The AI Council

#Overview

The AI Council is theLAB's multi-perspective AI response system. Instead of relying on a single AI model, it queries 5 specialized models running on Ollama across your local network, then synthesizes their responses into a unified consensus using Gemma3:27b.

Each council member has a distinct role and personality, providing different perspectives on any question. This approach helps overcome single-model biases and produces more well-rounded, thoroughly analyzed answers.

• 5 AI Models — diverse perspectives from specialized models.
• Consensus Engine — Gemma3:27b synthesizes all responses.
• Distributed — models across home lab servers.

#Council Members

The AI Council consists of 5 specialized models, each with a unique role. Models are distributed across localhost and 192.168.1.100:

#Architecture

The system uses a PHP proxy (ollama_proxy.php) to route requests to Ollama instances across the network:

#Response Modes

The Council supports two response modes selectable via dropdown:

#Sequential Mode

Models respond one at a time in order (1→2→3→4→5). Each response streams in real-time with a typing cursor. Good for following the "conversation" between council members and seeing how each builds on the previous.

#Parallel Mode

All models are queried simultaneously using Promise.all(). Faster overall response time since all models work at once, but responses appear together when all finish.

// Council member configuration
const COUNCIL_MEMBERS = [
  {
    id: 'adi',
    name: 'Gemma3',
    fullName: 'Gemma3:12b',
    endpoint: 'localhost',
    model: 'gemma3:12b',
    role: 'Primary Assistant',
    description: 'Comprehensive analysis and detailed responses',
    avatar: 'adi'
  },
  {
    id: 'tinyllama',
    name: 'Qwen3',
    fullName: 'Qwen3:4b',
    endpoint: '192.168.1.100',
    model: 'qwen3:4b',
    role: 'Quick Insights',
    description: 'Fast, concise responses and key insights',
    avatar: 'tinyllama'
  },
  // ... wizardlm, historian, strategos
];

#Consensus Engine

After all 5 council members respond, the Consensus Engine synthesizes their perspectives into a unified answer. This calls consensus.php which uses Gemma3:27b — a larger model capable of understanding and reconciling multiple viewpoints.

The consensus is displayed in a special green-bordered panel with streaming output. Once complete, a speaker button appears to play the consensus via TTS.

#Role Prompts

Each council member receives a role-specific system prompt that shapes their personality:

const rolePrompts = {
  adi: `You are Gemma3, the primary AI assistant. Provide
        comprehensive, well-structured responses with clear
        explanations and examples when helpful.`,

  tinyllama: `You are Qwen3, focused on providing quick, concise
              insights. Give clear, direct answers that capture
              the key points efficiently.`,

  wizardlm: `You are WizardLM, the critical analyst. Question
             assumptions, present alternative viewpoints, and
             provide constructive criticism.`,

  historian: `You are Phi3, specializing in historical context.
              Relate topics to historical precedents and trends.`,

  strategos: `You are GPT-OSS STRATEGOS, the Council's staff
              planner. Turn problems into clear, executable plans
              with assumptions, numbered steps, verification checks,
              rollback paths, and risks. Prefer commands, exact
              paths, and short rationale.`
};

#Ollama Proxy

The ollama_proxy.php handles all communication with Ollama instances, supporting streaming responses via Server-Sent Events:

// ollama_proxy.php
<?php
header('Content-Type: text/event-stream');
header('Cache-Control: no-cache');

$input = json_decode(file_get_contents('php://input'), true);

$endpoint = $input['endpoint'] ?? 'localhost';
$model = $input['model'] ?? 'gemma3:12b';
$prompt = $input['prompt'] ?? '';

$ollamaUrl = "http://{$endpoint}:11434/api/generate";

$payload = json_encode([
    'model' => $model,
    'prompt' => $prompt,
    'stream' => true
]);

$ch = curl_init($ollamaUrl);
curl_setopt($ch, CURLOPT_POST, true);
curl_setopt($ch, CURLOPT_POSTFIELDS, $payload);
curl_setopt($ch, CURLOPT_HTTPHEADER, ['Content-Type: application/json']);
curl_setopt($ch, CURLOPT_WRITEFUNCTION, function($ch, $data) {
    echo $data;
    ob_flush();
    flush();
    return strlen($data);
});
curl_exec($ch);
echo '[STREAM_END]';
curl_close($ch);

#Setup Guide

#1. Install Ollama on Your Servers

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull required models
ollama pull gemma3:12b
ollama pull gemma3:27b    # For consensus
ollama pull qwen3:4b
ollama pull wizardlm2:7b
ollama pull phi3:3.8b

# Enable network access
sudo systemctl edit ollama
# Add: Environment="OLLAMA_HOST=0.0.0.0"
sudo systemctl restart ollama

#2. Deploy PHP Files

/var/www/ai-council/
├── index.php           # Main UI (Tailwind + vanilla JS)
├── ollama_proxy.php    # Ollama API proxy with streaming
└── consensus.php       # Consensus engine (Gemma3:27b)

#3. Update Endpoints

Edit the COUNCIL_MEMBERS array to match your network. Change endpoint values to point to your Ollama servers.