Per FULLSTACK_MASTER §E: "Amica = open VRM avatar shell that talks to ANY LLM backend. SOV33's role: BE THE BACKEND Amica calls. Character memory/personality/care/identity in SOV33; VRM body is the shell."
Amica (and any OpenAI-compatible VRM avatar shell) can call SOV33 as if it were any other LLM backend. The avatar speaks; SOV33 thinks and answers. Memory, identity, care-floor all stay in SOV33.
┌─────────────────────────────────┐
│ Amica VRM Shell (the body) │
│ • Face, lipsync, gestures │
│ • ANY LLM backend config │
└────────────┬────────────────────┘
│ OpenAI-compatible HTTP
▼
┌─────────────────────────────────┐
│ SOV33 Amica Adapter │
│ POST /v1/chat/completions │
│ GET /v1/models │
└────────────┬────────────────────┘
│
▼
┌─────────────────────────────────┐
│ SOV33 Sovereign Substrate │
│ • Care-floor 0.95 │
│ • 5 OWEMs (compliance, defense, │
│ intuition, voice, general) │
│ • 61-model router │
│ • SIGIL-signed every action │
│ • Article 0 binding │
└─────────────────────────────────┘| Model ID | OWEM | Specialty |
|---|---|---|
compliance | EU AI Act, UK AI Bill, Article 50 | Regulatory knowledge |
defense | Kill switch, intrusion, foreign-access | Defensive AI |
intuition | Patterns, predictions | Sense-making |
voice | Sovereign truths, Charter | Speech-style |
general | General knowledge | Fallback |
Try the SOV33 backend with the same payload an Amica avatar would send:
// Amica config (or any OpenAI-compatible VRM shell)
const response = await fetch('https://os.meok.ai/v1/chat/completions', {
method: 'POST',
headers: {'Content-Type': 'application/json'},
body: JSON.stringify({
model: 'voice', // picks the OWEM
messages: [
{role: 'user', content: 'What is Article 0?'}
]
})
});
const data = await response.json();
const text = data.choices[0].message.content;
// → "Article 0 of the Charter states: ..."