Set Up Enterprise RAG with h2oGPTe
- Generative AI
- Enterprise h2oGPTe
This lesson walks through getting started with Enterprise h2oGPTe's retrieval augmented generation platform, beginning with signing into the free trial at genai.h2o.ai. It shows how to create a data collection, choose an embedding model, and crawl a website, plus how to turn on private data detection to redact sensitive text before it reaches an external model. It then compares an answer from a large language model alone against one generated after adding the crawled data, showing how retrieval augmented generation grounds responses in referenced source pages.
Key takeaways
- How to create a data collection in Enterprise h2oGPTe and choose an embedding model for the source language
- What private data detection does, either blocking or redacting sensitive text before it reaches an external model
- How crawling a website populates a collection that h2oGPTe can then search for relevant chunks
- Why answers change when switching from a plain LLM response to retrieval augmented generation on your own data
- The reference and LLM prompt views show exactly which document chunks fed a given answer
Start owning your AI