TypeScriptNext.js · Node.js · Edge

Monitor Vercel AI SDK with TraceHawk

Trace every generateText, streamText, and tool call in your Vercel AI SDK app. Works with Next.js, Node.js, and Edge runtime.

Install

npm install @tracehawk/sdk

Initialize

For Next.js App Router, add initialization in instrumentation.ts — runs once on server startup:

instrumentation.ts

// instrumentation.ts
export async function register() {
  const { init } = await import("@tracehawk/sdk");
  init({ apiKey: process.env.TRACEHAWK_API_KEY! });
}

For Node.js scripts, import and call at the top of your entry file:

import { init } from "@tracehawk/sdk";

init({ apiKey: "ao-..." });

Example — generateText with tools

agent.ts

import { init } from "@tracehawk/sdk";
init({ apiKey: process.env.TRACEHAWK_API_KEY! });

import { generateText } from "ai";
import { anthropic } from "@ai-sdk/anthropic";
import { tool } from "ai";
import { z } from "zod";

const result = await generateText({
  model: anthropic("claude-3-5-sonnet-20241022"),
  tools: {
    getWeather: tool({
      description: "Get weather for a location",
      parameters: z.object({ location: z.string() }),
      execute: async ({ location }) => {
        return { temperature: 22, condition: "sunny" };
      },
    }),
  },
  prompt: "What is the weather in Vienna?",
});

// Full trace: LLM call → tool decision → tool execution → final response
// All visible in TraceHawk with cost breakdown

Example — streamText in Next.js route

app/api/chat/route.ts

// app/api/chat/route.ts
import { init } from "@tracehawk/sdk";
init({ apiKey: process.env.TRACEHAWK_API_KEY! });

import { streamText } from "ai";
import { anthropic } from "@ai-sdk/anthropic";

export async function POST(req: Request) {
  const { messages } = await req.json();

  const result = streamText({
    model: anthropic("claude-3-5-sonnet-20241022"),
    messages,
    onFinish({ usage, finishReason }) {
      // TraceHawk automatically captures token usage here
    },
  });

  return result.toDataStreamResponse();
}

What you see in TraceHawk

✓Every generateText / streamText call as a span with latency
✓Streaming token usage and cost captured on finish
✓Tool call decisions with parameters and results
✓Per-user / per-session cost attribution (pass userId in headers)
✓Latency breakdown: model response time vs tool execution time

Per-user attribution

Pass a X-User-Id header from your frontend and TraceHawk will group costs and traces by user automatically — no extra SDK calls needed.

Ready to ship?

Free tier — 50K spans/month. No credit card required.

Get started free →