README.md

metadata

title: EPA Demo
emoji: 🧬
colorFrom: blue
colorTo: gray
sdk: gradio
sdk_version: 4.20.0
app_file: app.py
pinned: false

EPA (Ethical Prompt Architecture) Demo

This Space demonstrates the EPA scaffold in action...

EPA: Ethical Prompt Architecture (Python Scaffold)

Overview

EPA is a composable, auditable meta-layer for integrating language models, designed to ensure ethical, context-rich, and traceable operations.
It embodies the Universal Flourishing Objective (ϕ₁) and supports modular extension for future ethical, technical, or operational requirements.

Core Modules

• Contextualization Module: Injects roles, time, user input, constraints, memory state, and custom context into all prompts.
• Ethical Gating Module: Prunes or allows actions based on ethical score (F) relative to threshold (θ₀).
• Audit & Trace Module: Generates GoldenDAG, Trace ID, and Codex ID for every EPA invocation.
• Qualia & PII Redaction Module: (Stub) Redacts or flags qualia-sensitive and PII content (extend as needed).
• Memory & State Management: Allows dynamic updates to memory/context.
• Extensibility Hooks: Supports adding constraints and custom context.

Example Usage

from epa.core import EPA

epa = EPA(system_role="Code Ethics Advisor")
prompt, ids = epa.build_epa("Refactor the code for efficiency.", {"project": "EPA Scaffold"})
print(prompt)
print(ids)

# Ethical gating example
F = 1.0  # Calculated ethical score
if not epa.ethical_gate(F, θ₀=0.5):
    raise PermissionError("Action pruned: does not meet Universal Flourishing Objective (ϕ₁).")

Extending EPA

• Implement custom logic in redact_qualia_pii for advanced redaction.
• Override/add modules for domain-specific constraints, memory, or ethics.

License

MIT License (or as specified by project owner)