AI Agent Framework
AI Agent Framework is a TypeScript framework for building AI applications with composable primitives and an agent runtime.
The project is currently focused on a practical MVP:
- runnable-based composition (
RunnableLambda,RunnableSequence,RunnableParallel,RunnableMap) - prompt and output parser layers
- model adapter runnable
- tool-capable agent runtime
What This Documentation Covers
This documentation is designed for three audiences:
- Developers integrating the framework into AI apps.
- Contributors extending primitives and runtime behavior.
- Reviewers validating architecture and release quality.
If you are new, start with:
If you are implementing or reviewing changes, start with:
Current MVP Snapshot
As of v0.1.0-mvp:
- Runnable composition works end-to-end.
- Prompt formatting and JSON parsing layers are available.
- OpenAI provider adapter is integrated.
- Agent runtime supports tool execution and memory context loop.
- Core behavior is covered by unit tests and contract tests.