Using Collections and RAG in h2oGPTe to Query a Page
- Generative AI
- Enterprise h2oGPTe
This lesson covers Collections in Enterprise h2oGPTe, the place where a user uploads different sources of information and queries them by chatting with an LLM. It demonstrates importing a Wikipedia page by URL into a collection, showing how the tool converts the web page into a document and moves its text into a vector database. It then starts a chat using the RAG option, rather than the LLM alone, so the answer to a question about the page draws directly on the uploaded document instead of the model's general knowledge.
Key takeaways
- How a Collection in h2oGPTe lets a user upload many different sources of information to query in one place
- How importing a web page by URL turns it into a document and moves its text into a vector database
- Why choosing RAG instead of the LLM alone grounds an answer in the specific document that was uploaded
- How a question about an uploaded Wikipedia page returned a more detailed, document grounded answer using RAG
Start owning your AI