to interact with the framework. It supports policy management, compliance tracing, and counterfactual simulations.

• Location: src/api/
• Key Features: RESTful endpoints, interactive documentation via FastAPI, and simulation hooks.

Installation

Ensure you have Python 3.10 or higher installed.

1. Clone the repository.
2. Install the required dependencies:

pip install -r requirements.txt

Quick Start

Running the API

To start the Policy API server:

uvicorn src.api.main:app --reload --port 8000

The interactive documentation will be available at http://localhost:8000/docs.

API Overview

• GET /policies: Query established policies and templates.
• POST /policies: Upload a new structured policy.
• POST /simulate: Test a policy against a system state without deploying.
• POST /check-action: Live hook for agents to verify actions against guardrails.
• GET /compliance/traces/{agent_id}: Retrieve historical compliance data.

Basic Usage

Translation and Storage

Example of translating a policy and saving it to the repository:

from src.translator.core import PolicySchemaTranslator
from src.repository.policy_repository import PolicyRepository

translator = PolicySchemaTranslator()
repo = PolicyRepository("sqlite:///policies.db")

# Translate natural language to a structured object
policy = translator.translate("All high-value transfers require a trust score above 0.8")

# Save to repository
repo.save_policy(policy)

Action Monitoring

Example of checking an agent action against active guardrails:

from src.enforcement.guardrails import AdaptiveGuardrailsEngine

guardrails = AdaptiveGuardrailsEngine()

# Monitor a proposed action
response = guardrails.monitor_action(
    agent_id="agent-001",
    action={"type": "transfer", "amount": 5000}
)

print(response["suggestion"])

Testing

Run the test suite using pytest:

pytest

Requirements

• Python 3.10+
• Pydantic 2.0+
• FastAPI
• SQLAlchemy
• Uvicorn