KAI (Knowledge Augmented Intelligence) is an AI augmented engineering meta-framework that supports human in the loop development. The goal is to give engineers a predictable and modular way to work with large language models inside real software projects.
KAI focuses on three ideas.
- Contract driven development for predictable outputs
- Structured feedback loops between the model and the engineer
- Parallel feature workflows through worktrees and stacked pull requests
The framework is in early development. The repository currently contains a placeholder. Documentation and reference implementations will be added as the project matures.
KAI aims to provide a clear path for integrating model assisted workflows into greenfield and brownfield codebases. It will include examples, templates, conventions and helper scripts that support:
- deterministic code generation
- multi feature parallelization
- reproducible orchestration patterns
Active development is ongoing. The initial public version will appear once the core patterns complete internal testing.
Contributions will open once the first stable milestone is ready. Until then, feedback and ideas are welcome through the issue tracker.
KAI is licensed under the MIT License.