Open Source
License &
Dependencies
ptarmigan-flow is free and open-source software. It builds on a rich ecosystem of local AI and macOS tooling — all listed here with their respective licenses.
Project license
MIT License
ptarmigan-flow is released under the MIT License. You are free to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the software, subject to the conditions below.
Copyright © 2024–2025 ptarmigan-flow Contributors
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Dependencies
Key dependencies
All runtime packages run locally on your Mac. No data is sent to third-party servers by any of these libraries.
Core
- moonshine-voice — Apache-2.0 Efficient English speech recognition by Useful Sensors.
- transformers — Apache-2.0 Hugging Face transformers; used by the Granite and vLLM backends.
- mlx-whisper — MIT OpenAI Whisper optimised for Apple Silicon via MLX.
- mlx-audio — MIT MLX-native audio processing; powers the Voxtral backend.
- voxmlx — Apache-2.0 Voxtral real-time speech model bindings for Apple Silicon.
Audio & input
- sounddevice — MIT Cross-platform audio I/O via PortAudio. Used for microphone capture.
- pynput — LGPL-3.0 Global keyboard monitoring for push-to-talk hotkey detection.
- pyperclip — BSD-3-Clause Clipboard integration for the clipboard-paste output mode.
macOS integration
- pyobjc-core — MIT Python ↔ Objective-C bridge. Foundation for all macOS system calls.
- pyobjc-framework-Cocoa — MIT AppKit, Foundation, and CoreGraphics bindings.
- pyobjc-framework-WebKit — MIT WKWebView bindings; powers the macOS app's built-in browser window.
- pyobjc-framework-AVFoundation — MIT Microphone permission handling via AVCaptureDevice.
Configuration & utilities
- pydantic — MIT Type-validated config models and runtime settings.
- tomli-w — MIT TOML serialisation for writing the config file.
- numpy — BSD-3-Clause Numerical arrays for audio buffer processing.
- websockets — BSD-3-Clause WebSocket server for vLLM realtime audio streaming.
- mistral-common — Apache-2.0 Common utilities for Mistral/Voxtral model inference.
For the complete, authoritative list of all transitive dependencies and their exact versions, see pyproject.toml in the repository.
Models
STT model licenses
Models run entirely on your Mac. They are downloaded from Hugging Face Hub on first use. Each model is subject to its own license — see the relevant Hub page for the full terms.
| Model | Provider | License |
|---|---|---|
| Moonshine tiny / base | Useful Sensors | Apache-2.0 |
| Granite 4.0 1B Speech | IBM | Apache-2.0 |
| Whisper large-v3-turbo | OpenAI | MIT |
| Voxtral Mini 4B | Mistral AI | Apache-2.0 |
| vLLM (custom) | User-supplied | Varies by model |
ptarmigan-flow does not bundle or redistribute any model weights — they are fetched directly from Hugging Face or your configured vLLM endpoint.