w3gpt

TypeScript client for the Web3GPT skill endpoint.

The SDK is intentionally simple:

• call with no params to create a new chat
• keep the returned chatId
• pass that chatId on later calls to continue the same thread
• set history=true or full=true if you want the full simplified history back

Installation

npm install w3gpt
# or
pnpm add w3gpt
# or
bun add w3gpt

Usage

import { w3gpt } from "w3gpt";

const client = w3gpt();

const started = await client.startChat();
console.log(started.chatId);

const reply = await client.chat({
  chatId: started.chatId,
  message: "Deploy an ERC20 on Polygon mainnet with 1,000,000 supply",
});

console.log(reply.response);

Read History

const result = await client.chat({
  chatId: "your-chat-id",
  history: true,
});

console.log(result.history);

API

w3gpt(config?)

const client = w3gpt({
  baseUrl?: string; // defaults to https://w3gpt.ai
  fetch?: typeof fetch; // optional custom fetch implementation
});

client.startChat(agentId?)

Create a new chat and receive a chatId.

client.chat(params?)

const result = await client.chat({
  agentId?: string;
  chatId?: string;
  message?: string;
  history?: boolean; // defaults to false
  full?: boolean; // alias for history=true
});

Response:

type W3GPTChatResponse = {
  agentId: string;
  chatId: string;
  response: string | null;
  history?: Array<{
    id: string;
    role: "user" | "assistant" | "system";
    text: string;
  }>;
};

Notes

• chatId is the secret for continuing the thread.
• If you call client.chat() with no params, a new chat is created.
• Deployments happen through the agent conversation itself.
• Polygon mainnet deployment is supported through the skill endpoint.