Eve library

@oshun/persona-policy-lilith

Other

Revision 1

On this page

Canonical Lilith contemplative persona policy adapter contracts for OSHUN.

This package defines the product-facing Lilith interface that OSHUN uses when it needs contemplative teacher policy packs, tone guidance, spiritual-domain safety overlays, abstain/refer behavior, topic-scope checks, and voice-safety requirements for persona-linked guidance.

It deliberately normalizes the fragmentation that already exists across Lilith:

  • moderation policy packs own persona-specific boundaries and abstain/refer rules
  • spiritual-guidance services own contemplative personas, tone, and topic scope
  • safety services own crisis detection, resource routing, and disclaimer logic
  • voice services own watermarking, clone abuse scoring, and quality gates

The adapter contract exposes one canonical policy pack, one safety/tone evaluation surface, and one prompt-overlay model for OSHUN shell and admin integration work.

Scope#

  • canonical persona-pack, tone-guidance, safety-assessment, topic-scope, and prompt-overlay DTOs
  • low-level adapter interface for Lilith persona policy, spiritual-guidance, safety, registry, archetype, and voice-safety primitives
  • canonical adapter factory (createCanonicalLilithPersonaPolicyAdapter) that wraps an injected low-level Lilith adapter
  • normalization helpers that align fragmented Lilith persona and safety inputs with OSHUN policy contracts

Usage#

ts
import { createCanonicalLilithPersonaPolicyAdapter } from '@oshun/persona-policy-lilith';

const lilith = createCanonicalLilithPersonaPolicyAdapter({ apiAdapter });

const evaluation = await lilith.evaluateInteraction({
  consumer: 'assistant',
  activeDomain: 'tara',
  userMessage: 'Can you guide me through a grief meditation?',
  topic: 'grief meditation',
  userExperienceLevel: 'beginner',
});