Instructor Provides a Free API to Get Structured Output from LLMs

Instructor is a tool that patches OpenAI, Anthropic, and other LLM clients to return Pydantic models instead of raw text, ensuring structured, validated, and typed output every time.

Details

The article discusses the challenges of working with large language models (LLMs) that return unstructured text output, which can be difficult to parse and validate. To address this, the author introduces Instructor, a tool that patches popular LLM clients like OpenAI to return Pydantic models instead of raw text. Pydantic models are Python classes that define the structure and types of the data, ensuring the output is always valid and easy to work with. Instructor supports complex data structures like nested objects and lists, allowing developers to extract structured information from LLMs with ease. The article provides code examples demonstrating how to use Instructor to get a validated User object from an LLM response, as well as more complex examples involving nested Address and Company models. Overall, Instructor aims to simplify the process of working with LLMs and make it easier to integrate their capabilities into applications.