Switch between OpenAI, Anthropic, and Google by changing one line — a library that erases the pain of learning each provider's SDK 🔌 Title: andrewyng/aisuite URL: 🔌 Overview A lightweight Python library by Andrew Ng. It lets you call multiple LLM providers through a unified, OpenAI-style interface. Just specify the model as "provider:model-name" and run the same code across providers. ❓ Challenges Solved Working with multiple LLM providers means memorizing distinct SDKs, APIs, and parameter structures — real development friction. ・aisuite solves this fragmentation problem ・It standardizes requests and responses so you focus on logic, not SDK differences 💡 Key Features ・Chat Completions API: a provider-agnostic layer (supports temperature, max_tokens, tools) ・Agents API: structured agents with tool handling, state persistence, and governance via tool policies ・Toolkits: pre-sandboxed tool families for files, git, and shell operations ・MCP support: natively use any MCP server's tools ・OpenCoworker: a desktop AI agent app built on aisuite It uses a provider adapter pattern, auto-discovering and instantiating providers from the model string per naming conventions. 🌍 Supported Providers / Usage Supports OpenAI, Anthropic, Google, Mistral, Hugging Face, AWS, Cohere, Ollama, OpenRouter, and more (extensible). Create ai.Client() and call with model="openai:gpt-4o" — switching providers is just changing the model string. MIT-licensed. #LLM# #OpenSource#