​
Why Use an SDK?

SDKs aim to simplify the integration process by providing:

  • Abstraction: Wrappers around the direct HTTP API calls, making it easier to call from your application code.
  • User Identification: Helpers for managing userId and ensuring consistent variant assignment for users across sessions if needed.
  • Context Management: Utilities for easily capturing and sending relevant context with assignment and log calls.
  • Simplified Logging: Easier methods to log inputs, outputs, and custom metrics without needing to manually construct the JSON request bodies.
  • Potential for Local Caching: Some SDKs might offer client-side caching of variant assignments to reduce latency or network requests for repeated assignments.

​
Current Status & Direct API Usage

The initial open-source release of AgentTest focuses on providing this robust and flexible backend server. Official, full-featured SDKs for various languages and frameworks are a key part of our future roadmap.

In the meantime, you can integrate AgentTest by making direct calls to the HTTP API endpoints:

Refer to the API Reference sections for detailed information on request and response formats.

​
Planned SDKs

We plan to develop and release SDKs for popular languages and AI frameworks, including:

  • Python SDK: For general Python applications and popular AI/ML libraries.
  • JavaScript/TypeScript SDK: For Node.js backends and potentially front-end applications (though careful consideration of API key security is needed for client-side usage).
  • LangChain Integration: An SDK or module specifically designed for easy integration with LangChain projects, allowing you to A/B test chains, prompts within chains, or agent components.
  • LangGraph Integration: Similar to LangChain, specific support for LangGraph to experiment with different graph configurations or agent states.

​
Community & Contributions

The development of SDKs is an area where community contributions can be highly impactful. If you are interested in:

  • Contributing to the development of an official SDK.
  • Building your own SDK for a language or framework not yet covered.
  • Sharing feedback on what features you’d like to see in an SDK.

Please join our community discussions or check out the AgentTest GitHub repository (link to be updated with actual SDK repo or main org repo when available).