node-red-contrib-factory-agent-states

A Node-RED node for factory agent systems that collects and manages state information from multiple MQTT topics.

Overview

The factory-agent-states node is designed to work within factory automation systems using agents. It provides a way to:

• Cache incoming messages from different MQTT topics
• Store system prompts and environment descriptions for agents
• Output collected states when triggered by a specific signal
• Clear the cache after output for the next cycle

Installation

You can install this node directly from the Node-RED palette manager, or via npm:

npm install node-red-contrib-factory-agent-states

Features

• Caching system: Stores the latest message from each topic until triggered
• Configurable fields: Allows setting System Prompt and Environment Description
• Initial delay: Automatically sends the first state after a configurable delay upon receiving the first message
• Trigger mechanism: After the first automatic trigger, constantly monitors the flow context variable agentstate and outputs cached states when it equals "recieved"
• Automatic cache clearing: Clears the topic cache after sending out the state

Node Configuration

The node provides a simple configuration interface with the following options:

• Name: Optional name for the node instance
• System Prompt: Text to be stored in msg.sysPrompt
• Environment Description: Text to be stored in msg.envPrompt

Input

The node accepts messages with payloads that will be cached based on their topics.

Flow Context Variables

The node monitors a flow context variable:

• agentstate: When this variable is set to "recieved", the node will output all cached states and then reset the variable

Output

When triggered, the node will output a message containing:

• msg.sysPrompt: The System Prompt configured in the node
• msg.envPrompt: The Environment Description configured in the node
• msg.state: An object containing all cached topic payloads, with topics as keys

Usage Example

1. Add MQTT input nodes for various factory sensors/data sources
2. Connect these to the factory-agent-states node
3. Configure the node with appropriate prompts and descriptions
4. Set up a trigger mechanism (e.g., a function node that sets the flow context variable: flow.set("agentstate", "recieved"))
5. Connect the output to your agent processing node

Example Flow

[
    {
        "id": "mqtt1",
        "type": "mqtt in",
        "topic": "factory/temperature",
        "wires": [["factory-states"]]
    },
    {
        "id": "mqtt2",
        "type": "mqtt in",
        "topic": "factory/pressure",
        "wires": [["factory-states"]]
    },
    {
        "id": "factory-states",
        "type": "factory-agent-states",
        "name": "Factory States",
        "systemPrompt": "You are a factory optimization agent",
        "environmentDescription": "Factory with temperature and pressure sensors",
        "wires": [["agent-processor"]]
    },
    {
        "id": "trigger",
        "type": "function",
        "payload": "",
        "payloadType": "str",
        "repeat": "300",
        "func": "flow.set('agentstate', 'recieved');\nreturn msg;",
        "wires": [["factory-states"]]
    }
]

License

MIT