0xplaygrounds/rig

Rig is a Rust library for building applications that use large language models, the kind of AI that powers tools like ChatGPT. If you want to write a program that can chat with users, answer questions, generate text, or take actions based on AI responses, Rig gives you a structured way to do that without writing a lot of repetitive setup code. The main idea behind the library is a unified interface. Instead of writing different code for each AI provider, like OpenAI, Anthropic, or Google, you write your application once and can swap the model underneath without changing much. The library currently supports more than 20 AI model providers. Similarly, if your application needs to search through a large collection of documents or data to find relevant context before answering a question, which is a common pattern called retrieval-augmented generation, Rig provides built-in connections to more than 10 different vector database services under the same kind of single interface. The library supports agents that can carry on multi-turn conversations, use external tools, stream responses back in real time, and handle both text generation and embedding workflows. It also supports audio transcription, audio generation, and image generation if the underlying model provider offers those capabilities. The core library is compatible with WebAssembly, which means it can run in environments beyond traditional servers. To add it to a Rust project, you run a single command and then write a few lines of code to configure a client and send prompts. The code example in the README shows creating an agent with a system prompt in under ten lines. Rig is used in production by a range of organizations according to the README, including medical research tools, incident management platforms, AI agent frameworks, and browser-based coding assistants. The project is open source and warns that it is still evolving quickly, so breaking changes between versions should be expected.