🎙️ Menie

100% Local-First AI Meeting & Audio Intelligence Desktop Assistant

Key Features • Architecture • Upcoming Roadmap • Getting Started • Contributing

Provenance: Menie is an experimental derivative of Meetily by Zackriya Solutions. The upstream MIT copyright is preserved. See ATTRIBUTION.md before citing or contributing to this repository.

---

🌟 What is Menie?

Menie is an open-source, ultra-private, 100% local-first AI meeting assistant and audio intelligence desktop app. Built with Tauri v2, Next.js, Rust, and native hardware-accelerated speech engines (Whisper & Parakeet), Menie captures, transcribes, speaker-diarizes, and summarizes your meetings entirely on your device.

No audio or transcript data ever leaves your computer. No cloud subscription fees, no mandatory API keys, and zero telemetry tracking.

---

🛠️ Core Workflow & Pipeline

┌─────────────────────────────────────────────────────────────────────────────────┐
│                           MENIE 100% LOCAL AI PIPELINE                          │
├──────────────────┬───────────────────────┬───────────────────┬──────────────────┤
│  1. Audio Input  │  2. Transcription     │  3. Diarization   │  4. Intelligence │
├──────────────────┼───────────────────────┼───────────────────┼──────────────────┤
│ • Mic / WebCam   │ • NVIDIA Parakeet TDT │ • Neural Speaker  │ • Local LLM      │
│ • System Audio   │ • Whisper ONNX        │   Embeddings      │   Summarization  │
│ • File Imports   │ • GPU Accelerated     │ • Auto-Clustering │ • SQLite FTS5    │
└──────────────────┴───────────────────────┴───────────────────┴──────────────────┘

---

✨ Key Features

🔒 100% On-Device Privacy & Offline First

• Zero Cloud Dependence: Transcriptions and AI summaries run 100% locally on your CPU/GPU.
• Air-Gapped Operation: Use Menie in confidential environments without internet connectivity.
• SQLite Database with Encryption: Your transcripts stay securely on your local file system.

🎙️ Real-Time & Batch Audio Intelligence

• Dual Channel Capture: Capture microphone input and system audio simultaneously.
• Multi-Format Import: Import existing audio files (.mp3, .wav, .m4a, .flac, .ogg, .webm).
• Parakeet & Whisper ONNX Engines: Sub-second fast transcription using NVIDIA Parakeet TDT or Whisper models.

⚡ Local Hardware Acceleration

• Apple Silicon Metal: Native Metal GPU and CoreML acceleration for macOS.
• NVIDIA CUDA & Vulkan: High-throughput GPU offloading on Windows and Linux.
• OpenBLAS CPU Fallback: Optimized multi-threaded execution for standard CPUs.

🧠 Local LLM Meeting Summarization

• llama-helper & Ollama Integration: Generate action items, meeting minutes, decision logs, and key takeaways using local GGUF models (LLaMA 3, Qwen, Mistral, Gemma).
• Custom Summarization Templates: Tailor summary structures for 1-on-1s, engineering standups, executive syncs, or sales calls.

👥 Speaker Diarization & Labeling

• Speaker Attribution: Track distinct speakers during real-time meetings or imported recordings.
• Custom Speaker Profiles: Assign names, roles, and colors to meeting participants.

🔍 Search & Export Flexibility

• Full-Text SQLite FTS5 Search: Instant full-text search across thousands of hours of historical meeting transcripts.
• Rich Export Options: Export meeting minutes to Markdown, PDF, JSON, or send payload webhooks to custom local endpoints.

---

🏗️ Architecture

                               ┌────────────────────────────────────────┐
                               │           Menie Frontend               │
                               │      Next.js 14 + React + Tailwind     │
                               └───────────────────┬────────────────────┘
                                                   │ IPC (Tauri Bridge)
                               ┌───────────────────▼────────────────────┐
                               │            Tauri Rust Core             │
                               │  CPAL Audio Engine + SQLite FTS5 Store │
                               └─────────┬──────────────────────┬───────┘
                                         │                      │
                   ┌─────────────────────▼──────┐        ┌──────▼─────────────────────┐
                   │    Whisper / Parakeet RS   │        │     llama-helper Engine    │
                   │ (Metal / CUDA / Vulkan GPU)│        │ (Local GGUF LLM Execution) │
                   └────────────────────────────┘        └────────────────────────────┘

---

🚀 Upcoming Roadmap

Here are the upcoming features planned for Menie's future releases:

• 🎯 Real-Time Acoustic Speaker Separation: Live acoustic embedding extraction to automatically differentiate multi-speaker conversations without prior training.
• 📚 Local Vector RAG Knowledge Base: Semantic vector search across your entire historical meeting archive to ask questions across months of meeting history offline.
• 📱 Mobile Companion Local Handoff: Seamless peer-to-peer Wi-Fi transfer from mobile voice memos to your desktop workstation.
• ⚡ Automated Local Action Pipelines: Direct integration with local note-taking apps like Obsidian, Logseq, and Notion Local.
• 🌐 Offline Live Streaming Translation: Real-time cross-lingual translation stream during live international conference calls.

---

💻 Getting Started

Installation

Download the latest pre-compiled desktop binary for your operating system from Releases.

macOS

1. Download Menie-aarch64.dmg (Apple Silicon) or Menie-x64.dmg (Intel).
2. Drag Menie.app into your Applications folder.

Windows

1. Download Menie-x64-setup.exe.
2. Run the installer and launch Menie.

---

🛠️ Building from Source

Prerequisites

• Node.js (v18+)
• pnpm (npm install -g pnpm)
• Rust (1.77+)

Quick Build Instructions

# Clone the repository
git clone https://github.com/0xSuleman/Menie.git
cd Menie/frontend

# Install frontend dependencies
pnpm install

# Run desktop dev environment
pnpm tauri:dev

To build a standalone production release:

pnpm tauri:build

---

📄 License

Menie is open-source software licensed under the MIT License.

---

Made with ❤️ by 0xSuleman