instructor

You extract structured data from LLM responses by passing a Pydantic model to define the expected schema, which validates the LLM output for type safety. The process automatically retries if the LLM fails to generate valid data matching your schema.

Validating LLM outputs is best handled using Pydantic schemas to enforce type safety on the generated text. The system automatically detects validation failures and triggers retries to ensure the final output strictly matches the defined model.

Yes, you can process partial results as they are generated by the LLM. This streaming capability allows you to incrementally populate and interact with a Pydantic model while the text generation is still in progress.

Yes, structured data extraction using Pydantic validation works with both OpenAI and Anthropic models. The implementation relies on their respective dependencies to ensure type-safe data parsing and automatic retries across different LLM providers.

You need structured LLM outputs when performing complex data parsing, classification, or information retrieval tasks that require reliable formats. Using Pydantic validation ensures the extracted data is type-safe and automatically corrected upon failure.

LLM data extraction often fails due to malformed or unstructured text outputs. Pydantic validation prevents this by enforcing a strict schema for type safety, automatically triggering retries to correct any outputs that fail validation.