LangExtract
Google's open-source Python library that uses LLMs to extract structured information from unstructured text, maps each extraction back to its exact location in the source and renders the results as an interactive HTML review page.
About LangExtract
Pulling entities, attributes and relationships out of long reports, clinical notes or literature with an LLM gives answers that are hard to check, because the output does not show where in the source each value came from.
Who it’s for
- Python developers turning long documents into structured records with Gemini, OpenAI or local Ollama models
- Analysts and researchers who need to verify each extracted value against the original text
When to consider it
Consider LangExtract when you want to define an extraction task with a prompt and a few examples, run it over long documents with chunking, parallel calls and multiple passes, and review every result highlighted in its original context.
Tradeoffs & limitations
- Apache-2.0 licensed and free to install from PyPI, but cloud models such as Gemini or OpenAI need your own API key and are billed by the provider; local models run through Ollama. The README states it is not an officially supported Google product, and health-related use is also subject to Google's Health AI Developer Foundations Terms of Use.
SOTA overview · Documentation-based assessment · Sources & review method
Updates
No updates shared yet.
Discussion
Newest firstAsk a question or share how you use LangExtract.
Keep it helpful. Community rules
Loading discussion…
Sources & review method
Documentation-based assessment · Nov 2, 2026 · Prepared with AI assistance; not a hands-on benchmark.
Checked by SOTA · AI-assisted documentation review. Selection advice is our assessment; verify current requirements for your deployment.
Import history & original evidence
LangExtract on GitHub
Google's open-source Python library that uses LLMs to extract structured information from unstructured text, maps each extraction back to its exact location in the source and renders the results as an interactive HTML review page.
Based on official pages and announcements checked on 2026-11-02. No hands-on test, performance benchmark or popularity ranking is claimed.