Spiegami

Spiegami is a powerful and intuitive app that helps you understand and rewrite text using the power of on-device AI. Whether you need to summarize a long article, explain a complex concept, or change the tone of your writing, Spiegami has you covered.

Demo

Features

• Summarize: Get the key points of any text in a fraction of the time
• Explain: Simplify complex topics and understand them with ease
• Paraphrase: Rewrite text in your own words while retaining the original meaning
• Enhance: Improve the grammar, clarity, and flow of your writing
• Change Tone: Adapt your text to different audiences with casual and corporate tone options
• Multi-language Support: Work with text in English, Spanish, French, German, Italian, and Portuguese
• Privacy First: All processing is done locally on your device, ensuring complete privacy and data security
• Open Source: Built with open source technologies, including the Qwen2.5-0.5B-Instruct model

Internal App Functioning

Spiegami leverages the power of the Cactus package to run a sophisticated on-device AI model. Here's a brief overview of its inner workings:

1. Model Initialization: Upon launching the app, the AI model is downloaded from the URL specified in AppConfig.modelUrl and initialized with a context size of 2048 tokens. This ensures that the app is ready to process text efficiently.
2. User Input: The user provides the text they want to process and selects one of the available actions (e.g., Summarize, Explain, Paraphrase). Each action is associated with a specific prompt that guides the AI model's response.
3. Prompt Engineering: The app constructs a detailed prompt for the AI model by combining the user's text, the selected action's prompt, and the chosen language for the output. This ensures that the model receives clear instructions to generate the desired result.
4. On-Device Processing: The complete prompt is then sent to the on-device AI model for processing. All computations are performed locally, which means that the user's data remains private and secure.
5. Displaying Results: The AI model's response is streamed to the ResultScreen, where the user can view the processed text in real-time. The ResultScreen is responsible for constructing the full prompt from the user's text and the selected actions, sending it to the model, and rendering the response token by token. It also provides functionality to copy the result to the clipboard and displays important notices about privacy and the nature of AI-generated content. This provides a seamless and interactive user experience.

Platforms

Available for mobile devices only.

Installation

Mobile App Stores

Available on major mobile app stores (coming soon)

Building from Source

To build Spiegami from source, you'll need the Flutter SDK installed.

1. Clone the repository:

   git clone https://github.com/your-username/spiegami.git
   cd spiegami

2. Install dependencies:

   flutter pub get

3. Run the app:

   flutter run

Technology

Spiegami is powered by:

• Flutter - Cross-platform mobile framework
• Qwen2.5-0.5B-Instruct - On-device language model by Alibaba Cloud
• Apache 2.0 licensed open source technologies

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the Apache 2.0 License. See the LICENSE file for details.

Third-party Licenses

This app uses the following open source software:

• Qwen2.5-0.5B-Instruct - Licensed under Apache 2.0 License (Alibaba Cloud)