terrytangyuan/aisuite

Aisuite lets you talk to many different AI models through one consistent interface. Instead of learning each provider's specific API and rewriting your code when you want to switch or compare models, you write your code once and can swap between providers like OpenAI, Anthropic, Google, AWS, and others with minimal changes. At a high level, the library wraps the existing Python client libraries for each supported provider. You specify which model you want to use with a simple format like openai:gpt-4o or anthropic:claude-3-5-sonnet-20240620, and the library handles calling the right provider with the right parameters. The interface mirrors the OpenAI API style, so if you have used that before, the code will look familiar. You set your API keys as environment variables, create a client, and call client.chat.completions.create() with your messages and the model you want. This is useful for developers and product teams who want to compare responses from different AI providers side by side, or who want the flexibility to switch providers without rewriting their application code. For example, if you are building a chatbot and want to test whether GPT-4o or Claude gives better responses for your use case, you can run the same prompt through both in a loop and compare. It also future-proofs your work: if a new provider comes out next month, you can add it without overhauling your existing code. The project is deliberately kept lightweight, calling itself a "thin wrapper" rather than a heavy framework. Right now it focuses on chat completions, which covers the most common use case of sending messages and getting text back. The README notes that more use cases will be added over time. Adding support for a new provider follows a straightforward naming convention, so the project is designed to grow its list of supported providers through community contributions.