Nodes
Learn what each node does, when to use it, and jump into focused guides.
When to use which node
- User Utterance Node — Wait for user input and pass it to your flow.
- Model Node — Call an LLM (with system prompt, tools, and parsers).
- Reply Message Node — Send a response (plain text or via Jinja).
- Custom Node — Run your own Python and return JSON.
- Custom Tool Node — Define a function the LLM can call during reasoning.
- REST API Node — Call external services (optionally expose as a tool).
- Human Agent Node — Escalate to a human via a configured channel.
- Knowledge Base Node (EA) — Connect to a knowledge base (experimental).
- Sub Graph — Reuse another flow inside the current one.
Quick reference
| Node | Use for | Link |
|---|---|---|
| User Utterance | Pause and capture user input | User Utterance |
| Model | Call an LLM; prompts, tools, parsers | Model |
| Reply Message | Return output to the user; Jinja pull | Reply Message |
| Custom Node | Custom Python that returns JSON | Custom Node |
| Custom Tool | LLM-callable function (arguments/returns) | Custom Tool |
| REST API | External API calls; optional tool exposure | REST Api |
| Human Agent | Route to a human channel | Human Agent |
| Knowledge Base (EA) | Retrieval from KB (in progress) | Knowledge Base (Early Access) |
| Sub Graph | Reuse an existing flow | Sub Graph |
Tip
If you're unsure, start with User Utterance → Model → Reply Message, then add tools or APIs as needed.