Agents

​

What are Agents?

Agents are intelligent assistants powered by Large Language Models (LLMs) that can:

• Collaborate together
• Use external tools & APIs to perform actions or fetch data
• See (Vission)
• Listen (Audio input)
• Talk (TTS with custom voices)

and do more crazy things that I still don’t know about :)

​

Creating Agents

You can visit the Scoopika platform to create new agents, The process takes mere seconds. You give your agent an avatar, name, and description; define the LLM powering it and give it a system prompt.

​

Prompting Your Agent

A system prompt is a set of instructions guiding the LLM’s interaction and behavior with users. Learn more about system prompts in this page.

​

Accessing Agent Information

Use the agent.load method to retrieve information about an agent (name, avatar, description, etc.) within your application:

const agent = new Agent("ID", scoopika);
await agent.load();

console.log(agent.agent);

Alternatively, use agent.info to obtain specific details:

const agent = new Agent("ID", scoopika);
const name = await agent.info("name"); // Type-safe

console.log(name); // name is string

​

Running agents

Once you have an agent instance, use the agent.run method to execute it. This method accepts an object containing inputs and hooks.

Explore these resources for further understanding:

• Agent Run.
• Inputs object.
• Streaming Hooks.

Here’s a basic example:

const agent = new Agent("ID", scoopika);
await agent.run({
	inputs: {
		message: "Hello"
	},
	hooks: {
		// Capture each token the agent outputs
		onToken: (token) => console.log(token)
	}
})

​

Speech

You can make your agents talk in custom voices now! learn more about Agents speaking. this is a simple example:

// English
await agent.speak("Good morning");

// French
await agent.speak("Bonjour", "fr");

.speak returns the URL to the audio generated from the text provided.

​

Using on client-side

To use agents on the client-side We recommend you check the Scoopika for the Web guide as It will walk you through how you can setup your server and client to run agents with streaming hooks on the client-side. It takes only a few minutes ;)