The new extraction service is a bare-bones web application built on top of FastAPI and Postgresql, which exposes a standard REST API for defining and persisting "extractors" that contain the key components needed to configure an LLM for extraction. These components include a schema that describes the structure of the data to be extracted, a prompt that provides context for the extraction and primes the model, and reference examples that show the model some examples of extractions. The service allows users to submit files for extraction using an existing extractor and supports parsing of PDFs and HTML files, with the option to extend support to other types of files using existing LangChain document loaders and parsers. The API is designed to always extract multiple entities, which is usually the desired behavior when extracting from text. Users can specify examples that will be incorporated to improve the quality of extracted results, and save and share extractors in a database for easy reuse.