# Third-party software notices The project depends on third-party npm packages whose own licenses apply. This file records the packages added for the browser-local SDR and on-device voice features; the complete resolved dependency inventory remains in `package-lock.json`. ## Web RTL-SDR - Package: `@jtarrio/webrtlsdr` 3.0.6 - Author: Jacobo Tarrio Barreiro; portions copyright Google Inc. - Source: - License: Apache License 2.0 - License text: ## Signals - Package: `@jtarrio/signals` 0.10.1 - Author: Jacobo Tarrio Barreiro - Source: - License: Apache License 2.0 - License text: ## On-device voice Loaded only when voice runs on this device. | Package | Version | Source | License | | --- | --- | --- | --- | | `@litert-lm/core` (runtime and wasm served from the build) | 0.17.1 | | Apache-2.0 | | `@huggingface/transformers` (speech recognition) | 4.3.0 | | Apache-2.0 | | `onnxruntime-web` (with transformers.js 4.3.0) | 1.31.0-dev.20260914 | | MIT | | `kokoro-js` (natural voice) | 1.2.1 | | Apache-2.0 | | `@huggingface/transformers` (used by kokoro-js) | 3.8.1 | | Apache-2.0 | | `onnxruntime-web` (used by kokoro-js) | 1.22.0-dev.20250409 | | MIT | Natural voice also uses `phonemizer` 1.2.1 (), which embeds eSpeak NG (, GPL-3.0-or-later). It is not part of this repository or its build: when natural voice first starts, the user's browser downloads the pinned file from `cdn.jsdelivr.net/npm/phonemizer@1.2.1` and checks its SHA-256 before use. `GEV_NATURAL_VOICE=off` turns natural voice off.