Create assistant message

The Create assistant message endpoint streams an Ask AI reply for a published docs site. There is no API key. The site is selected by {domain} — the hostname readers visit.

For Velu's own docs at https://veludocs.com/docs, {domain} is veludocs.com. Pass /docs (or a page path like /docs/quickstart) in currentPath. Each reply consumes 1 credit from the site owner's plan.

Integration with useChat

The useChat hook from Vercel's AI SDK is the recommended way to embed Ask AI in your product.

Install AI SDK

npm i ai @ai-sdk/react

Use the hook

import { useState } from "react";
import { useChat } from "@ai-sdk/react";
import { DefaultChatTransport } from "ai";

function DocsChat({ domain }: { domain: string }) {
  const [input, setInput] = useState("");

  const { messages, sendMessage } = useChat({
    transport: new DefaultChatTransport({
      api: `https://api.getvelu.com/v1/assistant/${domain}/message`,
      body: {
        fp: "anonymous",
        retrievalPageSize: 5,
        currentPath: "/docs",
      },
    }),
  });

  return (
    <div>
      {messages.map((message) => (
        <div key={message.id}>
          {message.role === "user" ? "User: " : "Assistant: "}
          {message.parts
            .filter((part) => part.type === "text")
            .map((part) => part.text)
            .join("")}
        </div>
      ))}
      <form
        onSubmit={(e) => {
          e.preventDefault();
          if (input.trim()) {
            sendMessage({ text: input });
            setInput("");
          }
        }}
      >
        <input value={input} onChange={(e) => setInput(e.target.value)} />
        <button type="submit">Send</button>
      </form>
    </div>
  );
}

Required configuration:

• transport — Use DefaultChatTransport to configure the API connection.
• body.fp — Fingerprint identifier (anonymous or a unique user id).
• body.retrievalPageSize — Number of search results to ground the reply (recommended: 5).

Optional configuration:

• body.threadId — Continue a thread. The stream finish event includes threadId.
• body.currentPath — Path the reader is viewing. Maximum 200 characters.
• body.context — Snippets from the page. Each object has:
  • type — code or textSelection
  • value — The snippet
  • path (optional) — Source file or page path
  • elementId (optional) — UI element id
• body.filter.language — Language code filter, for example en.

See useChat and Transport in the AI SDK documentation.

Rate limits

• 10 requests per reader (fp) per minute
• 200 requests per docs site per minute