🤖 MISSI — Mistral Intelligent Super System Interface

A voice-first AI operating system powered by 6 specialized Mistral models, 28 autonomous tools, and 10,000+ integrations — built for the Mistral AI Worldwide Hackathon 2026.

🌐 Live Demo · 🎬 Demo Video · 🏢 MiMi Tech AI

---

What is MISSI?

MISSI is a voice-first AI operating system that understands you in 10 languages, intelligently routes every query to the optimal Mistral model, executes multi-step workflows autonomously, and responds with natural human-like speech.

Speak → MISSI routes to the best model → Executes tools autonomously → Speaks back naturally.

Why MISSI Stands Out

Feature	MISSI	Typical Chatbots
Model Selection	6 specialized Mistral models, auto-routed per query	Single model
Voice	Voxtral STT + ElevenLabs TTS, continuous conversation	Text only or basic TTS
Tools	28 built-in + 10,000+ via Composio	Limited or none
Languages	10 languages, auto-detected from browser	English only
Integrations	Gmail, Calendar, GitHub, Slack, Notion, Drive	None or manual
Documents	Professional HTML reports with charts & tables	Plain text

---

🧠 6-Model Intelligent Routing

MISSI analyzes every message and selects the optimal Mistral model in real-time:

Query Type	Model	Trigger Examples
General intelligence	mistral-large-latest 🧠	Research, analysis, questions
Fast responses	mistral-small-latest ⚡	Greetings, simple queries
Deep reasoning	magistral-medium-latest 🧮	Math, logic, proofs, comparisons
Creative writing	mistral-medium-latest 🎨	Stories, poems, creative content
Code generation	codestral-latest 💻	Programming, debugging, scripts
Image analysis	pixtral-large-latest 👁️	Photo analysis, OCR, diagrams

Additional models used internally:

• voxtral-small-latest — Speech-to-text (Mistral-native STT)
• mistral-ocr-latest — Document OCR and analysis

---

🛠️ 28 Agent Tools

Core Intelligence

Tool	Description
web_search	Real-time web search via Brave API
read_webpage	Extract content from any URL
wikipedia	Wikipedia knowledge in any language
news_headlines	Latest news via Google News RSS
get_weather	Weather + 3-day forecast (Open-Meteo)
get_time	Current time in any timezone
get_location	Browser geolocation

Finance & Data

Tool	Description
get_stock_price	Real-time stock prices (Yahoo Finance)
get_crypto_price	Live crypto prices (CoinGecko)
calculate	Mathematical computations
analyze_data	Statistical analysis & pattern finding
unit_convert	Temperature, length, weight, volume, speed, data
generate_chart	Bar, line, pie, area charts (SVG)

Creation & Documents

Tool	Description
run_code	Execute JavaScript in sandbox
generate_code	Production code via Codestral
create_document	Professional HTML reports with tables & metrics
analyze_document	OCR via mistral-ocr-latest
translate	Multi-language translation
summarize_text	Intelligent text summarization
define_word	Dictionary with synonyms & examples
random_fact	Curated facts & trivia

Integrations (Composio — Permission-Gated)

Tool	Description
search_gmail / read_gmail	Gmail inbox search & reading
get_calendar	Google Calendar events
get_github	GitHub repos, issues, create issues
search_files	Google Drive file search
set_reminder	Browser notification reminders
change_voice	Switch between 6 ElevenLabs voices

---

🎙️ Voice System

MISSI's voice pipeline is designed for natural, continuous conversation:

1. Voxtral STT — Mistral-native speech-to-text (not browser APIs)
2. Intelligent Processing — Auto-routes to optimal model
3. ElevenLabs TTS — Full response spoken as continuous natural speech
4. Auto-Relisten — Immediately listens for your next input

Voice Features

• Continuous Conversation — Speak naturally back-and-forth
• 10 Languages — Auto-detected from browser, switchable in UI
• 6 Switchable Voices — Sarah, Aria, Rachel, Eric, Roger, Charlie
• VoiceOrb Visualization — Animated state indicator (idle → listening → thinking → speaking)
• Keyboard Shortcuts — Space (voice), Escape (stop), ⌘K (clear)

---

📄 Canvas Panel

MISSI generates professional documents rendered in a side panel:

• HTML Documents — Styled reports with tables, metrics, charts
• Traffic Light Controls — Close (🔴), minimize (🟡), fullscreen (🟢)
• Download — HTML or Markdown format
• Print-Ready — Optimized CSS for printing

Documents include CSS components: stat cards, info cards, comparison grids, progress bars, timelines, and data tables.

---

🔌 10,000+ Integrations (Composio)

One-click OAuth connection via Composio:

Integration	Status	Features
📧 Gmail	✅ Connected	Search, read, send emails
📅 Google Calendar	✅ Connected	View, create, manage events
🐙 GitHub	✅ Connected	Repos, issues, create issues
💬 Slack	Available	Messages, channels
📝 Notion	Available	Pages, databases
📁 Google Drive	Available	Files, sharing
➕ 10,000+ more	Via Composio	Managed OAuth & token refresh

---

🌍 Internationalization

MISSI auto-detects the browser language and switches the entire UI:

• 10 Languages: 🇩🇪 DE · 🇺🇸 EN · 🇫🇷 FR · 🇪🇸 ES · 🇮🇹 IT · 🇵🇹 PT · 🇯🇵 JA · 🇰🇷 KO · 🇨🇳 ZH · 🇷🇺 RU
• Welcome screen, badges, placeholders, footer — all localized
• AI responses match the selected language
• TTS automatically selects the optimal voice model per language

---

🏗️ Architecture

┌─────────────────────────────────────────────────┐
│            MISSI Frontend (Next.js 16)          │
│  VoiceOrb ← Web Audio API ← ElevenLabs TTS     │
│  Voxtral STT → Chat → Vision → Canvas Panel    │
└──────────────────────┬──────────────────────────┘
                       │ SSE Streaming
┌──────────────────────▼──────────────────────────┐
│          Intelligent Model Router                │
│    Query Analysis → Optimal Model Selection      │
│  ⚡small 🧠large 💻codestral 👁️pixtral 🧮magistral 🎨medium │
└──────────────────────┬──────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────┐
│        Autonomous Planning Engine                │
│    Task Decomposition → Multi-Step Execution     │
│    Up to 8 rounds × 28 tools                    │
└──────────────────────┬──────────────────────────┘
                       │
┌──────────────────────▼──────────────────────────┐
│     28 Built-in Tools + 10,000+ via Composio    │
│  Search│Weather│Code│Finance│Docs│Gmail│GitHub│… │
└─────────────────────────────────────────────────┘

---

🚀 Tech Stack

Layer	Technology
Frontend	Next.js 16.1.6, React 19, TypeScript, Tailwind CSS 4
AI Models	Mistral AI SDK 1.14.1 (6 models + intelligent routing)
Speech-to-Text	Voxtral Small (Mistral-native)
Text-to-Speech	ElevenLabs (Turbo v2.5 + Multilingual v2)
Integrations	Composio V3 Tool Router Session API
Charts	Custom SVG generation engine
Markdown	react-markdown + remark-gfm
Deployment	Vercel

---

🏃 Quick Start

git clone https://github.com/MimiTechAi/missi.git
cd missi
npm install

# Configure environment
cp .env.example .env.local
# Add your API keys (see below)

npm run dev
# Open http://localhost:3000

Environment Variables

MISTRAL_API_KEY=your_mistral_api_key
ELEVENLABS_API_KEY=your_elevenlabs_api_key
COMPOSIO_API_KEY=your_composio_api_key    # Optional: for Gmail, Calendar, GitHub

---

🏆 Hackathon Categories

• 🌍 Global Winner — Full Mistral ecosystem: 6 models + 28 tools + autonomous planning
• 🎙️ Best Voice Use Case (ElevenLabs) — Continuous voice conversation in 10 languages
• 🤖 Best Use of Agent Skills — 28 tools with multi-step autonomous execution
• 🔗 Best Integration (Composio) — Gmail, Calendar, GitHub + 10,000+ tools

---

👥 Team

MiMi Tech AI — Michael Bemler & Michael Soppa

🌐 mimitechai.com · 🐙 github.com/MimiTechAi

---

📄 License

MIT

---

Built with ❤️ for the Mistral AI Worldwide Hackathon 2026