instructor

To extract structured data from an LLM response, validate the text against Pydantic models to ensure type-safe outputs. This approach automatically retries failed extractions by feeding validation errors back to the LLM for correction.

The best way to validate LLM outputs for type safety is using Pydantic models to structure the extraction. This ensures the generated data conforms to specific types and automatically handles validation failures with retry logic.

Yes, you can stream partial structured data from LLMs in real time. Streaming support allows you to process partial Pydantic model results incrementally as the LLM generates the response, enabling immediate downstream data consumption.

Structured data extraction works across multiple LLM providers through provider adapters. It supports multi-LLM backends like OpenAI and Anthropic, ensuring consistent, type-safe Pydantic validation and automated retries regardless of the provider.

To handle validation failures when extracting nested models from LLMs, the system automatically retries the extraction. It feeds the specific Pydantic validation errors back to the LLM, prompting it to correct and regenerate the structured output.