## Reasoning

For reasoning models, set `reasoning_effort` on a chat request to `"none"`, `"low"`, `"medium"`, or `"high"`.

```json
{
  "model": "deepseek/deepseek-v4-pro",
  "messages": [{ "role": "user", "content": "Plan a three-step migration." }],
  "reasoning_effort": "high"
}
```

The model's thinking comes back separately from its answer, as `reasoning_content`:

- Non-streaming: `choices[0].message.reasoning_content`
- Streaming: `delta.reasoning_content` on each chunk

Models that do not reason ignore the field.

## Tracing

Every response carries an `X-Request-Id` header. To follow one call across your own services, send your own `X-Request-Id` on the request and the gateway keeps it.

```bash
curl https://api.lune.io/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "X-Request-Id: order-8841-step-2" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek/deepseek-v4-pro","messages":[{"role":"user","content":"Hi"}]}'
```

The same id appears in your request logs in the console. Open a row on the [Usage](/docs/usage-and-analytics) page to see it. Include it when you contact support.