Skip to main content

Agent cooperation

We already mentioned signals on a couple of occasions, but let’s take a closer look at how communication on the platform actually works. This is important to understand if you want to build a more complex system, where multiple agents cooperate with each other.

Sending signals between agents

Apart from communicating with their users, we already mentioned the agents can communicate with each other. For example, your developer agent, Agent Smith, could be notified each time a new agent is created, to keep track of all the new users in the system. All we need to do is have Nola send a signal to Smith whenever a new agent of this type is created.

For this we just need to define a new signal. Signals are nothing else but Java classes that contain the data we want to communicate. Signals directed at a particular receiver are called Messages. So let's create a new directory and file in which we'll structure our message, java/signals/

package signals;

import com.mindsmiths.ruleEngine.model.Agent;
import com.mindsmiths.sdk.core.api.Message;

import lombok.*;

public class AgentCreated extends Message {
Agent agent;

Now, we’ll add some special behavior for the first message the user sends. To send the signal, we'll add a new rule to Conversation.drl:

import com.mindsmiths.ruleEngine.model.Initialize
import signals.AgentCreated

rule "First contact"
initialize: Initialize() from entry-point "signals"
agent: Nola()
agent.sendMessage("Okay, turning my engines on. Let's start!");
agent.send("SMITH", new AgentCreated(agent));

To pinpoint the very first contact the user makes with the platform, we make use of the Initialize() signal the platform sends when agents are created.

As mentioned, Messages are signals sent to a specific entity (agent or service). That's why you need to specify the id of the receiving agent ("SMITH"), along with the object you’re sending (the AgentCreated signal). All signals are by default always sent to the entry-point "signals".

Now all that's left is to process the message on Agent Smith’s end! Add a new file rules/smith/Agents.drl with the following content:

package rules.smith;

import com.mindsmiths.ruleEngine.util.Log

import signals.AgentCreated

rule "Register new agent"
signal: AgentCreated(agent: agent) from entry-point "signals"
then"New agent: " + agent);

Notice that in this simple case, the total number of active agents will correspond to the number of registered users (with the exception of Agent Smith, who doesn’t have a user). In more complex scenarios, this might not be the case.

That’s it! Write forge reset in the Terminal first to clear the database and start forge run to try it out!


If you want to learn more about signals, check out the documentation.

P.S. You can find the entire code on our GitHub →!


To spare you some time - here is Nola.drl file →, and here is file →