google/langextract

Supports multiple LLM providers (Google Gemini, OpenAI, local Ollama models) with pluggable custom providers, and uses optimized chunking and parallel processing to handle long documents efficiently. Enforces schema-compliant outputs through controlled generation on supported models, while automatically detecting hallucinated extractions that don't ground to source text. Generates interactive HTML visualizations that map each extracted entity back to its precise character position in the original document for verification and review.

34,668 stars and 173,955 monthly downloads. Actively maintained with 8 commits in the last 30 days. Available on PyPI.