BigChat

BigChat is a WinUI 3 chat application that provides a modern desktop interface for interacting with large language models. It is designed for experimentation, developer workflows, and local/cloud model usage on Windows.

Highlights

• Native WinUI 3 desktop UI with a responsive chat experience.
• Flexible provider support: local Ollama models and OpenAI-compatible services.
• Conversation management with persistent, on-device storage (SQLite + EF Core).
• Fine-grained model controls: temperature, top-p, max output tokens, frequency/presence penalties, and restore defaults.
• Streaming responses and custom Markdown rendering for readable replies.
• Open-source and extensible — aimed at developers and advanced users.

Features

• Modern WinUI 3 chat interface for creating and managing conversations.
• Multiple AI provider support:
  • Ollama (local model server via OllamaSharp)
  • OpenAI-compatible endpoints (use your API key)
• Local data store using Entity Framework Core with SQLite – conversations and messages are stored on-device.
• Settings page to configure provider, endpoints, API keys, and model parameters.
• Custom Markdown rendering using a modified MarkdownTextBlock from the Community Toolkit.

Prerequisites

• Windows 11 or later.
• .NET 10 SDK.
• Visual Studio (recommended) with WinUI 3 workload to build from source.
• For Ollama: install and run a local Ollama server before using Ollama models.
• For cloud providers: valid API key(s) and correct endpoint configuration in Settings.

Getting Started

1. Clone the repository.
2. Restore dependencies and build in Visual Studio or via the .NET CLI.
3. Run the app (BigChat.WinUI).
4. Open Settings and choose your AI provider. Configure endpoint, model ID, and API key if required.
5. Start a new conversation and type your message in the input box. Responses stream back as they are generated.

Configuring AI Providers

Open Settings to select the active provider and set provider-specific values:

• Ollama: set the local endpoint (default http://localhost:11434) and choose a local model.
• OpenAI-compatible: set endpoint, API key, and model ID.

Privacy & Security

• Conversation history is stored locally in a SQLite database by default.
• Model inference is performed by the configured provider. If you use a remote provider (OpenAI or Azure), your data is sent to that service according to its policy. Review provider privacy terms before connecting.

Limitations & Notes

• Experimental: the app relies on preview and alpha libraries and is intended for learning and experimentation.

License

Licensed under the MIT License. See LICENSE.txt for details.

Acknowledgments

• Built with WinUI 3, .NET 10, Entity Framework Core, and CommunityToolkit components.

---

Efficient, modern, and open.