Instructor: Complete Guide 2026
Python/TypeScriptAI Agent Framework8k+ stars
Overview
A lightweight library for getting structured outputs from LLMs using Pydantic models. Instructor patches LLM client libraries to return validated, typed Python objects instead of raw text, with automatic retries on validation failures.
Key Features
✓Pydantic-based structured output extraction
✓Automatic retry on validation failures
✓Streaming support for partial objects
✓Multi-modal input support
✓Provider-agnostic with OpenAI, Anthropic, and more
✓Parallel and batch extraction
Use Cases
- → Extracting structured data from unstructured text
- → Building type-safe LLM-powered APIs
- → Data transformation and classification
- → Form-filling and document parsing
Pros & Cons
Pros
- +Dead-simple API for structured LLM outputs
- +Excellent type safety and validation
- +Works with any LLM provider
- +Minimal overhead and dependencies
Cons
- -Focused on extraction rather than full agent workflows
- -Not a complete agent framework
- -Requires Pydantic knowledge for complex schemas
Frequently Asked Questions
What is Instructor?▾
A lightweight library for getting structured outputs from LLMs using Pydantic models. Instructor patches LLM client libraries to return validated, typed Python objects instead of raw text, with automatic retries on validation failures.
What language is Instructor built in?▾
Instructor is primarily built in Python/TypeScript.
Is Instructor good for production?▾
Instructor has 8k+ GitHub stars. Dead-simple API for structured LLM outputs for extracting structured data from unstructured text.