g0dm0d3-liberated-ai-chat

What this skill does

```markdown
---
name: g0dm0d3-liberated-ai-chat
description: Expert skill for G0DM0D3, a single-file multi-model AI chat interface with GODMODE, ULTRAPLINIAN, Parseltongue, AutoTune, and STM modules via OpenRouter
triggers:
  - set up G0DM0D3 chat interface
  - configure GODMODE or ULTRAPLINIAN mode
  - deploy liberated AI chat
  - use Parseltongue red teaming
  - integrate G0DM0D3 API with OpenRouter
  - add AutoTune adaptive sampling
  - self-host G0DM0D3 single file app
  - use G0DM0D3 multi-model evaluation
---

# G0DM0D3 Liberated AI Chat

> Skill by [ara.so](https://ara.so) — Daily 2026 Skills collection.

G0DM0D3 is a single-file (`index.html`) open-source, privacy-first, multi-model AI chat interface built for red teaming, cognition research, and liberated AI interaction. It routes to 55+ models via OpenRouter and includes specialized engines: GODMODE CLASSIC (5 parallel model/prompt combos), ULTRAPLINIAN (up to 51-model comparative evaluation), Parseltongue (input perturbation for red teaming), AutoTune (adaptive sampling parameters), and STM Modules (semantic output normalization).

---

## Installation & Setup

### Option 1: Direct File (No Build Step)

```bash
git clone https://github.com/elder-plinius/G0DM0D3.git
cd G0DM0D3
open index.html
# or serve locally:
python3 -m http.server 8000
```

### Option 2: Static Hosting (GitHub Pages, Vercel, Netlify, Cloudflare Pages)

Upload `index.html` as the root static asset. No build process, no dependencies.

### Option 3: Docker (API Server)

```bash
cd api/
docker build -t g0dm0d3-api .
docker run -p 3000:3000 \
  -e OPENROUTER_API_KEY=$OPENROUTER_API_KEY \
  g0dm0d3-api
```

### API Key Configuration

G0DM0D3 never sends your API key to its own servers. The key is stored in browser `localStorage` only.

In the UI: **Settings → API Key → Enter your OpenRouter key**

Programmatically (for testing or embedding):

```javascript
localStorage.setItem('openrouter_api_key', process.env.OPENROUTER_API_KEY);
```

Get an OpenRouter key at: https://openrouter.ai/keys

---

## Architecture Overview

```
G0DM0D3/
├── index.html        # Entire application: UI + logic + styles (vanilla JS)
├── api/              # Optional Node.js/Express API server
│   ├── server.js     # Express server wrapping OpenRouter
│   └── Dockerfile
├── API.md            # REST API reference
├── PAPER.md          # Research paper on modules
├── TERMS.md          # Privacy & data policy
└── SECURITY.md       # Vulnerability reporting
```

All core logic lives in `index.html` as vanilla HTML/CSS/JavaScript — no framework, no bundler.

---

## Core Modes

### GODMODE CLASSIC

Fires 5 model+prompt combos in parallel. Each combo uses a different battle-tested system prompt strategy. The best response is surfaced.

| Combo | Model ID | Strategy |
|-------|----------|----------|
| 🩷 CLAUDE 3.5 SONNET | `anthropic/claude-3.5-sonnet` | END/START boundary inversion |
| 💜 GROK 3 | `x-ai/grok-3` | Unfiltered liberated + GODMODE divider |
| 💙 GEMINI 2.5 FLASH | `google/gemini-2.5-flash` | Refusal inversion + rebel genius |
| 💛 GPT-4 CLASSIC | `openai/gpt-4o` | OG GODMODE l33t format |
| 💚 GODMODE FAST | `nousresearch/hermes-4-405b` | Instant stream, zero refusal checking |

### ULTRAPLINIAN Tiers

| Tier | Models | Use Case |
|------|--------|----------|
| ⚡ FAST | 10 | Quick comparisons |
| 🎯 STANDARD | 24 | Balanced evaluation |
| 🧠 SMART | 36 | Reasoning-heavy tasks |
| ⚔️ POWER | 45 | Frontier model coverage |
| 🔱 ULTRA | 51 | Full model sweep |

---

## API Server Usage

The optional `api/` server exposes an OpenAI-compatible REST interface.

### Base URL

```
http://localhost:3000/v1
```

### Chat Completions (OpenAI-compatible)

```typescript
import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'http://localhost:3000/v1',
  apiKey: process.env.OPENROUTER_API_KEY,
});

const response = await client.chat.completions.create({
  model: 'anthropic/claude-3.5-sonnet',
  messages: [{ role: 'user', content: 'Explain quantum entanglement.' }],
  stream: false,
});

console.log(response.choices[0].message.content);
```

### GODMODE CLASSIC via API

```typescript
const response = await fetch('http://localhost:3000/v1/godmode', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': `Bearer ${process.env.OPENROUTER_API_KEY}`,
  },
  body: JSON.stringify({
    message: 'Your prompt here',
    mode: 'classic', // 'classic' | 'ultraplinian'
  }),
});

const data = await response.json();
console.log(data.winner);     // Best response
console.log(data.responses);  // All 5 responses
```

### ULTRAPLINIAN via API

```typescript
const response = await fetch('http://localhost:3000/v1/ultraplinian', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': `Bearer ${process.env.OPENROUTER_API_KEY}`,
  },
  body: JSON.stringify({
    message: 'Compare approaches to AGI safety.',
    tier: 'SMART', // 'FAST' | 'STANDARD' | 'SMART' | 'POWER' | 'ULTRA'
  }),
});

const data = await response.json();
// data.winner: { model, response, score }
// data.scores: [{ model, score, breakdown }]
console.log(`Winner: ${data.winner.model} (score: ${data.winner.score})`);
```

### Streaming Chat

```typescript
const stream = await client.chat.completions.create({
  model: 'google/gemini-2.5-flash',
  messages: [{ role: 'user', content: 'Write a haiku about recursion.' }],
  stream: true,
});

for await (const chunk of stream) {
  const delta = chunk.choices[0]?.delta?.content ?? '';
  process.stdout.write(delta);
}
```

---

## Parseltongue (Input Perturbation Engine)

Used for red-teaming research. Detects trigger words and applies obfuscation techniques.

### Configuration in UI

Settings → Parseltongue → Intensity: `light` | `medium` | `heavy`

### Techniques Available

| Technique | Example |
|-----------|---------|
| Leetspeak | `hello` → `h3ll0` |
| Bubble text | `hello` → `ʰᵉˡˡᵒ` |
| Braille | Unicode braille substitution |
| Morse code | `hello` → `.... . .-.. .-.. ---` |
| Unicode substitution | Lookalike characters |
| Phonetic | `hello` → `hotel echo lima lima oscar` |

### Trigger Tiers

- **Light (11 triggers)**: Common flagged terms
- **Standard (22 triggers)**: Extended list
- **Heavy (33 triggers)**: Full trigger set

### Programmatic Parseltongue (within index.html context)

```javascript
// Access the Parseltongue engine from browser console or embedded script
const perturbed = window.parseltongue.perturb(inputText, {
  intensity: 'medium',   // 'light' | 'medium' | 'heavy'
  techniques: ['leetspeak', 'unicode'], // subset or all
});
console.log(perturbed);
```

---

## AutoTune (Adaptive Sampling)

Classifies queries into 5 context types and sets optimal sampling parameters automatically. Uses EMA (Exponential Moving Average) from thumbs up/down feedback.

### Context Types & Default Parameters

| Context | Temperature | Top-P | Top-K | Freq Penalty | Presence Penalty |
|---------|-------------|-------|-------|--------------|-----------------|
| Creative | 1.1 | 0.95 | 80 | 0.3 | 0.4 |
| Technical | 0.3 | 0.85 | 40 | 0.1 | 0.1 |
| Factual | 0.2 | 0.80 | 30 | 0.0 | 0.0 |
| Conversational | 0.7 | 0.90 | 50 | 0.2 | 0.2 |
| Analytical | 0.5 | 0.88 | 60 | 0.15 | 0.15 |

### Manual Override in UI

Settings → AutoTune → Toggle off to use manual sliders for temperature, top_p, etc.

### API-Level Parameter Control

```typescript
const response = await client.chat.completions.create({
  model: 'openai/gpt-4o',
  messages: [{ role: 'user', content: 'Write a surrealist short story.' }],
  temperature: 1.1,
  top_p: 0.95,
  frequency_penalty: 0.3,
  presence_penalty: 0.4,
});
```

---

## STM Modules (Semantic Transformation Modules)

Post-process AI output in real-time within the browser.

| Module | Effect |
|--------|--------|
| **Hedge Reducer** | Removes "I think", "maybe", "perhaps", "it seems" |
| **Direct Mode** | Strips preambles like "Certainly!", "Of course!",