A modular, extensible AI chat engine built with a pipeline-based architecture.
Battle-tested in production - Powers Resonate, a modern AI chat application.
This repository is a monorepo containing the following packages:
| Package | Description | Version |
|---|---|---|
kai_engine |
The core AI chat engine with a pipeline-based architecture | TBD |
kai_engine_firebase_ai |
Firebase AI adapter for the Kai Engine | TBD |
The Kai Engine is a flexible framework for building AI-powered chat applications with a clean, modular architecture. It follows a pipeline-first pattern, allowing developers to easily customize and extend the processing pipeline with domain-specific logic.
The core framework provides essential abstractions for building conversational AI applications while remaining unopinionated about concrete implementations, allowing maximum flexibility.
- Modular Pipeline Architecture: Each processing step is a separate component that can be customized or replaced.
- Extensible Design: Unlimited extensibility through component composition.
- Generic Type Support: Full generic support for using your own message types with MessageAdapter.
- Stream-Based Responses: Real-time streaming responses for better user experience.
- Optimistic UI Updates: Immediate UI feedback with rollback on errors.
- Flexible Context Building: Advanced prompt engineering with parallel and sequential context building.
- Tool Calling Support: Native support for AI function/tool calling with type-safe schemas.
- Template Engine: Built-in flexible template engine for dynamic content generation.
- Post-Response Processing: Process AI responses after generation with custom pipelines.
- Type Safety: Strong typing throughout the system for better developer experience.
- Comprehensive Testability: Designed for easy unit and integration testing.
Add the dependency to your pubspec.yaml:
dependencies:
kai_engine: ^1.0.0For Firebase AI integration:
dependencies:
kai_engine_firebase_ai: ^1.0.0See the individual package READMEs for detailed documentation:
Contributions are welcome! Please read our Contributing Guide for details on how to submit pull requests, report issues, or request features.
This project is licensed under the MIT License - see the LICENSE file for details.