Jason

Instructor is an open-source library that helps developers extract structured data from large language models with type safety, validation, and greater control. Created by Jason Liu, it is free to use under the MIT license and is designed to make structured outputs simpler to build into AI-powered applications.

The library provides consistent APIs for Python, TypeScript, PHP, Ruby, Elixir, and Go. It supports integrations with OpenAI and offers adapter patterns for providers and tools including Anthropic, Google, Vertex AI, Mistral, Ollama, and others. Developers can define expected output structures using type systems such as Pydantic and Zod, then validate model responses and automatically retry when results do not meet requirements.

Instructor also includes streaming for partial results and iterables, prompt templates, custom validators, and event hooks for logging, monitoring, and modifying interactions. Its documentation, examples, API reference, and community resources help users get started and develop their implementations. The project welcomes contributors to areas such as documentation, testing, and feature development.