@uservane/vercel-ai/server: MintFeedbackTokenResult

Type Alias: MintFeedbackTokenResult

MintFeedbackTokenResult = object

Defined in: server.d.ts:64

@uservane/vercel-ai/server - server-only helpers for agent-native capture.

Re-exports mintFeedbackToken from @uservane/agent-core/server.

SECURITY: This module MUST run only on the customer’s server (Next.js route handler / server action / edge function). It accepts the project SECRET key and POSTs to /v1/sdk/tokens. Never import this entry from client components or ship the secret key to the browser.

Correlation model (verified, 2026-07-27):

• Langfuse + Vercel AI SDK: the developer controls sessionId via propagateAttributes({ sessionId }, () => generateText(…)). The trace id is auto-generated; we do NOT auto-read it.
• Langfuse accepts session-level scores (sessionId alone).
• Therefore: mint a token bound to the same sessionId the server already set for Langfuse, then thread the showToken + sessionId to the client.

Example

// app/api/chat/route.ts (server)
import { mintFeedbackToken } from "@uservane/vercel-ai/server";
import { propagateAttributes } from "@langfuse/tracing";

const sessionId = crypto.randomUUID();
await propagateAttributes({ sessionId }, async () => {
  const result = await generateText({ ... });
  const { tokens } = await mintFeedbackToken({
    secretKey: process.env.USERVANE_SECRET_KEY!,
    respondentId: userId,
    sessionId,
    surveyId: "surv_post_task",
  });
  // stream tokens[surveyId] + sessionId to the client; client calls bind()
});

Properties

issuedAt

issuedAt: string

Defined in: server.d.ts:69

---

suppression

suppression: string[]

Defined in: server.d.ts:68

Survey ids already answered/dismissed for this respondent.

---

tokens

tokens: Record<string, string>

Defined in: server.d.ts:66

surveyId -> v2 show-token bound to sessionId. Empty when gated.