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

# Tavily Extract

Tavily is a web data retrieval API designed for developers and AI applications. This document describes the Tavily passthrough endpoints exposed by the platform gateway. It is based on Tavily's Extract API with platform route prefixes applied.

Base URL example: `https://api.novita.ai`

Extract web page content from one or more specified URLs.

## Request Headers

All endpoints require platform API authentication.

**Content-Type** `string` *required*

Use `application/json`.

**Authorization** `string` *required*

Platform API key, formatted as `Bearer <api_key>`.

## Request Body

**urls** `string | string[]` *required*

URL(s) to extract.

**query** `string`

Query for snippet reranking.

**chunks\_per\_source** `integer`

Max number of content snippets per source to return. Default is `3`; range is `1` to `5`. Available only when `query` is provided.

**extract\_depth** `string`

Extraction mode. Common values include `basic` and `advanced`. `basic` is set as the default value.

**include\_images** `boolean`

Returns a list of images from the response URLs.

**include\_favicon** `boolean`

Include favicon URL from result.

**format** `string`

Extracted Content Format. Common values include `markdown` and `text`.

**timeout** `float`

URL extraction time limit in seconds. Range is `1.0` to `60.0`. `10` for `basic`, `30` for `advanced`.

## Request Example

```bash theme={"system"}
curl -s -X POST 'https://api.novita.ai/v3/tavily/extract' \
  -H 'Authorization: Bearer sk-123' \
  -H 'Content-Type: application/json' \
  -d '{
    "urls": ["https://en.wikipedia.org/wiki/Lionel_Messi"],
    "query": "career highlights and achievements",
    "chunks_per_source": 3,
    "extract_depth": "basic",
    "include_images": false,
    "format": "markdown",
    "timeout": 15
  }'
```

## Response

**results** `object[]`

Extracted results from URL(s). See [Result Object](#result-object).

**failed\_results** `object[]`

Unprocessable URL(s). See [Failed Result Object](#failed-result-object).

**response\_time** `float`

Request duration in seconds.

**request\_id** `string`

Unique request identifier.

### Result Object

**url** `string`

Response content URL.

**raw\_content** `string`

Full page content.

**images** `string[]`

Image URLs from page URL. Returned if `include_images` is `true`.

**favicon** `string`

Site favicon URL. Returned if `include_favicon` is `true`.

### Failed Result Object

**url** `string`

Unprocessed URL(s)

**error** `string`

URL processing error message.

## Errors

The platform may return standard HTTP errors before forwarding the request, and Tavily may return upstream errors after forwarding.

**400** `status`

Invalid request body or unsupported parameter value. May occur if over 20 URLs are provided.

**401** `status`

Missing or invalid API key.

**403** `status`

Access denied by platform or upstream provider.

**404** `status`

Route or requested resource not found.

**429** `status`

Rate limit exceeded.

**500** `status`

Internal server error.

**502** `status`

Upstream provider error.

**503** `status`

Service unavailable.

## Notes

* All request bodies are JSON.
* Extra Tavily parameters not listed here may be passed through.
* Response shapes can vary depending on request options.
* This document intentionally omits billing-related fields.

## References

For more details, see the [Tavily Extract API reference](https://docs.tavily.com/documentation/api-reference/endpoint/extract).


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