Configure Chats in h2oGPTe
- Generative AI
- Enterprise h2oGPTe
This lesson walks through the three tabs in h2oGPTe's New Chat interface, starting with the Documents tab for selecting a collection and uploading files. It covers the Configuration tab, including choosing a language model, enabling vision for image inputs, turning on the Use Agent toggle, which adds tools like chart generation and web search, and picking a generation mode of LLM only, RAG, or automatic. It also covers the Prompts tab, where a system prompt and template define the AI's role and tone. It shows how automatic LLM routing balances cost and performance, unless an agent is running, in which case accuracy takes priority.
Key takeaways
- What the Documents tab does, letting you select a collection and check which files are indexed
- How turning on the Use Agent toggle adds tools like chart generation, document conversion, and web search
- Why choosing RAG instead of LLM Only changes whether responses draw on your uploaded documents
- The automatic LLM routing mode balances cost and performance unless an agent is active, then favors accuracy
- How the Prompts tab lets you set a system prompt and template to shape the AI's tone and role
Start owning your AI