Meetily is an open-source, privacy-focused, on-device AI meeting assistant. Unlike typical cloud-based meeting summarization services that transmit audio data to external servers, raising privacy concerns, this tool performs 100% local audio capture, real-time transcription, summarization, and question answering within the user's personal computer (PC). Its core architecture is designed around the desktop cross-platform framework Tauri, combining Next.js for the front-end and Rust for the back-end, which requires high-performance computing. This allows it to operate offline, completely disconnected from external networks, and leverages Whisper and Parakeet ONNX models locally to ensure excellent processing speed. Existing AI meeting assistant tools primarily use a method of recording by having a virtual bot participate in video conferences, which has the limitations of visually exposing the recording to other participants and being restricted by corporate security policies. Meetily adopts a bot-free approach, directly capturing system audio in the background, allowing secure and confidential transcriptions to be obtained without the attention or security restrictions of other participants. This is like having your own private record-keeper that is not exposed externally, and it operates flexibly regardless of the video conferencing platform (Zoom, Teams, etc.). Furthermore, by using a local Ollama server and open-source large language models (LLMs) for summarization instead of external paid APIs, it eliminates long-term costs and guarantees complete data sovereignty. Biotech researchers and pharmaceutical developers conduct numerous confidential meetings that require extreme security, such as discussions on the structural formulas of new drug candidates before patent applications or the design of clinical trials. When using cloud-based tools, they had to rely on manual recording due to concerns about technology leakage, but by combining Meetily and Ollama's Llama 3 model, meeting minutes can be automatically generated in complete security on an offline workstation. The researcher activates the audio capture function, conducts the meeting, and then uses the local Whisper model to convert one hour of presentation audio data into high-quality Korean text in just over two minutes. The converted text is stored in a SQLite local database, allowing keyword searches at any time, and the core milestones and action items are quantitatively extracted through Ollama's summarization function and safely integrated as knowledge assets in the internal RAG (Retrieval-Augmented Generation) system.

Meetily is an open-source, privacy-focused, on-device AI meeting assistant. Unlike typical cloud-based meeting summarization services that transmit audio data to external servers, raising privacy concerns, this tool performs 100% local audio capture, real-time transcription, summarization, and Q&A within the user's personal computer (PC). Its core architecture combines the desktop cross-platform framework Tauri, the front-end framework Next.js, and Rust for the back-end, which requires high-performance computing.