Instructor

What Does Instructor Do?

Instructor is a Python library that extracts typed, structured data from any LLM response using Pydantic schemas. Instead of parsing raw JSON strings manually, you define a Pydantic model and Instructor validates the LLM output against it — automatically retrying on validation failure. It works with OpenAI, Anthropic, Google Gemini, and local models, and returns real Python objects instead of strings.

Also see: LiteLLM · Haystack

What Is Structured Output from an LLM?

Structured output means the model returns data that conforms to a defined schema — a JSON object, a Pydantic model, or a typed dataclass — rather than free-form text. This makes LLM responses reliably parseable by downstream code without brittle string parsing. Instructor enforces this by validating the output and retrying the model call if the response doesn't match the schema.