H2O GPT Overview: Open Source Chat and Document Q&A
- Generative AI
This lesson walks through H2O's open source h2oGPT tool and the large language model concepts behind it. It covers what GPTs and LLMs are in plain terms, then demonstrates two versions of the chat interface, including simple conversational prompts and a data source tab for uploading a document, adding a URL, or pasting text so the model can answer questions grounded in that content. The lesson also covers the tradeoffs between hosted commercial LLMs and open source models that can run on premises for data privacy.
Key takeaways
- How to upload a PDF, add a URL, or paste text into h2oGPT so it answers questions from that specific source
- Why open source models like h2oGPT can run on premises to keep sensitive data off external servers
- The difference between a general chat answer and one grounded in an uploaded document or web page
- What to try when a summary comes back too generic: ask a more specific question or regenerate the answer
Start owning your AI