LangChain Extract is an open-source reference application designed to demonstrate how large language models can be used to extract structured data from unstructured text and document files. The project implements a lightweight web service that allows developers to define extraction schemas and apply them to various sources such as plain text, HTML, or PDF documents. Built using FastAPI and the LangChain framework, the application exposes a REST API that can process documents and return structured outputs that match user-defined JSON schemas. Developers can create reusable “extractors” that define what type of information should be pulled from a document, along with example prompts that improve extraction quality through in-context learning.
Features
- FastAPI web server exposing a REST API for LLM-based extraction
- Schema-driven extraction using JSON schema definitions
- Support for extracting structured data from text and document files
- Few-shot examples to improve extraction accuracy
- Database storage for reusable extractors and configurations
- LangServe endpoint integration for LangChain pipelines