The Problem: Current AI meeting tools (Otter.ai, Fireflies) require visible bots joining your calls and send all audio to their servers. This is awkward for client calls and impossible for GDPR/HIPAA environments.

How Meetily Works: • Captures system + microphone audio locally (no visible bot in meetings) • Runs Parakeet / Whisper.cpp for live transcription (GPU-accelerated) • Generates summaries via local LLMs (Ollama) or API (Claude/GPT) • Everything stays on your hardware – zero cloud dependency

Tech Stack: • Desktop: Tauri (Rust) + Next.js • Audio: System-level capture (no virtual cameras/bots) • Models: Parakeet, Whisper, Open source LLMs, OpenAi / Claude / Groq • Platforms: macOS, Windows, Linux (You might have to build. Build instructions available in Repo)

It's MIT licensed and designed for teams that can't send sensitive data to third parties (Gov, legal, healthcare, finance).

GitHub: https://github.com/Zackriya-Solutions/meeting-minutes Website: https://meetily.ai

Would love feedback on the architecture, especially around audio capture reliability and model performance trade-offs!