> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lev8.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Deep Search

> Run a synchronous deep search from a natural-language question.

[`POST /v1/deep-search`](/api-reference/deep-search/run) answers an open-ended research question and returns the generated answer in the same response.

Use Deep Search for research and question answering when the input can be expressed as one natural-language `query`.

## Headers

| Header                           | Required    | Description                                                                                            |
| -------------------------------- | ----------- | ------------------------------------------------------------------------------------------------------ |
| `Content-Type: application/json` | Yes         | Other media types return `422`.                                                                        |
| `Accept: application/json`       | Recommended | Declares that the client expects a JSON response.                                                      |
| `Authorization`                  | Yes         | `Bearer lev8_live_...` -- an active Lev8 API key with Deep Search access.                              |
| `Idempotency-Key`                | No          | Stable key for safely replaying the same logical request. Lev8 generates and returns one when omitted. |

## Request body

| Field   | Type   | Required | Description                                                                    |
| ------- | ------ | -------- | ------------------------------------------------------------------------------ |
| `query` | string | Yes      | Natural-language research question, from 1 to 4,000 characters after trimming. |

## Response

On success, `success` is `true` and `answer` contains the generated answer as any non-null JSON value. Lev8 passes this value through without converting it or enforcing a fixed string or object shape.

The current search service commonly returns an object with `Result` for the generated response and `Citations` for source attribution. Treat those fields as an example of the current upstream format, not as a stable schema guaranteed by the Lev8 API.

```json theme={null}
{
  "success": true,
  "answer": {
    "Citations": [],
    "Result": "OpenAI is an AI research and product company. Its CEO is Sam Altman."
  }
}
```

A business-level failure can return `success: false` with a non-null explanatory `answer`. Missing or `null` answers are treated as upstream protocol failures and returned through the standard error envelope.

Errors outside the Deep Search workflow use the standard [error envelope](/api-reference/errors).