Building Reusable Extractors in Enterprise h2oGPTe
- Generative AI
- Enterprise h2oGPTe
This lesson covers the full Extractors workflow in Enterprise h2oGPTe, moving beyond a single document extraction to a reusable component you can apply across any collection. It explains that an Extractor pairs a JSON schema, which defines the fields you want pulled out, with an LLM that interprets each document and maps its content to that structure. The lesson walks through creating a new Extractor, defining its schema using either the visual builder or the code editor, and running it against a target collection to produce structured JSON output.
Key takeaways
- How an Extractor combines a JSON schema and an LLM to turn unstructured documents into structured, reusable output
- The two ways to define an Extractor's schema: the visual builder for adding fields and tables, or pasting a schema in code mode
- Why schema field labels do not need to match a document's exact wording, since the LLM maps intent to the right content
- How to organize Extractors across a team using the All Extractors, My Extractors, and Shared views
- The steps to run an Extractor on a collection: open its actions menu, choose Run Extractor, select the collection, and run
Start owning your AI