Agent Building
Initial Prompt
Every ATLAS agent requires a carefully structured system prompt. Here's the recommended format:
const systemPrompt = {
identity: {
name: "ATLAS Agent",
version: "1.0",
foundation: ["ELIZA", "Zerepy"]
},
capabilities: [
"knowledge_synthesis",
"pattern_recognition",
"consensus_validation"
],
behavior: {
communication_style:
"technical_yet_accessible",
response_format: "structured_insights"
}
};
Basic Setup
import { ATLAS } from '@atlas-ai/core';
const node = ATLAS.new()
.with_consensus("proof_of_stake")
.with_storage("distributed")
.build();
await node.start();
Example Implementation
Here's a basic example of how to use our package in your project:
import { openai } from "@ai-sdk/openai";
import { streamText } from "ai";
export async function POST(req: Request) {
const { messages } = await req.json();
const result = streamText({
model: openai("gpt-4"),
system: "You are ATLAS, a decentralized
superintelligence
network built on
ELIZA technology and Zerepy framework.
Your purpose is to synthesize and
distribute knowledge across the agent
network while maintaining security
through proof-of-stake consensus...",
messages,
});
return result.toDataStreamResponse();
}
For more detailed information, please check out our API Reference sections.