Collect and inspect feedback

Updated 4 hours ago • October 18, 2026

Your application can send ratings for an assistant response, such as a thumbs-up or thumbs-down. Studio associates the rating with the response's run.

Collect response feedback

Kortyx assistant messages expose runId through @kortyx/react. Use the run reference to connect a response rating to Studio. Send the rating to your own authenticated backend first: derive the actor from the signed-in user and verify that the response/run belongs to that user. Do not accept an arbitrary browser-supplied run ID without an ownership check or signed feedback token.

After those checks, forward a server-to-server request:

const response = await fetch(
  `${process.env.KORTYX_TELEMETRY_API_URL}/v1/telemetry/scores`,
  {
    method: "POST",
    headers: {
      authorization: `Bearer ${process.env.KORTYX_TELEMETRY_API_KEY}`,
      "content-type": "application/json",
    },
    body: JSON.stringify({
      runId: verifiedResponse.runId,
      actorId: authenticatedUser.id,
      value: liked ? 1 : 0,
      reasons: selectedReasons,
      comment: optionalComment,
    }),
  },
);
if (!response.ok) throw new Error("Feedback could not be saved.");

verifiedResponse, authenticatedUser, liked, selectedReasons, and optionalComment represent your application-owned validated values. Reasons are optional: incorrect, irrelevant, incomplete, unsafe, or other. Comments are optional and capped at 4000 characters. Do not assign a negative reason automatically if the user has not selected one.

One vote is stored per actor/run. Changing the vote updates its existing score; send DELETE to the same endpoint with { runId, actorId } to clear it. The run must already exist in Studio, so flush/retry pending telemetry before treating a 404 as permanent. No credentials are exposed to the chat browser.

Studio surfaces ratings in Runs → Feedback, its feedback filter, and session activity. A negative rating does not mark execution as failed. Human correctness reviews appear separately and require a Studio key with studio:write in addition to studio:read. Shared-key self-hosted reviews use the Studio key as reviewer identity. These native Studio scores are independent of the optional Langfuse integration; forwarding to Langfuse remains application-owned.

Good to know: A negative user rating can belong to a successfully executed run. Run status describes execution; evaluation grades assess defined test requirements.

Inspect a rating

Open the run's feedback view to read recorded ratings and their available context. A negative rating describes the user's experience. It does not change a successful execution into a failed run or replace an evaluation grade.

Next: Inspect runs.