Hands-On RAG with Enterprise h2oGPTe
- Generative AI
- Enterprise h2oGPTe
This lesson is Lab One of a hands-on RAG module using Enterprise h2oGPTe, following the theory covered in the prior lecture. The presenter creates a new collection, uploads the h2oGPT research paper, and shows how document ingestion works before querying the collection through the API using the h2oGPTe Python package. He runs example prompts asking for a summary, a LinkedIn post draft, and bullet points on the paper's impact, noting the answers cite the correct source document rather than hallucinating links. The lesson ends with homework to repeat the exercise on a new document.
Key takeaways
- How to create a new collection in Enterprise h2oGPTe and upload a document for retrieval augmented generation
- How to generate an API key in Enterprise h2oGPTe and query a collection programmatically with the h2oGPTe Python package
- Why RAG answers cite the correct source link instead of hallucinating one, since responses are grounded in the uploaded document
- What the homework assignment covers: uploading a LinkedIn profile PDF and refining queries with an added system prompt
- How document preprocessing works in Enterprise h2oGPTe, taking about 17 seconds to parse and store a single PDF
Start owning your AI