instructor

Extracts structured data from LLM responses using Pydantic validation and automatic retries.

Updated Apr 16, 2026

Search Tags:#pydantic #validation #llm #streaming #retry #structured-output #instructor

Ready to run

npx skills add https://github.com/jacardl/New-Radar --skill instructor-jacardl

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: instructor
Source: https://github.com/jacardl/New-Radar/tree/main/backend/frameworks/hermes-agent/optional-skills/mlops/instructor
Command: npx skills add https://github.com/jacardl/New-Radar --skill instructor-jacardl

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Extract structured data from LLM responses by applying Pydantic validation, providing a reliable and auditable data extraction workflow that reduces surface errors and inconsistencies.

Core Features & Use Cases

Structured output extraction with automatic validation against Pydantic models.
Automatic retries on validation failures and error-driven re-prompts to improve accuracy.
Streaming support to progressively deliver partial results for real-time processing across multiple providers.

Quick Start

Instruct the AI to return data that matches a predefined Pydantic model and let Instructor validate and stream the results.

Frequently Asked Questions about instructor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I extract structured data from LLM responses with Pydantic validation?▼

To extract structured data from LLM responses, you use the Instructor library with predefined Pydantic models to validate the outputs, ensuring type safety and reducing surface errors in production workflows.

What happens when LLM structured output extraction fails validation?▼

When structured output extraction fails Pydantic validation, the system automatically retries the failed extraction and issues error-driven re-prompts to the LLM to correct the output and improve accuracy.

Can I stream partial structured data results from an LLM in real-time?▼

Yes, you can stream partial structured data results in real-time across multiple LLM providers. The Instructor library provides streaming support to progressively deliver partial extractions for immediate processing.

Do I need predefined Pydantic models to parse complex JSON from LLMs?▼

Yes, predefined Pydantic models are required. They act as the target schema for type safety, enabling the Instructor library to validate parsed complex JSON and ensure the LLM output matches your expected data structure.

What is the best way to handle inconsistent JSON formatting from LLM outputs?▼

The best way to handle inconsistent JSON from LLMs is to apply Pydantic validation with Instructor, which automatically retries failed extractions and uses error-driven re-prompts to enforce a reliable, structured output format.

Does structured output extraction work across multiple LLM providers?▼

Yes, structured output extraction works across multiple LLM providers. The system supports streaming partial results and applying Pydantic validation regardless of the underlying provider used for the generation.