Did your OpenAI Agents SDK run actually help?

A run that completed is not a run that helped. This captures what the user thought at the point the work is genuinely done, and binds it to the run group you already use to stitch a conversation together.

Start free Read the docs

Install

One package. It sits beside your existing tracing rather than replacing any of it.

npm install @uservane/openai-agents @openai/agents
# peer: react >= 18 for the client UI

Wiring

Capture at true resolution, on the server

import { runWithFeedback } from "@uservane/openai-agents/server";
import { Agent, Runner } from "@openai/agents";

const groupId = sessionId; // your chat session id

const { finalOutput, feedback } = await runWithFeedback(agent, input, {
  secretKey: process.env.USERVANE_SECRET_KEY!, // uv_sk_... server only
  respondentId: userId,
  groupId,
});

What binds to what. The token is minted on your server against the group id you already use for the session, so the rating arrives attached to that run rather than floating free.

Limits

What it does not do

Worth reading before you wire it in, because both of these are things people reasonably assume and neither is true.

The secret key never reaches the browser: Minting happens on your server. A publishable key is what the client holds, and it cannot mint a token bound to somebody else's session.
It does not grade the model: A model scoring its own kind is a proxy. This is the human answer, which is the one your leadership is actually asking for.

The number

The number you get back

Every score ships with its margin of error and the sample it came from, and a sample too thin to report is suppressed rather than printed. An agent satisfaction number off twelve responses is not a metric, and a tool that shows it to you anyway is not doing you a favour.

See the interval around a score How the capture works

Free to start, no card. The SDKs are MIT and open source at github.com/uservane/uservane-js.

Start free @uservane/openai-agents on npm