Skip to content

voice_assistant_licenses.md

Latest commit

 

History

History
304 lines (228 loc) · 8.16 KB

voice_assistant_licenses.md

File metadata and controls

304 lines (228 loc) · 8.16 KB

Voice Assistant - License Information and On-Premise Deployment

Version: 1.0
Status: Enterprise Feature
Date: December 2025

Overview

This document provides comprehensive license information for all libraries used in the ThemisDB Voice Assistant feature, with specific focus on open-source compliance and on-premise deployment suitability.

Core Libraries

1. Whisper.cpp (Speech-to-Text)

Attribute Details
Repository https://github.com/ggerganov/whisper.cpp
Purpose High-accuracy speech transcription
License MIT License
Copyright Copyright (c) 2022 Georgi Gerganov
Open Source ✅ Yes
Commercial Use ✅ Allowed
On-Premise ✅ Fully supported
Attribution Required (MIT terms)

Key Features:

  • C++ port of OpenAI's Whisper model
  • No external API dependencies
  • Runs entirely offline
  • 100+ languages supported
  • Multiple model sizes (tiny to large)

Compliance:

  • MIT License is one of the most permissive open-source licenses
  • Allows commercial use, modification, and distribution
  • No copyleft requirements
  • Can be used in proprietary software
  • Only requires attribution in documentation

2. Piper TTS (Text-to-Speech)

Attribute Details
Repository https://github.com/rhasspy/piper
Purpose Neural text-to-speech synthesis
License MIT License
Copyright Copyright (c) 2023 Michael Hansen
Open Source ✅ Yes
Commercial Use ✅ Allowed
On-Premise ✅ Fully supported
Attribution Required (MIT terms)

Key Features:

  • Neural TTS with ONNX models
  • Multiple voice profiles
  • High-quality synthesis
  • No cloud dependencies
  • Lightweight and fast

Compliance:

  • MIT License
  • Suitable for commercial deployment
  • Voice models also open-source (mostly MIT/CC-BY)
  • No usage restrictions

3. llama.cpp (LLM Inference)

Attribute Details
Repository https://github.com/ggerganov/llama.cpp
Purpose Large Language Model inference
License MIT License
Copyright Copyright (c) 2023 Georgi Gerganov
Open Source ✅ Yes
Commercial Use ✅ Allowed
On-Premise ✅ Fully supported
Status Already integrated in ThemisDB v1.3.0+

Key Features:

  • Efficient LLM inference in C++
  • Support for LLaMA, Mistral, Phi-3, and more
  • GGUF model format
  • CPU and GPU acceleration
  • No external API calls

Compliance:

  • MIT License
  • Already approved for ThemisDB use
  • Core library is MIT (models have separate licenses)

4. ONNX Runtime (Neural Network Engine)

Attribute Details
Repository https://github.com/microsoft/onnxruntime
Purpose Neural network inference engine (used by Piper)
License MIT License
Copyright Copyright (c) Microsoft Corporation
Open Source ✅ Yes
Commercial Use ✅ Allowed
On-Premise ✅ Fully supported

Key Features:

  • High-performance ML inference
  • Cross-platform support
  • Hardware acceleration
  • Industry standard

License Summary

All Core Libraries: MIT License ✅

The MIT License allows:

  • ✅ Commercial use
  • ✅ Modification
  • ✅ Distribution
  • ✅ Private use
  • ✅ Sublicensing

The MIT License requires:

  • ✅ License and copyright notice (attribution)

The MIT License does NOT require:

  • ❌ Disclosure of source code
  • ❌ Copyleft (sharing modifications)
  • ❌ Trademark use
  • ❌ Patent grant
  • ❌ Liability or warranty

On-Premise Deployment

✅ Fully Supported

All libraries are designed for and support on-premise deployment:

  1. No External Dependencies

    • All processing happens locally
    • No API calls to external services
    • No internet connection required after model download

  2. Privacy & Compliance

    • Data never leaves your infrastructure
    • GDPR/DSGVO compliant
    • Suitable for sensitive data (healthcare, government, finance)
    • No third-party data processing

  3. Cost Benefits

    • No per-request API fees
    • No usage metering
    • One-time model download
    • Unlimited usage

  4. Control & Flexibility

    • Full control over versions
    • Can customize and fine-tune models
    • No vendor lock-in
    • Offline capability

  5. Performance

    • Low latency (no network calls)
    • Scalable (add more hardware)
    • GPU acceleration supported
    • Predictable performance

Enterprise Suitability

✅ Approved for Enterprise Use

Criterion Status Notes
Open Source ✅ Yes All MIT licensed
Commercial Use ✅ Allowed No restrictions
On-Premise ✅ Supported Designed for it
Data Privacy ✅ Compliant No external calls
Vendor Lock-in ✅ None Open standards
Modifications ✅ Allowed Can customize
Distribution ✅ Allowed Can bundle
Support ✅ Available Active communities

Model Licenses

Important: While the inference libraries are MIT licensed, AI models may have separate licenses.

Whisper Models

  • Source: OpenAI / HuggingFace (ggerganov/whisper.cpp)
  • License: Apache 2.0 / MIT
  • Commercial Use: ✅ Allowed
  • Models: tiny, base, small, medium, large (all open)

Piper Voice Models

  • Source: rhasspy/piper-voices (HuggingFace)
  • License: Mostly MIT, some CC-BY
  • Commercial Use: ✅ Most allow commercial use
  • Voices: 50+ languages, 100+ voices

LLM Models (for llama.cpp)

Different models have different licenses:

Model License Commercial Use
LLaMA 2 Meta License ✅ Allowed (<700M users)
Mistral Apache 2.0 ✅ Allowed
Phi-3 MIT License ✅ Allowed
Gemma Google ToS ✅ Allowed

Recommendation: Use Mistral or Phi-3 for unrestricted commercial use.

Attribution Requirements

As per MIT License terms, include the following attribution:

This software uses the following open-source libraries:

Whisper.cpp - Speech Recognition
Copyright (c) 2022 Georgi Gerganov
MIT License - https://github.com/ggerganov/whisper.cpp

Piper TTS - Text-to-Speech Synthesis  
Copyright (c) 2023 Michael Hansen
MIT License - https://github.com/rhasspy/piper

llama.cpp - LLM Inference
Copyright (c) 2023 Georgi Gerganov
MIT License - https://github.com/ggerganov/llama.cpp

ONNX Runtime - Neural Network Inference
Copyright (c) Microsoft Corporation
MIT License - https://github.com/microsoft/onnxruntime

Compliance Checklist

For on-premise enterprise deployment:

  • All libraries are open-source
  • All libraries use permissive licenses (MIT)
  • Commercial use is explicitly allowed
  • On-premise deployment is supported
  • No external API dependencies
  • No usage restrictions or metering
  • Source code modifications allowed
  • Can be bundled with proprietary software
  • Attribution requirements are minimal
  • No copyleft obligations
  • No patent concerns
  • GDPR/DSGVO compliant architecture
  • Suitable for sensitive data processing
  • No vendor lock-in

Legal Disclaimer

This document provides information about the licenses of third-party libraries used in the ThemisDB Voice Assistant feature. It is provided for informational purposes only and does not constitute legal advice. For specific legal questions, please consult with a qualified attorney.

Last Updated: April 2026
Review: Recommended annually or when updating library versions

References

Contact

For questions about licenses or enterprise deployment: