Repolode
Submit
567-labs

567-labs/instructor

A Python library that gets structured, Pydantic-validated data from LLMs across multiple providers, with automatic retries on validation failure.

View on GitHubWebsite
Stars
14,046
Forks
1,330
Open issues
214
Last commit
yesterday

What is it?

Instructor extracts typed data from LLM responses by letting you define the desired output as a Pydantic model. Instead of hand-writing JSON schemas and parsing raw text, you pass a response_model to the client and receive a validated Python object back.

The library wraps the APIs of several providers, including OpenAI, Anthropic, Google, Ollama and Groq, behind a single interface created with from_provider. When validation fails, Instructor sends the error message back to the model and retries, up to a configurable max_retries. It also supports streaming partial objects and nested models.

Besides Python, the project's documentation lists TypeScript, Ruby, Go, Elixir and Rust implementations.

Key features

  • 01Defines output schemas with Pydantic models and returns validated objects.
  • 02Works with multiple LLM providers through one from_provider interface.
  • 03Retries failed validations automatically, feeding the error back to the model.
  • 04Streams partial objects as they are generated.
  • 05Supports nested models and lists of models.

Use cases

  • Extracting fields such as names, dates or prices from free-form text.
  • Turning LLM outputs into typed data for downstream application code.
  • Building pipelines that need reliable JSON from models without manual parsing.

Quick start

  1. 1

    Install the library with pip

    pip install instructor
  2. 2

    Create a client for your provider and extract a Pydantic model

Things to know

  • Requires an API key or a running LLM provider such as OpenAI, Anthropic, or a local Ollama server.

Topics

#llm#structured-outputs#pydantic#python#openai#validation#json#openai-function-calli#openai-functions#pydantic-v2

FAQ

What is instructor?
A Python library that gets structured, Pydantic-validated data from LLMs across multiple providers, with automatic retries on validation failure. Instructor extracts typed data from LLM responses by letting you define the desired output as a Pydantic model. Instead of hand-writing JSON schemas and parsing raw text, you pass a response_model to the client and receive a validated Python object back.
Is instructor free to use? What license does it have?
instructor is open source under the MIT license. Check the license terms before using it in commercial products.
Is instructor still maintained?
Actively maintained. The last commit to instructor was on Oct 20, 2026.
What can I use instructor for?
Extracting fields such as names, dates or prices from free-form text. Turning LLM outputs into typed data for downstream application code. Building pipelines that need reliable JSON from models without manual parsing.
How do I install instructor?
Install the library with pip: pip install instructor Create a client for your provider and extract a Pydantic model
Where can I get instructor?
The source code and installation instructions are on GitHub: https://github.com/567-labs/instructor

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects