Skip to main content
Velor

Enterprise IT guides

Going live checklist

The order to do things in, what the Go-live readiness page checks and cannot check, and the install steps for a plain HTML page and for React or Next.js.

Your engineer can do all of this without contacting us. The product checks part of it for you on /dashboard/go-live (Go-live readiness); this page is the rest, in the order that avoids rework.

1. Before any customer sees the widget

  1. Teach the agent. Add at least three knowledge sources under /dashboard/knowledge; the readiness page counts three as the minimum for the knowledge check to pass. Three sources is a floor, not a quality bar: ask the questions your customers ask in /dashboard/playground and read the answers. The readiness check "You have tried the agent" links to /onboarding?step=4 until you have.
  2. Put a human on the receiving end. The agent hands a conversation to a person; with nobody invited, or nobody online, the customer waits. Invite a teammate under /dashboard/team/invite, and have someone keep the dashboard open during your support hours. Set your hours under /dashboard/settings/business-hours.
  3. Set the visitor consent notice under /dashboard/settings/privacy. The readiness page checks the setting; it cannot check that the text is adequate for your jurisdiction. That is your counsel's call, not ours.

2. Install the widget

Your workspace slug is the Slug field under Settings → General. The widget is served from https://app.velorplatforms.com/widget.js; there is no other host.

Plain HTML, before </body>:

<script src="https://app.velorplatforms.com/widget.js" async></script>
<script>
  window.Velor = window.Velor || { q: [] };
  Velor.q.push(["init", { orgSlug: "your-org-slug" }]);
</script>

Strict Content-Security-Policy (no 'unsafe-inline'): the inline stub above is refused by such a policy. Use the one-tag form from /docs, which reads data-org-slug off its own tag, and allow https://app.velorplatforms.com in script-src and connect-src.

React or Next.js (a client component mounted once, for example in your root layout):

"use client";
import { useEffect } from "react";

export function VelorWidget() {
  useEffect(() => {
    const script = document.createElement("script");
    script.src = "https://app.velorplatforms.com/widget.js";
    script.async = true;
    (window as any).Velor = (window as any).Velor || { q: [] };
    (window as any).Velor.q.push(["init", { orgSlug: "your-org-slug" }]);
    document.body.appendChild(script);
    return () => {
      script.remove();
      (window as any).Velor?.destroy?.();
    };
  }, []);
  return null;
}

React's development Strict Mode runs the effect, its cleanup and the effect again; the cleanup removes the widget before it is mounted again, so you end up with one launcher, not two. Do not render the component on every route: mount it once in the layout so the conversation survives navigation.

Then open your site and start a chat. The readiness page reads Ready for the widget only after a real chat session has arrived from your site; it will say no visitor has opened the widget until then.

3. Lock it to your sites

The list of allowed origins under /dashboard/settings/widget starts empty, which accepts any website so installs work immediately. Before launch, set it to your production hostnames (and staging, if you test there). A custom domain verified under /dashboard/settings/custom-domain is added to that list; it does not host anything on your hostname.

4. Identify signed-in visitors (optional, recommended)

If your customers are signed in, tell the widget who they are with Velor.identify, and turn on identity verification under /dashboard/settings/widget. The widget settings page shows a signing snippet with your secret in it; sign the user id on your server, never in the browser, because anyone who has the secret can impersonate any of your users. With identity verification on, a visitor whose signature is missing or wrong cannot start a chat: the session request is refused (403) and the widget shows an error. If you turn it on before generating a secret, every signed-in visitor is refused. Test a signed-in visitor end to end before you rely on it.

5. API, webhooks and the other channels

  • API: create a key under /dashboard/settings/api-keys, grant only the scopes the integration needs, and call the endpoints listed at /developers/api. The key is shown once.
  • Webhooks: verify the signature on every delivery with the samples at /docs/webhooks, and reply 2xx quickly. Verify against the raw request body, not re-serialised JSON.
  • Voice, SMS, WhatsApp, Slack: each has its own step list under Settings → Channels and at /docs/getting-started. These run on your own provider accounts, and their approval times (US text-message registration, Meta business verification, number porting) belong to those providers; we do not state a time and cannot shorten it. Email intake is set up with us and is not self-serve; there is no committed date.
  • SSO and SCIM: optional, owner-only, covered in the SSO and SCIM guides. SSO is verified by automated tests and has not yet been exercised against a live identity provider in production.

6. What the readiness page cannot see

It does not know whether a test chat was answered from your sources (the playground shows the sources but does not record them), whether your site shows a privacy link near the widget, or whether voice, SMS and WhatsApp numbers are working (those are checked at setup). A green page means the things it can check are in place, not that the agent is good. Run your own test conversations, including a request for a person.

7. After launch

Read the first conversations yourself in the first few days. See Incidents and status for how we tell you about a problem and what you should do on your side.

In the product: /dashboard/go-live (sign-in required).