Node-RED Nodes
Smartunit provides a small set of Node-RED nodes for building agent flows.
| Type | Kind | Purpose |
|---|---|---|
agent | Flow node | Runs the agent and produces responses. |
agent-in | Flow node | Converts ordinary messages, tool results, or upstream agent responses into agent requests. |
agent-skill | Flow node | Exposes selected Node-RED flow actions as callable skills. |
agent-model | Config node | Stores model names and temperature. |
agent-connection | Config node | Stores provider connection details. |
agent-behaviour | Config node | Stores instructions, iteration limit, and debug mode. |
agent-memory | Config node | Connects facts and events storage. |
agent-budget | Config node | Sets optional token or spend limits. |
The agent node receives agent requests and produces responses or operational events.
It requires:
- connection
- behaviour
- model
- memory
- optional budget
- context mode
- optional skill set
Context mode controls how much stored history is sent to the model: full, summary,
lastn, or none.
Agent In
Section titled “Agent In”agent-in is the recommended entry point before the agent node.
| Mode | Use |
|---|---|
Message | Wrap a user message as an agent request. |
Tool Response | Return skill output to the agent. |
Agent Response | Chain one agent’s response into another agent’s request. |
In message mode, the node resolves session ID, message text, and optional user ID from configured Node-RED fields.
Agent Skill
Section titled “Agent Skill”agent-skill defines one or more actions. Each action has:
- name
- description
- parameter list
- output mapping
Skill names and descriptions should be clear enough for the agent to choose the correct action. Do not include secrets or private implementation notes in skill descriptions.
Config Nodes
Section titled “Config Nodes”agent-connection stores provider connection details. Never copy real keys into docs,
examples, screenshots, or support messages.
agent-model stores the primary model, optional secondary model, and temperature.
agent-behaviour stores system instructions, max iterations, and debug mode. Debug mode
can record sensitive request and response details, so use it carefully.
agent-memory stores facts/events storage settings and a scope of session or user.
agent-budget can limit tokens or spend over session, daily, weekly,
monthly, or lifetime periods.