# Instructor

> A Python library for structured data extraction from LLMs using Pydantic validation and automatic retries.

Instructor is a Python library that simplifies extracting structured data from large language models (LLMs). Built on top of Pydantic, it provides a seamless way to define data schemas and automatically validate LLM outputs, ensuring type-safe and reliable data extraction. The library supports automatic retries with error correction, making it robust for production use cases.

- **Pydantic Integration** - Define your data models using familiar Pydantic syntax, and Instructor handles the conversion to LLM-compatible prompts and validates responses automatically.

- **Automatic Retries** - When LLM outputs don't match your schema, Instructor automatically retries with error context, improving success rates without manual intervention.

- **Multi-Provider Support** - Works with OpenAI, Anthropic, Google, Cohere, Mistral, and other major LLM providers through a unified interface.

- **Streaming Support** - Extract structured data from streaming responses, enabling real-time data processing and partial results.

- **Validation Hooks** - Add custom validators to your Pydantic models for complex business logic validation beyond type checking.

- **Multimodal Capabilities** - Extract structured data from images and other multimodal inputs supported by compatible LLMs.

- **Parallel Extraction** - Process multiple extraction tasks concurrently for improved throughput in batch processing scenarios.

To get started, install Instructor via pip with `pip install instructor`. Import the library, patch your OpenAI client, and define a Pydantic model for your desired output structure. Call the patched client with your model as the `response_model` parameter, and Instructor handles the rest. The library includes comprehensive documentation with examples for common use cases like entity extraction, classification, and data transformation.

## Features
- Structured data extraction from LLMs
- Pydantic model validation
- Automatic retry with error correction
- Multi-provider LLM support
- Streaming response handling
- Custom validation hooks
- Multimodal input support
- Parallel extraction processing
- Type-safe outputs
- OpenAI function calling integration

## Integrations
OpenAI, Anthropic, Google AI, Cohere, Mistral, Pydantic, Python

## Platforms
DEVELOPER_SDK

## Pricing
Open Source

## Links
- Website: https://www.useinstructor.com
- Documentation: https://www.useinstructor.com
- Repository: https://github.com/jxnl/instructor
- EveryDev.ai: https://www.everydev.ai/tools/instructor