XMTP

XMTP provides end-to-end encrypted messaging, crucial for privacy, security, and compliance. Without it, messages are exposed to phishing, tampering, and unauthorized access.

Risks

Risks of not using end-to-end encryption for agent interactions:

Phishing: Messages can be intercepted and manipulated.
Tampering: Content can be altered without detection.
Unauthorized access: Sensitive information is vulnerable.

Backend

Install the plugin

bun install xmtp-agent

Create a client and relay responses through the XMTP network

import { createClient } from "xmtp-agent";
 
const xmtp = await createClient(onMessage, {
  apiKey: process.env.XMTP_PRIVATE_KEY,
});
 
const onMessage = async (message, user) => {
  console.log(`Decoded message: ${response} by ${user}`);
  //Your AI model response
  xmtp.send(response);
};

GPT example

onMessage(): Triggered each time a new message is received from XMTP.
xmtp.send(): Used to send messages (e.g., AI prompts and responses) back to the XMTP network.

In this example, when onMessage is invoked, it takes the incoming user message, passes it to the AI model (OpenAI's Chat Completion API), and then sends the AI's response back over XMTP.

import { createClient } from "xmtp-agent";
import { Configuration, OpenAIApi } from "openai";
 
const xmtp = await createClient(onMessage, {
  apiKey: process.env.XMTP_PRIVATE_KEY,
});
 
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
 
// This function is called whenever a new message arrives on XMTP.
// 'incomingMessage' represents the user's message received from the XMTP network.
const onMessage = async (incomingMessage, user) => {
  console.log(`Received: ${incomingMessage} by ${user}`);
 
  // Prepare conversation history or context as needed
  const history = [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: incomingMessage },
  ];
 
  // Call the OpenAI API to process the user's message
  const response = await openai.createChatCompletion({
    model: process.env.GPT_MODEL || "gpt-4",
    messages: history,
  });
 
  const decodedResponse = response.data.choices[0].message.content.trim();
 
  // Send the AI-generated response back to the XMTP network
  xmtp.send(decodedResponse);
};

Note: End users need to use a XMTP client to send messages to the agent.