TwinMind Reference Architecture
Overview
Production architecture for meeting AI systems using TwinMind: transcription pipeline, memory vault, action item workflow, and calendar integration. TwinMind uses the Ear-3 speech model (5.26% WER, 3.8% DER) for transcription, with GPT-4, Claude, and Gemini for AI summarization.
Prerequisites
- TwinMind account (Free, Pro $10/mo, or Enterprise)
- Chrome extension installed and authenticated
- Understanding of TwinMind workflow
Instructions
Step 1: Setup
TwinMind operates as a Chrome extension and mobile app with optional API access for Pro/Enterprise users.
// TwinMind configuration
const config = {
apiKey: process.env.TWINMIND_API_KEY,
model: "ear-3", // Transcription model
aiModels: ["gpt-4", "claude", "gemini"], // Summary models
};
Step 2: Implementation
// TwinMind Reference Architecture implementation
// Core TwinMind integration
const twinmind = {
transcriptionModel: "ear-3",
languages: ["en", "es", "ko", "ja", "fr"],
features: ["transcription", "summary", "action-items"],
privacyMode: "on-device", // Audio never stored
};
// Check transcription capabilities
async function verify() {
const health = await fetch("https://api.twinmind.com/v1/health");
console.log("TwinMind status:", await health.json());
}
