> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openserv.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Compare Raw mode with SERV

> Run the same request with and without SERV Reasoning.

Use the Playground for a visual comparison. To compare in code, send one normal request and repeat it with the `x-openserv-disable-braid: true` header.

The header bypasses SERV processing for that request and bills it at the upstream tier. SERV prompt generation, Prompt Guard, the content filter, Shadow Agent, Kronos, and Multipath do not run. Raw mode is useful for a controlled comparison or fallback path.

```js theme={null}
const messages = [
  { role: "system", content: "You are a careful refund-policy reviewer." },
  { role: "user", content: "Does this refund request meet the policy? Explain why." },
];

const serv = await client.chat.completions.create({ model: "gpt-5.4-mini", messages });
const rawResponse = await fetch(
  "https://inference-api.openserv.ai/v1/chat/completions",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.SERV_API_KEY}`,
      "Content-Type": "application/json",
      "x-openserv-disable-braid": "true",
    },
    body: JSON.stringify({ model: "gpt-5.4-mini", messages }),
  },
);
const raw = await rawResponse.json();

console.log({ serv: serv.choices[0].message.content, raw: raw.choices[0].message.content });
```

Keep the model, input, system prompt, tools, and sampling settings identical between runs. Use an evaluation set that includes normal, edge, incomplete, and adversarial cases. Compare task success, failure rate, latency, token usage, and total cost across the full set.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.