Kimi K2 API Integration: A No-Fluff Getting Started

Kimi K2 is Moonshot AI's flagship Mixture-of-Experts model, capable of understanding images alongside text in the same chat-completions interface. This guide provides a straightforward approach to integrating K2 into your applications.

To get started with cURL, set your API key and send a request to the v1 endpoint with a JSON payload containing your model, messages, and desired token limit. For Python, use the OpenAI SDK to create a client with your API key and base URL, then call the chat completions endpoint with similar parameters.

The key difference between K2 and text-only models is its ability to process images natively. Provide image URLs in the message content array, and K2 will analyze them alongside the text input. Keep image sizes reasonable to optimize token usage.

For streaming responses, set the stream parameter to True and iterate through the returned chunks, printing each one. This allows for real-time processing of the model's output.

When defining function calls, structure your tools with a function name, description, and required parameters. The model can then invoke these functions as needed, making it easy to incorporate external APIs or custom logic into your applications.