# OpenAI Chat Completions

Trace a non-streaming Chat Completions request through the OpenAI client's fetch hook.



Datool supplies a fetch wrapper for **non-streaming Chat Completions**. It records each HTTP attempt, model input/output, status, and reported usage. Responses API calls and streaming requests pass through without this wrapper's model spans. Use [manual tracing](/docs/tracing/instrumentation) for those paths.

## Install [#install]

Use Node.js 22.18 or newer:

```sh
npm init -y
npm install @datool/sdk@0.3.1 openai @opentelemetry/api@1 @opentelemetry/sdk-trace-node@2
```

Configure the three `DATOOL_*` variables from [your first trace](/docs/get-started/first-trace) and an `OPENAI_API_KEY` in `.env`. The Datool key needs `traces:write`; the OpenAI key needs access to the selected model. The call uses your OpenAI account's quota.

## Make one traced request [#make-one-traced-request]

Save `openai.mjs`:

```js
import OpenAI from "openai"
import { trace } from "@opentelemetry/api"
import { NodeTracerProvider } from "@opentelemetry/sdk-trace-node"
import { DatoolSpanProcessor } from "@datool/sdk/otel"
import { traceOpenAIChatFetch } from "@datool/sdk/openai"

const processor = new DatoolSpanProcessor()
const provider = new NodeTracerProvider({ spanProcessors: [processor] })
provider.register()
const openai = new OpenAI({
  fetch: traceOpenAIChatFetch(trace.getTracer("docs-openai")),
})

try {
  const result = await openai.chat.completions.create({
    model: "gpt-4.1-mini",
    messages: [{ role: "user", content: "Say hello in one sentence." }],
  })
  console.log(result.choices[0]?.message.content)
} finally {
  try {
    await processor.forceFlush()
  } finally {
    await provider.shutdown()
  }
}
```

```sh
node --env-file=.env openai.mjs
```

Open the resulting trace and inspect `openai.chat.completions`. Check the request model, returned model, input messages, assistant response, and usage. Missing provider usage or price coverage remains unavailable. OpenAI client retries can produce multiple attempt spans.

Do not apply this wrapper to an [AI SDK](/docs/tracing/ai-sdk) client already emitting model spans. Configure Datool's processor only once per provider. Review [capture and redaction](/docs/tracing/instrumentation#delivery-and-sensitive-content) before recording production messages.