instructor-ai/instructor
Instructor
Instructor is a schema enforcement and validation library designed to transform language model outputs into structured, type-safe data formats. It functions as a validation layer that uses Pydantic to ensure model responses conform to specific data models, acting as a tool for forcing large language models to return data in predefined schemas.
The project differentiates itself through a recursive error-feedback loop that automatically retries requests when structural errors occur, passing validation failure messages back to the model to guide corrections. It also includes a streaming parser capable of processing partial fragments of structured objects in real time as they are generated.
The library covers broad capabilities for structured data extraction, including the parsing of complex hierarchical information and nested structures into machine-readable formats. It utilizes prompt injection to translate type definitions into schema instructions and provides a type-safe wrapper interface to map raw responses directly into typed objects.
Features
- Structured Output Enforcements - Forces language model outputs to adhere to strict data models for reliable programmatic processing.
- Error-Correction Feedback Loops - Sends structural validation failures back to the model to guide the correction of malformed data.
- Model Output Formatting - Ensures model responses conform to strict schemas for seamless integration with other software.
- Schema Enforcement Layers - Utilizes a validation layer to force model outputs to adhere to specific JSON schemas.
- Validation-Based Retries - Automatically retries requests when model outputs fail to conform to the required structural schema.
- Schema Enforcement Tools - Provides a system for validating model outputs and automatically retrying requests on structural errors.
- Structured Data Extraction - Provides a framework for extracting structured data from language model outputs using predefined schemas.
- Structured Output Converters - Forces large language models to return data in predefined schemas and type-safe formats.
- Model Response Validation Wrappers - Uses Pydantic as a validation layer to ensure LLM responses conform to specific Python data models.
- Incremental JSON Parsing - Parses incomplete JSON fragments from a stream into structured objects in real time.
- Data Serialization and Parsing - Transforms nested and hierarchical information from language models into structured lists and objects.
- LLM JSON Constraints - Injects JSON schema instructions into prompts to constrain the model's output format.
- Structured Collection Streams - Processes fragments of structured objects incrementally as they are generated by the model.
- JSON - Provides a parser that processes partial JSON fragments in real time as they are generated by a language model.
- Structured Object Streaming - Enables real-time processing of structured objects by parsing partial fragments as they are generated.
- Type-Safe Response Parsers - Wraps client calls to map raw model responses directly into type-safe objects.
- Structured Output Generation - Library for extracting structured data using Pydantic schemas.
- Data Pipelines - Extracts structured data from LLMs for downstream processing.
