@tozil/openai

Tozil instrumentation for the OpenAI SDK.

Automatically tracks every chat.completions.create() call — model, tokens, latency — with zero code changes to your OpenAI usage.

Install

npm install @tozil/core @tozil/openai

Usage

import { init } from "@tozil/core";
import { openai } from "@tozil/openai";

init({
  apiKey: "tz_...",
  instrumentations: [openai()],
});

// Use OpenAI as normal — tracking is automatic
import OpenAI from "openai";
const client = new OpenAI();

const completion = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Hello" }],
});

How it works

The plugin monkey-patches chat.completions.create() on the OpenAI SDK prototype. It captures:

• Model name
• Input/output token counts (from usage in the response)
• Latency in milliseconds
• User and endpoint context (if set via @tozil/core)

It does not capture prompts, completions, or any content.

Streaming support

Streaming with stream: true is fully supported. Token usage is captured from the final chunk's usage field.

License

MIT