Tool call errors have a different trace than tool call errors

[AbhiPrasad]

Trace, Issue

Repro:

import "./instrument";
import { generateText, streamText, tool } from "ai";
import { openai } from "@ai-sdk/openai";
import dotenv from "dotenv";
import { z } from "zod";
import * as Sentry from "@sentry/node";

dotenv.config();

async function main() {
  const result = await generateText({
    model: openai("o1", { structuredOutputs: true }),
    experimental_telemetry: {
      functionId: "Simple Agent",
      recordInputs: true,
      recordOutputs: true,
      isEnabled: true,
    },
    tools: {
      calculateTool: tool({
        description: `Calculate the result of a math problem. Returns a number.`,

        parameters: z.object({
          a: z.number().describe("First number"),
          b: z.number().describe("Second number"),
        }),
        type: "function",
        execute: async ({ a, b }) => {
          throw new Error("Not implemented");
          return a + b;
        },
      }),
    },
    maxSteps: 2,
    system: `You helps users with their math problems.`,
    prompt: "What is 1 + 1?",
  });
  console.log("result", result.text);
}

main();

[linear]

JS-758 Tool call errors have a different trace than tool call errors

[Divyanshu7001]

Hi @AbhiPrasad ,
I’m new to the Sentry environment and this repo. I can understand the issue conceptually, but to see the issue/trace in action, do I need to create an account or get access to https://telemetry-experience.sentry.io/?
I’m a bit lost and would appreciate any guidance on how to replicate or observe the problem in action so I can move forward.

[mydea]

Hey,

we have likely identified the problem, which happens if this tools are called without a wrapping root span. This should only be the case if there is e.g. not a http server that triggers calls. In this case, you'll need to manually wrap the call with a root span to ensure things are consistent. We are currently looking to verify this and add tests showing this behavior.