Brieva

Third-Party Notices

Last updated: August 8, 2026

Brieva is commercial software (© Brieva, all rights reserved) that stands on excellent open-source work. Everything below runs locally on your Mac; nothing is sent to Brieva or anyone else.

AI models

Models are downloaded once, at first run, directly from huggingface.co and sherpa-onnx release assets on GitHub - pinned to exact revisions and checksums - and are governed by their own licenses:

ModelDeveloperLicenseUsed for
Qwen3 4B (GGUF)Qwen team, Alibaba CloudApache-2.0Local language model for summaries and briefs (default)
Qwen3 8B (GGUF)Qwen team, Alibaba CloudApache-2.0Local language model, medium tier
Qwen3.5 35B-A3B (GGUF)Qwen team, Alibaba Cloud (GGUF by Unsloth)Apache-2.0Local language model, large tier
Parakeet TDT 0.6B v3NVIDIA (MLX conversion by mlx-community)CC BY 4.0On-device speech-to-text
pyannote segmentation-3.0 (ONNX)Herve Bredin (pyannote)MITSpeaker segmentation for diarization
3D-Speaker ERes2Net (ONNX)3D-Speaker team, Alibaba (speech lab)Apache-2.0Speaker embeddings for diarization

Speech-to-text is powered by Parakeet TDT 0.6B v3, created by NVIDIA and licensed under Creative Commons Attribution 4.0. Brieva uses the mlx-community conversion of the model to Apple's MLX format; the model weights are otherwise unmodified.

Bundled software

ComponentLicenseUsed for
llama.cppMITLocal LLM inference (summaries, briefs); includes ggml
MLXMITApple Silicon ML framework
parakeet-mlxApache-2.0Parakeet inference on MLX
CPython / python-build-standalonePSF-2.0Bundled Python runtime for transcription
Node.jsMIT-styleBundled runtime for the local service layer
sherpa-onnxApache-2.0Speaker diarization engine
ONNX RuntimeMITNeural-net inference for diarization
libsndfileLGPL-2.1-or-laterAudio file reading (dynamically linked)
soxrLGPL-2.1-or-laterAudio resampling (dynamically linked)
TauriMIT / Apache-2.0Application shell

License claims on this page were last verified against each source on August 8, 2026. Questions about licensing: support@brieva.app.