langextract

LangExtract is a Python library by Google that leverages Large Language Models (LLMs) to extract structured information from unstructured text based on user-defined instructions. It processes documents like clinical notes or reports, identifying and organizing key details while precisely grounding extracted data back to the source text. Optimized for long documents through chunking, parallel processing, and multi-pass extraction, it ensures reliable structured outputs via schema enforcement and few-shot examples. It supports flexible LLM integration (cloud and local via Ollama) and provides interactive visualization for reviewing results.

• Precise Source Grounding with Character Spans
• Reliable Structured Output and Schema Enforcement
• Optimized Extraction for Long Documents
• Interactive Visualization of Extracted Entities
• Flexible Multi-LLM Support (Cloud & Local)