Agent complaints Error executing agent task: No LLM API key or LLM Client configured. An LLM API key or a custom LLM Client is required to use act, extract, or observe.

[guptaraghvendra]

Hello Team,

I just started to use stagehand to validate if it can be a good fit for my requirement at our organization and I keep running into an issue just with agent tasks.
We have self hosted open source models (qwen3-VL, gemma3 etc) for our use at our enterprise using APIs and I am using AISDK to configure it in my code and act, observe works fine but when it comes to agent tasks I get following error:
Error executing agent task: No LLM API key or LLM Client configured. An LLM API key or a custom LLM Client is required to use act, extract, or observe.

Following is a snippet of my code, What I understand is that for other tasks (extract/act/observe) my llmclient: custom client works fine but somehow for agent tasks it is not picking up my customClient. I might be wrong but please guide me to fix this issue:

import "dotenv/config";
import { Stagehand } from "@browserbasehq/stagehand";
import { createOpenAICompatible } from "@ai-sdk/openai-compatible"; 
import { AISdkClient } from "@browserbasehq/stagehand";


const customProvider = createOpenAICompatible({
  name: 'name',
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: 'https://base-url/gemma-27b/v1',
  includeUsage: true, 
  supportsStructuredOutputs: true
});


const customClient = new AISdkClient({
  model: customProvider('/app/models/gemma-3-27b-it'),
});

async function main() {
  const stagehand = new Stagehand({
    env: "LOCAL",
    llmClient: customClient,
  });

  await stagehand.init();

  console.log(`Stagehand Session Started`);


  const page = stagehand.context.pages()[0];

  await page.goto("https://stagehand.dev/");

  const extractResult = await stagehand.extract(
    "Extract the value proposition from the page."
  );
  console.log(`Extract result:\n`, extractResult);

  const actResult = await stagehand.act("Click the 'Evals' button.");
  console.log(`Act result:\n`, actResult);


  const observeResult = await stagehand.observe("What can I click on this page?");
  console.log(`Observe result:\n`, observeResult);

  const agent = stagehand.agent({
    systemPrompt: "You're a helpful assistant that can control a web browser.",
  });

  const agentResult = await agent.execute(
    "What is the most accurate model to use in Stagehand?"
  );
  console.log(`Agent result:\n`, agentResult);

  await stagehand.close();
}

main().catch((err) => {
  console.error(err);
  process.exit(1);
});