AI You Control: Choose your models. Own your data. Eliminate vendor lock-in.

Thunderbolt is an open-source, cross-platform AI client designed for enterprises seeking on-premise deployment to maintain data ownership and eliminate vendor lock-in. It empowers users to choose and integrate their own AI models, supporting local, frontier, and on-premise options for maximum flexibility.

How It Works

This client provides a unified interface for interacting with diverse AI models. Users self-host a backend using Docker Compose or Kubernetes, configuring model providers via settings. Options include OpenAI-compatible APIs or local inference engines like Ollama and llama.cpp. This architecture prioritizes user control over data and infrastructure, enabling custom deployments.

Quick Start & Requirements

• Installation involves self-hosting the backend using Docker Compose or Kubernetes.
• Prerequisites include setting up model providers (e.g., Ollama, llama.cpp, or OpenAI-compatible APIs) and potentially disabling default authentication/search features during local testing.
• Official documentation, FAQ, deployment guides, development setup, and architecture diagrams are available.

Highlighted Details

• Cross-platform availability on web, iOS, Android, Mac, Linux, and Windows.
• Compatibility with a wide range of frontier, local, and on-premise AI models.
• Enterprise features, dedicated support, and Full Disk Encryption (FDE) are offered.
• The project is currently undergoing a security audit and preparing for enterprise production readiness.

Maintenance & Community

The project is under active development, with ongoing efforts to enhance documentation, build community channels, and refine the roadmap. Users are encouraged to file GitHub issues for support, and contributions are welcomed from everyone.

Licensing & Compatibility

Licensed under the Mozilla Public License 2.0 (MPL 2.0). This permissive license allows commercial use and modification, provided derivative works are shared under the same license terms.

Limitations & Caveats

The project is in an early development stage and not yet fully offline-first, requiring an internet connection for authentication and search functionality (though these can be disabled). Users must provide and configure their own model inference endpoints.