Skip to main content
You use the Vercel AI SDK for streaming generations. Track each generation as an LLM call inside an Apie run. 
import { withVercelAiGeneration } from "@apie-sh/sdk";

await apie.withRun({ inputSummary: "Generate response" }, async (run) => {
  const result = await withVercelAiGeneration(
    apie,
    {
      runId: run.id,
      model: { provider: "openai", name: "gpt-4o" },
      inputSummary: "Summarize incident feed",
    },
    async () => generateText({ model: openai("gpt-4o"), prompt: "Summarize..." }),
  );
});
Events: agent.llm.called, agent.llm.completed, agent.llm.failed

​
What you’ll see

LLM generation events in the run timeline with model provider, name, and input summary.
The Vercel AI SDK is primarily a JavaScript ecosystem. The Python wrapper exists for API parity but is most relevant in JS/TS projects.

​
Next steps

Custom telemetry

LLM calls and workflow steps.

Integrations hub

All integrations.