SayIt

No more copying and pasting to ChatGPT! SayIt is always there, understanding what you're doing and ready to help. Working with local models for full privacy, or cloud for power. On Mac, Linux, Windows

Comment

SayIt - Local AI Assistant Agent

Inspiration

The inspiration for SayIt came from the growing need for privacy-focused AI assistants that you can interact with using your voice. It's faster, more natural and funnier than typing.

What it does

SayIt is a cross-platform desktop AI assistant that:

  • Runs AI models locally on your machine for enhanced privacy
  • Provides system-wide assistant integration across Windows, macOS, and Linux
  • Offers real-time AI processing with multiple model support
  • Features high-quality text-to-speech synthesis
  • Integrates seamlessly with the operating system
  • Maintains user privacy through local processing
  • Supports customizable settings and model parameters

How we built it

We developed SayIt using a modern tech stack centered around Electron, React, and TypeScript. The architecture is built on the Redux Saga pattern, which helps manage complex AI operations and state management. Key components include:

  • Electron for cross-platform desktop support
  • React with TypeScript for a robust frontend
  • Redux + Redux Saga for state management and handling async operations
  • Integration with Hugging Face Transformers for AI capabilities
  • ElevenLabs for advanced text-to-speech features
  • Electron Vite for optimized build tooling
  • Tailwind CSS for modern UI design

Challenges we ran into

While developing SayIt, we faced several technical challenges:

  • Optimizing performance for local AI model processing
  • Implementing seamless system-wide integration across different operating systems
  • Managing complex asynchronous AI operations while maintaining application responsiveness
  • Balancing feature richness with resource efficiency
  • Ensuring consistent behavior across Windows, macOS, and Linux platforms

Accomplishments that we're proud of

We successfully created an agent that:

  • Can run completely locally, protecting user privacy
  • Supports multiple AI models and customization options
  • Achieves seamless system-wide integration
  • Maintains high performance despite complex processing requirements
  • Provides a clean, modern user interface
  • Works consistently across all major desktop platforms

What we learned

Through building SayIt, we gained valuable experience in:

  • Managing complex AI processing flows in desktop applications
  • Implementing the Redux Saga pattern for robust state management
  • Creating cross-platform applications with Electron
  • Optimizing performance for local AI model processing
  • Balancing feature richness with resource efficiency

What's next for SayIt

Future development plans include:

  • More advanced and seamless personas feature
  • Middleware LLM for better transcription
  • Integrations with system and external resources

Who are we?

  • Michał Warda - Software Engineer, AI enthusiast. Responsible in the project for overall vision and architecture.
  • Paweł Sierant - Software Engineer. Responsible for model integration and work on API connections.
  • Dawid Kiełbasa - Software Engineer. Responsible for views, styling and design of application, integration with different systems.

Built With

Share this project:

Updates