Render Options

Basic Options

Basic options for rendering such as setting the URL or HTML, and viewport width and height

url

The URL or domain of the website you want to screenshot. We will automatically prepend http:// if it is missing.

url Examples

html

The HTML you want to render.

html Examples

format

The output format of the resulting render.

The available values are:

png jpeg webp avif svg pdf html mp4 webm md

format Examples

width

default: 1280

The viewport width of the browser, in pixels.

width Examples

height

default: 1024

The viewport height of the browser, in pixels.

height Examples

full_page

default: false

Specify whether to capture the full scrollable area of the website. For PDFs, full_page mode will attempt to capture the whole website onto one single page PDF document. It's likely you'll want to also hide any cookie banners that crop up during a full page screenshot, so we recommend you use click_accept and hide_cookie_banners too.

full_page Examples

full_page_slices

default: false

Split the full page screenshot into smaller vertical slices, each stored as its own image. The JSON response includes a slices array containing the url, offset_y, width and height of every slice. Very tall screenshots are often downscaled or rejected by AI vision models, so sending smaller slices one at a time usually produces better analysis results. Requires full_page to be true. Control the size of each slice with full_page_slice_height and full_page_slice_overlap_height.

full_page_slices Example

full_page_slice_height

default: 4000

The maximum height in pixels of each slice when full_page_slices is enabled. Must be between 1 and 16000. If the full page screenshot is shorter than this value, a single slice is returned.

full_page_slice_height Examples

full_page_slice_overlap_height

default: 0

The number of pixels of vertical overlap between adjacent slices when full_page_slices is enabled. Overlap means content cut at a slice boundary appears in full in at least one slice, which helps when analysing slices independently. The full_page_slice_height minus the overlap must be at least 100.

full_page_slice_overlap_height Examples

selector

Take a screenshot of the element that matches this selector. By default, if the selector is not found, Urlbox will take a normal viewport screenshot. If you prefer Urlbox to fail the request when the selector is not found, pass fail_if_selector_missing=true.

selector Examples

clip

Clip the screenshot to the bounding box specified by x,y,width,height.

clip Examples

gpu

Only available on: ultra and above

default: false

Enable GPU acceleration to render 3D scenes and heavy WebGL content. This is a beta feature and requires pre-approval. Please contact [email protected] to enable this feature on your account.

gpu Examples

response_type

For render link requests, setting this option to json will change the response type of the Urlbox request to JSON. For the API, the default response type is JSON.

The available values are:

json binary

response_type Examples

secure_mode

Only available on: ultra and above

Enforces zero data retention through secure mode to ensure sensitive data remains protected throughout the rendering process. All request data is automatically purged within 90 seconds of the render completing. In addition to this option, you need to use one of the storage approaches described in our Secure Screenshots guide.

Blocking Options

Options for blocking or dismissing certain page elements, such as cookie banners.

block_ads

Blocks requests from popular advertising networks from loading.

block_ads Examples

hide_cookie_banners

Automatically hides cookie banners from most websites, by setting their style to display: none !important;

hide_cookie_banners Examples

click_accept

Similar to the hide_cookie_banners option, but instead of hiding the banners, this option attempts to click on the 'Accept' button, in order to accept cookies.

click_accept Examples

press_escape

Attempts to press the Escape (ESC) key before capturing the page. Useful for dismissing pop-ups, overlays, or advertising banners that appear on load.

block_urls

Block requests from specific domains from loading. You can use wildcard characters such as * to match subdomains.

block_urls Examples

block_images

Blocks image requests

block_images Example

block_fonts

Blocks font requests

block_fonts Examples

block_medias

Block video and audio requests

block_medias Examples

block_styles

Prevent stylesheet requests from loading

block_styles Examples

block_scripts

Prevent requests for javascript scripts from loading

block_scripts Examples

block_frames

Prevents iframe and frame content from loading by blocking non-navigation document requests. The main page will load normally, but any embedded frames/iframes will be blocked.

block_frames Example

block_fetch

Block fetch requests from the target URL.

block_xhr

Block XHR requests from the target URL.

block_xhr Example

block_sockets

Prevents WebSocket connections from being established, blocking real-time communication features like live chat, notifications, or dynamic updates.

block_data_urls

Block data URLs such as data:image/png;base64,...

hide_selector

Hide specific HTML elements on the page before rendering. This option accepts a comma-delimited string of CSS element selectors that will be hidden by setting their style to visibility: hidden !important; pointer-events: none !important;. This preserves the page layout while making elements invisible.

This is particularly useful for:

• Hiding pop-ups, banners, or cookie notices
• Removing advertisements or promotional overlays
• Excluding navigation menus or sidebars
• Hiding specific content sections

Selector types supported:

• Element selectors (e.g., h1, div, img) - Hide all elements of that type
• Class selectors (e.g., .popup, .banner) - Hide elements with specific CSS classes
• ID selectors (e.g., #header, #sidebar) - Hide elements with specific IDs
• Complex selectors (e.g., .nav ul li, div.content > p) - Use any valid CSS selector
• Multiple selectors - Combine multiple selectors with commas

Tip: To find the selector for any element, open your browser's DevTools, right-click the element in the Elements tab, and select "Copy > Copy selector".

hide_selector Examples

Customize Options

Customize the look of the page before rendering a screenshot

js

Only available on: ultra and above

Execute custom JavaScript in the context of the page. The JS gets executed after the page's dom has loaded, but before the screenshot is taken. No need to use load etc event handlers to run code, as these events will already have fired by the time this JS gets executed. You can use await to wait for promises to resolve.

js Examples

css

Inject custom CSS into the page

css Examples

dark_mode

default: false

Emulate dark mode on websites by setting prefers-color-scheme: dark

dark_mode Examples

reduced_motion

Prefer less animations on websites by setting prefers-reduced-motion: reduced

reduced_motion Examples

show_timestamp

Shows a timestamp in a header above the rendered screenshot. Can be paired with show URL. If you're rendering a PDF, you can achieve this with the show_header option.

show_timestamp Examples

show_url

Shows a URL in a header above the rendered screenshot. Can be paired with show timestamp. If you're rendering a PDF, you can achieve this with the show_header option.

show_url Examples

Screenshot Options

Options relating to the generated screenshot image

retina

default: false

Take a 'retina' or high-definition screenshot, equivalent to setting a device pixel ratio of 2.0 or @2x. Please note that retina screenshots will be double the normal dimensions and will normally take slightly longer to process due to the much bigger image size.

retina Examples

thumb_width

The width of the generated thumbnail, in pixels. Omit for a full-size screenshot. For generating one or more separate thumbnail files alongside the full-size render (rather than resizing the main image), use the thumbnails option instead.

thumb_width Examples

thumb_height

The height of the generated thumbnail, in pixels. Omit for a full-size screenshot. For generating one or more separate thumbnail files alongside the full-size render (rather than resizing the main image), use the thumbnails option instead.

thumb_height Examples

img_fit

default: cover

How the screenshot should be resized or cropped to fit the dimensions when using thumb_width and/or thumb_height options

The available values are:

• cover - Preserving aspect ratio, attempt to ensure the image covers both thumb_width and/or thumb_height by cropping/clipping to fit.

• contain - Preserving aspect ratio, contain within both thumb_width and/or thumb_height using letterboxing where necessary.

• fill - Ignore the aspect ratio and stretch to both thumb_width and/or thumb_height.

• inside - Preserving aspect ratio, resize the image to be as large as possible while ensuring its dimensions are less than or equal to thumb_width and/or thumb_height.

• outside - Preserving aspect ratio, resize the image to be as small as possible while ensuring its dimensions are greater than or equal to thumb_width and/or thumb_height.

img_fit Examples

img_position

default: center

How the image should be positioned when using an img_fit of cover or contain.

The available values are:

north northeast east southeast south southwest west northwest center centre

img_position Examples

img_bg

Background colour to use when img_fit is contain, or img_pad is used, defaults to black without transparency

img_bg Examples

img_pad

Pad the screenshot, giving it a border. Can either be a single pixel value that gets added to each side, or a comma delimited string of top,right,bottom,left pixel values.

img_pad Examples

quality

default: 80

The image quality of the resulting screenshot (JPEG/WebP only)

quality Examples

transparent

default: false

If a website has no background color set, the image will have a transparent background (PNG/WebP only)

max_height

For extremely lengthy websites, it may be preferable to limit the screenshot to a maximum height to prevent Urlbox from spending time scrolling and generating an enormous screenshot.

max_height Example

download

Pass in a filename which sets the content-disposition header on the response. E.g. download=myfilename.png This will make the Urlbox link downloadable, and will prompt the user to save the file as myfilename.png

download Example

Thumbnail Options

Options relating to generating thumbnails of the main screenshot. Each spec's fit, bg and position fall back to the img_fit, img_bg and img_position options in Screenshot Options, which also serve the legacy thumb_width/thumb_height single-resize.

thumbnails

Generate up to 5 additional thumbnail images from the same render, each uploaded as its own file alongside the main screenshot. thumbnails is independent of the main screenshot's own dimensions, and unrelated to the legacy thumb_width/thumb_height options above, which resize the primary screenshot itself.

Each entry in the array is an object with the following keys, all optional:

• key - A short identifier (max 10 characters) for this thumbnail. Used as its filename suffix, and as its key in the JSON response (or in thumbnails_object's object response).
• preset - One of xs, sm, md, lg, xl, 2xl, 3xl, 4xl, 5xl, 1/2, 1/4, 3/4. Scales the thumbnail as a percentage of the original screenshot's dimensions (e.g. md is 30%, 1/2 is 50%). Takes priority over size/width/height when set.
• size - A single pixel value (10-2000), used as both the width and the height.
• width - The thumbnail width in pixels (10-2000).
• height - The thumbnail height in pixels (10-2000).
• fit - How the thumbnail should be resized or cropped to fit its dimensions: cover, contain, fill, inside or outside. Falls back to the top-level img_fit option, then cover.
• bg - Background color used for letterboxing. Falls back to the top-level img_bg option, then bg_color, then black.
• position - How the image should be positioned within its fit. Any img_position value, including attention and entropy, which crop around the most visually interesting or highest-entropy region of the image. Falls back to the top-level img_position option, then north.
• suffix - A custom filename suffix, used when key isn't set.
• presigned_url - A presigned URL to upload this specific thumbnail to, instead of Urlbox's own storage.

thumbnails can be supplied as a JSON array in a POST body:

"thumbnails": [{ "preset": "md" }, { "width": 320, "key": "w320" }]

or, over a GET/query-string request, using qs-style bracket notation:

&thumbnails[0][preset]=md&thumbnails[1][width]=320&thumbnails[1][key]=w320

Each generated thumbnail appears in the JSON response (see response_type) as an entry in a thumbnails array, each with a key, location and size - or, when thumbnails_object is true, as an object keyed by that key/suffix instead of an array.

thumbnails Example

thumbnails_object

default: false

Reshapes the thumbnails response from an array into an object keyed by each thumbnail's key/suffix, with key omitted from each value (since it's now the object's own key). Useful for looking up a specific thumbnail directly instead of scanning the array.

For example, with thumbnails: [{ preset: "md", key: "w320" }], the default array response is:

"thumbnails": [{ "key": "w320", "location": "...", "size": 12345 }]

and with thumbnails_object: true, it becomes:

"thumbnails": { "w320": { "location": "...", "size": 12345 } }

Has no effect unless thumbnails is also set.

thumbnails_object Examples

PDF Options

Options relating to PDF document generation.

pdf_page_size

default: A4

Sets the PDF page size.

Setting this option will take precedence over pdf_page_width and pdf_page_height.

The available values are:

A0 A1 A2 A3 A4 A5 A6 Legal Letter Ledger Tabloid

pdf_page_size Examples

pdf_page_range

Sets the PDF page range to return.

By default, the page is split into a multi page document and returns all page. Use this option to restrict which pages should be returned.

pdf_page_range Examples

pdf_page_width

Sets the PDF page width, in pixels.

pdf_page_width Examples

pdf_page_height

Sets the PDF page height, in pixels.

pdf_page_height Examples

pdf_margin

default: none

Sets the margin of the PDF document.

The available values are:

none default minimum

pdf_margin Examples

pdf_margin_top

Sets a custom top margin on the PDF.

pdf_margin_top Example

pdf_margin_right

Sets a custom right margin on the PDF.

pdf_margin_right Example

pdf_margin_bottom

Sets a custom bottom margin on the PDF.

pdf_margin_bottom Example

pdf_margin_left

Set a custom left margin on the PDF.

pdf_margin_left Example

pdf_auto_crop

Automatically remove white space from PDF. Occasionally a PDF will have a lot of trailing white space at the bottom of the page. This option will attempt to automatically crop the PDF to remove this white space.

pdf_scale

default: 1

Sets the scale factor of the website content in the PDF. Valid values are numbers between 0.1 and 2.

pdf_scale Examples

pdf_orientation

default: portrait

Sets the orientation of the PDF.

The available values are:

portrait landscape

pdf_orientation Examples

pdf_background

default: true

Sets whether to print background images in the PDF

pdf_background Examples

disable_ligatures

Prevents ligatures from being used. Useful when rendering a PDF, and you want to extract text which contains ligatures.

media

By default, when generating a PDF, the print CSS media query is used. To generate a PDF using the screen CSS, set this option to screen.

When generating an image, the screen CSS media query is used by default. To generate an image using the print CSS, set this option to print.

media Examples

pdf_show_header

Whether to show the default pdf header on each page of the pdf. The template of the header can be changed by setting the pdf_header option.

pdf_show_header Example

pdf_header

Change the default pdf header that is shown on each page of the pdf when pdf_show_header option is set.

You have the option to show the following variables in the header (or footer) of the pdf:

• current date
• title of the page
• url of the page
• current pageNumber
• the totalPages in the pdf document

You can display these variables by creating empty divs or spans, with special css class names relating to the variable you want to show.

For example, if you want to show the date followed by the url, you could use the following pdf header template:

<div class="date"></div><div class="url"></div>.

The pdf header template you set are inserted as the innerHTML of a parent div which is a flex container, and has align-items set to flex-start.

There are also some helper classes for aligning the divs or spans. The following classes are available:

• left - adds some left padding to the element and sets flex: none.
• center - aligns the element and text to the center.
• right - adds some right padding to the element and sets flex: none.
• text - sets the text to 8pt.
• grow - sets flex: auto on the element, allowing it to grow to fill the available space.

The default pdf header is:

<div class='date text left'></div><div class='title text center'></div>.

You can see exactly how the pdf page is constructed by looking at the chromium pdf template in the chromium source repository.

pdf_header Examples

pdf_show_footer

Whether to show the default pdf footer on each page of the pdf. The template of the footer can be changed by setting the pdf_footer option.

pdf_show_footer Example

pdf_footer

Change the default pdf footer that is shown on each page of the pdf when pdf_show_footer option is set.

You have the option to show the following variables in the footer (or header) of the pdf:

• current date
• title of the page
• url of the page
• current pageNumber
• the totalPages in the pdf document

You can display these variables by creating empty divs or spans, with special css class names relating to the variable you want to show.

For example, if you want to show the date followed by the url, you could use the following pdf footer template:

<div class="date"></div><div class="url"></div>.

The pdf footer template you set are inserted as the innerHTML of a parent div which is a flex container, and has align-items set to flex-end.

There are also some helper classes for aligning the divs or spans. The following classes are available:

• left - adds some left padding to the element and sets flex: none.
• center - aligns the element and text to the center.
• right - adds some right padding to the element and sets flex: none.
• text - sets the text to 8pt.
• grow - sets flex: auto on the element, allowing it to grow to fill the available space.

The default pdf footer is:

<div class='url text left grow'></div><div class='text right'><span class='pageNumber'></span>/<span class='totalPages'></span></div>.

You can see exactly how the pdf page is constructed by looking at the chromium pdf template in the chromium source repository.

pdf_footer Examples

readable

Make the pdf into a readable document by removing unnecessary elements such as navigation bars, ads, etc.

readable Example

pdf_title

Sets the title metadata field of the PDF document. This is visible in PDF readers under document properties. If not set, the page's <title> tag will be used as the PDF title.

pdf_subject

Sets the subject metadata field of the PDF document. This is visible in PDF readers under document properties.

pdf_author

Sets the author metadata field of the PDF document. This is visible in PDF readers under document properties.

pdf_keywords

Sets the keywords metadata field of the PDF document. Pass a comma-separated list of keywords, e.g. "pdf_keywords": "screenshot,web,api". This is visible in PDF readers under document properties and can help with document organization and search.

pdf_creator

Sets the creator metadata field of the PDF document. This typically indicates the application that created the original content. Visible in PDF readers under document properties.

Video Options

Options for rendering MP4 or WebM videos of a website. Set format to mp4 or webm to record a video instead of taking a screenshot. Pair these with the Video Scrolling Options below to control what the video shows.

video_time

default: 3000

How long to record for, in milliseconds. When scrolling is enabled the scroll choreography runs within this time; without scrolling the recording simply lasts this long. video_duration is an alias that takes seconds instead of milliseconds. When using video_scroll_to, the video length is derived from your sections automatically and video_time caps the scroll choreography (the section scrolls and waits, with lead-in, scroll-back and tail riding on top).

video_time Example

video_fps

default: 60

Frames per second to record at. Higher values give smoother scrolling at the cost of larger files and slightly longer render times.

video_fps Example

video_width

default: 1280

Width of the output video in pixels. The page is rendered at the usual viewport size (width/height) and the recording is scaled to video_width x video_height.

video_height

default: 1024

Height of the output video in pixels.

video_height Example

video_quality

default: 23

Constant rate factor for the H.264 encoder. Lower is higher quality and larger files. Typical values are 18 (visually lossless) to 28.

video_preset

default: medium

Encoder speed/compression trade-off. Slower presets compress better (smaller files at the same quality) but take longer to encode.

The available values are:

ultrafast superfast veryfast faster fast medium slow slower veryslow

video_preset Example

video_bitrate

Target bitrate for the recording, in kbps. An alternative to video_quality when you need to hit a specific file-size budget.

video_codec

default: h264

Video codec for the output. h264 produces MP4s with the widest playback support; vp9/vp8 produce WebM. Setting format to webm selects a WebM codec automatically.

The available values are:

h264 vp9 vp8

video_codec Example

Video Scrolling Options

Options controlling how the page is scrolled while a video records. Enable a simple top-to-bottom scroll with video_scroll, or choreograph stops at specific sections with video_scroll_to.

video_scroll

default: false

Smoothly scroll down the page while recording, section by section, then scroll back to the top. Scroll speed, easing, pauses and distance are controlled by the options below.

video_scroll Example

video_scroll_to

Scroll to specific sections of the page in turn, pausing at each for a set time, for example to walk through key features or synchronise a scroll with a voiceover. Each entry is a CSS selector (automatically piercing open shadow DOM, however deeply nested) or a text= locator matching visible text, plus optional per-section settings: wait (pause at the section), duration (how long the scroll to it takes), ease (easing name, lower-case, see video_ease), and offset (stop N pixels above the element, useful under sticky headers). In a render link, repeat the parameter once per section using ;key=value modifiers, e.g. video_scroll_to=%23reviews;wait=4s. In a JSON body, pass an array of the same strings or of objects. Durations accept milliseconds (2500) or unit strings ("2.5s", "500ms"). The video's length is derived from your sections (capped at 400s; an explicit video_time truncates). If a section isn't found, its full time slot still elapses at the current position so the rest of the timeline stays in sync. A literal ; inside a text= value needs the JSON object form. video_scroll_offset sets the default offset used for every section that doesn't specify its own.

video_scroll_to Examples

video_scroll_require_sections

default: false

Fail the render with a 400 error if any video_scroll_to section can't be found on the page, instead of holding position through its time slot. Use this when a wrong video is worse than no video.

video_scroll_offset

default: 0

Global default viewport offset for video_scroll_to stops, in pixels. Positive values stop the scroll with the viewport top N pixels above the target element, which keeps section headings visible and clears sticky navbars. Negative values scroll past the element. A per-section offset overrides this default for that section only.

video_scroll_duration

default: 1500

How long each scroll movement takes, in milliseconds. Also the default per-section scroll duration in video_scroll_to mode.

video_rest_duration

default: 0

How long to pause between scroll movements, in milliseconds. In video_scroll_to mode this becomes the default per-section wait (falling back to 2000ms if unset).

video_prescroll_duration

default: 1000

How long to hold at the top of the page before scrolling begins, in milliseconds.

video_postscroll_duration

default: 1000

How long to keep recording after scrolling finishes, in milliseconds.

video_scroll_back

default: true

Scroll back to the top of the page at the end of the recording. Set to false to end the video at the final scroll position.

video_scroll_back_duration

default: 2500

How long the final scroll back to the top takes, in milliseconds.

video_ease

default: quadratic.inout

Easing function for scroll movements. Controls how each scroll accelerates and decelerates. Values are lower-case. Also the default per-section ease in video_scroll_to mode.

The available values are:

linear.none quadratic.in quadratic.out quadratic.inout cubic.in cubic.out cubic.inout quartic.in quartic.out quartic.inout quintic.in quintic.out quintic.inout sinusoidal.in sinusoidal.out sinusoidal.inout exponential.in exponential.out exponential.inout circular.in circular.out circular.inout elastic.in elastic.out elastic.inout back.in back.out back.inout bounce.in bounce.out bounce.inout

video_ease Example

video_ease_end

default: quadratic.inout

Easing function for the final scroll back to the top. Accepts the same values as video_ease.

video_scroll_distance

Override the distance of each scroll movement, in pixels. By default each movement covers roughly one viewport. Ignored in video_scroll_to mode.

video_sections

Stop the scroll after this many scroll movements, rather than scrolling the whole page. Ignored in video_scroll_to mode.

video_jitter

Randomise scroll timing by up to this fraction, for a more human, less mechanical feel. For example 0.2 varies each movement and pause by up to 20%.

video_warmup

default: false

Perform an unrecorded scroll down and back up before recording starts, so lazy-loaded images and animations have already loaded when the real scroll happens.

Cache Options

Options to control how Urlbox caches your screenshots or PDF's. Please note that caching only applies to requests from render links. POST requests to the API are not cached.

To remove a cached render before its ttl expires, send a DELETE request to the exact same render link URL (same path and query string as the original GET, excluding force and the token). The cache entry is keyed on the full set of render options, so a DELETE with a different query string targets a different entry. A successful purge responds 200 with {"purged": true}; if no cached render matched the supplied options, the API responds 404 with error code CacheEntryNotFound. Check that your query string exactly matches the original render's. Note that clients may still serve a previously fetched copy from their own HTTP cache until your ttl elapses.

force

default: false

Generate a fresh render on each request, instead of getting a cached version.

unique

Pass a unique string such as a UUID, hash or timestamp, to have more control over when to generate a fresh screenshot or PDF.

ttl

default: 2592000

The duration to keep a screenshot or PDF in the cache, in seconds. ttl stands for 'time to live'. The default value is also the maximum value: 2592000 seconds (30 days).

ttl Examples

Request Options

Options to configure the browser, before navigating to the URL

proxy

Only available on: ultra and above

Pass in a proxy server address to make screenshot requests via that server in the format [address]:[port].

If proxy authentication is required, you can use the following format: [user]:[password]@[address]:[port].

use_proxy

Only available on: ultra and above

This uses the proxy you have stored on your project in use to make screenshot requests via that server.

use_stealth

default: false

Renders the page with a full, real Chrome browser set up the way a normal user's machine is. Standard renders streamline or disable some browser behaviours for speed; stealth renders leave everything in place, so the page sees an ordinary, consistent browsing environment.

Use it when a page renders incompletely, or returns a challenge page or 403, under a plain render. It combines well with retry_on and retry_with (enable stealth only on retries, so you only pay its cost when a render needs it), and with a proxy for sites that respond differently depending on the requesting network.

Stealth renders are noticeably slower than standard renders, so avoid enabling it as a blanket default.

hide_headless

default: false

A lighter-weight alternative to use_stealth. It runs Chrome with a set of plugins that fill in the properties headless Chrome normally leaves unset or empty (for example navigator.webdriver, the chrome runtime objects, and WebGL vendor strings), so the page sees the environment an ordinary desktop browser reports.

use_stealth is the stronger option and takes precedence if both are set. Reach for hide_headless when a page renders differently under headless Chrome but doesn't need the full stealth browser.

header

Only available on: hifi and above

Set a header on the request when loading the URL

Example: To set the header with key X-My-Header to the value SomeValue, you would pass header=X-My-Header%3DSomeValue.

This can be set multiple times, to set more than one header - e.g. header=X-My-Header%3DSomeValue&header=X-My-Other-Header%3DSomeOtherValue.

As with all options passed via the query string, the header value must be URL encoded - so X-My-Header=SomeValue becomes X-My-Header%3DSomeValue in order to be interpreted correctly by Urlbox.

header Examples

cookie

Sets a cookie on the request when loading the URL.

Example: To set the cookie with key Opt-In to the value yes, you would set the value of this option to Opt-In=yes.

Cookies can be passed as an array, to allow setting multiple cookies - e.g.["Opt-In=yes","Session-Id=DMTIzNDU"].

To achieve multiple cookies with render links, just set the cookie option multiple times, like cookie=Opt-In%3Dyes&cookie=Session-Id%3DDMTIzNDU.

To set a specific domain on a cookie, you can do the following: OptIn=yes;Domain=.mydomain.com.

You can set other attributes for the cookie such as Path, HttpOnly and SameSite

cookie Examples

user_agent

Sets the User-Agent string for the request. The user agent identifies what browser/device is making the request, which can affect how websites render content.

Presets:

• random - Uses a random user-agent to help avoid bot detection
• mobile - Uses a modern iPhone/Safari user-agent string
• desktop - Uses a modern Chrome/macOS user-agent string

Why use this?

• Some websites serve different content based on the user agent (e.g., mobile vs desktop layouts)
• Certain sites block requests from unknown or bot-like user agents
• You may want to emulate how a specific browser or crawler sees a page

Testing your user agent: Try rendering httpbin.org/user-agent to see exactly what user agent string is being sent.

For a comprehensive list of user agent strings, see useragents.me.

user_agent Examples

platform

default: MacIntel

Sets the navigator.platform that the browser will report for the request. Useful for getting around certain scripts that detect the platform.

platform Examples

accept_lang

default: en-US

Sets an Accept-Language header on requests to the target URL

accept_lang Examples

authorization

Sets an Authorization header on requests to the target URL. Can be used to pass an auth token through to the site in order to 'login' before rendering.

authorization Examples

tz

default: America/New_York

Emulate the timezone to use when rendering pages.

By default the rendering browser reports a US timezone (America/New_York or America/Los_Angeles, depending on which of our regions renders your screenshot) so that its clock agrees with where our renderers connect from. Set tz explicitly to pin it. Renders on the stable engine channel still report UTC until the next promotion of latest to stable (see engine_version).

This matters whenever the page derives anything from the browser's local clock: displayed dates and times, opening hours, "expires in" countdowns, and any cookie or token whose lifetime the page calculates client-side. If your renders need to be identical every time, or your page assumes UTC, set tz=UTC.

Example: tz=Europe/London. A list of timezone ID's can be found here: https://en.wikipedia.org/wiki/List_of_tz_database_time_zones

tz Examples

engine_version

default: latest

Sets the version of the urlbox rendering engine to use when rendering the page. Renders default to latest unless you choose a version here or on your project. Pin stable when you'd rather pick up engine changes in batches, at the scheduled promotions of latest to stable, than as they ship.

The available values are:

stable latest experimental

certify

Only available on: ultra and above

This creates a hash of the rendered file, timestamp and options providing proof that a render was taken at a given time. Returns a hash, timestamp and the options used to hash. Checkout our guide on certifying a render for more information.

Wait Options

Options to control how Urlbox waits for the page or elements to load before rendering.

delay

default: 0

The amount of time to wait before Urlbox captures a render in milliseconds.

timeout

default: 30000

The amount of time to wait for the requested URL to load, in milliseconds. The timeout value needs to be between 5,000 and 100,000 milliseconds. The default is 30000 or 30 seconds.

wait_until

default: loaded

Waits until the specified DOM event has fired before capturing a render.

The available options are:

• domloaded (the DOMContentLoaded event is fired)
• mostrequestsfinished (consider navigation to be finished when there are no more than 2 network connections for at least 500 ms)
• requestsfinished (there are no more than 0 network connections for at least 500 ms)
• loaded (the load event is fired)

The available values are:

domloaded mostrequestsfinished requestsfinished loaded

wait_until Examples

wait_for

Waits for the element specified by this selector to be present in the DOM before taking a screenshot or PDF.

By default, Urlbox will take a screenshot or PDF if the wait_for element is not found after waiting for the time specified by the wait_timeout option.

If you prefer Urlbox to fail the request when the wait_for element is not found, pass fail_if_selector_missing=true

wait_for Examples

wait_for_state

default: attached

Whether the element specified by wait_for should be visible or just present in the DOM. Visible means the element is attached to the DOM and has opacity > 0 and its width and height > 0 and doesn't have visibility:hidden set.

The available values are:

visible attached

wait_to_leave

Waits for the element specified by this selector to be absent from the DOM before taking a screenshot or PDF.

A typical use-case would be waiting for loading spinners to be absent before taking a screenshot.

By default, Urlbox will take a screenshot or PDF if the wait_to_leave element is still present after the time specified by the wait_timeout option.

If you prefer Urlbox to fail the request when the wait_to_leave element is still present, pass fail_if_selector_present=true

wait_to_leave Examples

wait_timeout

default: 30000

The amount of time to wait for the wait_for element to appear, or the wait_to_leave element to leave before continuing, in milliseconds.

Fail Options

Options to dictate how Urlbox handles certain scenarios.

fail_if_selector_missing

default: false

Fails the request if the elements specified by selector or wait_for options are not found on the page after waiting for wait_timeout.

fail_if_selector_missing Example

fail_if_selector_present

default: false

Fails the request if the element specified by wait_to_leave option is found on the page after waiting for wait_timeout.

fail_if_selector_present Example

fail_on

Pass in a specific HTTP status code (e.g., "400", "500") or an array of status codes (e.g., ["400", "404", "500"]) as strings. Urlbox will fail the request if the final response status code matches any of the specified codes.

fail_on Example

fail_on_4xx

default: false

If fail_on_4xx=true and the requested URL returns a status code between 400 and 499, Urlbox fails the request with a 400 error, an error code of fail_on_4xx, and a message naming the status code (for example Page returned 404 and fail_on_4xx was true).

fail_on_4xx Examples

fail_on_5xx

default: false

If fail_on_5xx=true and the requested URL returns a status code between 500 and 599, Urlbox fails the request with a 400 error, an error code of fail_on_5xx, and a message naming the status code (for example Page returned 500 and fail_on_5xx was true).

fail_on_5xx Examples

fail_on_except

Specify status codes to exclude from fail_on, fail_on_4xx, or fail_on_5xx checks. For example, if you want to fail on all 5xx errors except 503, use fail_on_5xx=true with fail_on_except=503.

retry_on

Only available on: ultra and above

Automatically retry renders when specific conditions occur. Pass a single condition, a comma-separated string, or an array of conditions.

When a condition matches, Urlbox will retry with exponential backoff (the delay doubles each time, starting at 1 second by default and capped at 30 seconds per wait) up to 2 times by default (3 total attempts including the original render). Use max_retries or max_attempts to change this. Retrying stops early if the request approaches the overall 5 minute render budget.

HTTP status conditions (retry based on the page's response):

• 4xx - Retry when the page returns any 4xx status. This range deliberately skips retrying when the failure looks like a problem with the request itself rather than a block (for example a missing selector or an invalid option), since those would fail identically on every attempt. List codes like 403/429 explicitly to always retry them.
• 5xx - Retry when the page returns any 5xx status
• 404, 429, 503, etc. - Retry on specific status codes

Engine conditions (retry when Urlbox encounters an internal error):

• timeout - Retry when the render times out
• crash - Retry when the browser crashes

Quality conditions:

• small_size - Retry when the screenshot is smaller than min_size_bytes

Special values:

• all - Retry on any of the conditions above

Examples:

• retry_on: "4xx" - Retry when the page returns any 4xx status
• retry_on: ["429", "503"] - Retry on rate limiting or service unavailable
• retry_on: "timeout,crash" - Retry on engine failures only
• retry_on: "5xx,timeout" - Retry on page 5xx errors or timeouts

Note: If a condition matches both retry_on and fail_on, the request will be retried first. Use fail_on_except to mark specific status codes as acceptable (won't retry or fail).

retry_with

Only available on: ultra and above

Change render options on retry attempts triggered by retry_on. Instead of retrying with the exact same options, each retry re-runs the render with the retry_with options merged over the original request, so you can escalate progressively: try a cheap, plain render first, and only add stealth or a proxy if the render fails.

retry_with has no effect on its own; it only applies when retry_on triggers a retry. Because the value is a nested object, use it with JSON POST requests to the render endpoints.

Single object - applied to every retry attempt:

{
  "url": "example.com",
  "retry_on": ["403", "429", "timeout"],
  "retry_with": { "use_stealth": true, "use_proxy": true }
}

Array of objects - progressive escalation. The first retry uses the first element, the second retry uses the second element, and so on. If there are more retries than elements, the last element is reused for the remaining retries. For example, if a site rate-limits or bot-blocks you (429 or 403 responses), you can escalate through your own proxies from cheapest to most expensive, so the expensive bandwidth is only spent on renders that actually got blocked:

{
  "url": "example.com",
  "retry_on": ["403", "429"],
  "max_retries": 3,
  "retry_delay_ms": 2000,
  "retry_with": [
    { "proxy": "user:[email protected]:8080" },
    { "proxy": "user:[email protected]:8080" },
    { "proxy": "user:[email protected]:9000" }
  ]
}

The first attempt runs without a proxy. If the page responds 403 or 429, the first retry goes through the cheap datacenter proxy, the second through the mid-tier one, and the third through the residential proxy. Entries aren't limited to the proxy option: any retry-compatible option works, for example { "use_stealth": true, "use_proxy": true } to enable use_stealth mode and route through the proxy saved on your project (like use_proxy). See the proxies guide for a walkthrough of this pattern.

Merging rules:

• Each retry's options are the original request options with that attempt's retry_with entry merged on top.
• Setting an option to null removes it from that retry attempt - e.g. { "proxy": null } retries without the proxy from the original request.
• Options you don't mention are left unchanged.

Retry-compatible options: only options that affect how the page is fetched and rendered can be changed on retry. These include proxy and network options (proxy, use_proxy, user_agent), stealth options (use_stealth, hide_headless), content blocking (block_ads, hide_cookie_banners, block_urls), interaction and waiting (click_accept, click, wait_until, wait_for, delay), viewport (width, height, retina), page modifications (js, css, hide_selector), request headers and cookies (header, cookie, authorization, accept_lang), and timeout. Options that change the output itself (like format or full_page) cannot be changed between attempts.

max_retries

Only available on: ultra and above

default: 2

Maximum number of retry attempts when using retry_on. Must be between 0 and 5. If max_attempts is also set, max_attempts takes precedence.

max_attempts

Only available on: ultra and above

default: 3

Maximum total number of render attempts (the original render plus retries) when using retry_on. Must be between 0 and 5. An alternative to max_retries: max_attempts is equivalent to max_retries + 1, and takes precedence over it when both are set.

retry_delay_ms

Only available on: ultra and above

default: 1000

Base delay in milliseconds between retry attempts. The delay doubles with each retry (exponential backoff), with each individual wait capped at 30 seconds. Must be between 100 and 60000.

min_size_bytes

Only available on: ultra and above

default: 1000

Minimum expected file size in bytes. Used with retry_on: "small_size" to retry when the screenshot is smaller than expected, which may indicate an error page was captured instead of the intended content. Must be at least 100.

Page Options

Options to modify the page state before taking a screenshot or PDF

scroll_to

Scroll, to either an element or to a pixel offset from the top, before taking a screenshot

scroll_to Examples

click

Specifies an element selector to click before generating a screenshot or PDF

Example: #clickme would click an element with id="clickme".

Can be used multiple times to simulate multiple sequential click events.

If the selector matches multiple elements, only the first element will be clicked.

click Examples

click_all

Specifies an element selector to click before generating a screenshot or PDF

Example: .clickme would click all elements with class="clickme".

Can be used multiple times to simulate multiple sequential click events.

If the selector matches multiple elements, all elements will be clicked.

click_all Example

hover

Specifies an element selector to hover over before generating a screenshot or PDF

Example: #hoverme would hover over the element with id="hoverme"

hover Examples

bg_color

Specify a hex code or CSS color string to use as the background color

Some websites don't set a body background colour, and will show up as transparent backgrounds with PNG, or black when using JPG. Use this setting to set a background colour. If the website explicitly sets a transparent background on the html or body elements, this setting will be overridden.

disable_js

default: false

Turns off javascript on the target URL.

disable_js Examples

show_certificate_errors

default: false

Shows browser certificate errors for pages with invalid HTTPS certificates. By default, Urlbox ignores certificate errors so pages with expired, self-signed, or otherwise invalid certificates can still render.

Full Page Options

Advanced options to control how Urlbox takes full page screenshots, when full_page=true

full_page_mode

default: stitch

Whether to use scroll and stitch algorithm (the default) to render a full page screenshot, or to use the native full page screenshot algorithm, which is faster, but can be less accurate on some sites.

The available values are:

stitch native

full_width

default: false

When full_page=true, specify whether to capture the full width of the website, for example if the site is horizontally scrolling.

allow_infinite

By default, when Urlbox detects an infinite scrolling page, it does not attempt to continue scrolling to the bottom, as this could result in infinite scrolling! If you want to override this behaviour, pass true for this option.

skip_scroll

default: false

Enabling skip_scroll will speed up renders by skipping an initial scroll through the page, which is used to trigger any lazy loading elements.

skip_scroll Examples

detect_full_height

Some pages have full-height backgrounds whose heights are set to 100% of the viewport. This can cause the backgrounds to get stretched when making a full page screenshot. If you are seeing this behaviour in your full page screenshots, pass true for this option.

max_section_height

default: 4096

When Urlbox takes a full_page screenshot, the maximum height of each image section is set to 4096 pixels. If a sites height is greater than this value, Urlbox will start splitting the screenshot into sections. Sometimes it is worthwhile experimenting with this number.

scroll_increment

Sets how many pixels to scroll when scrolling the page to trigger lazy loading elements. By default, the scroll increment is set to the browser viewport height. Some pages' lazy loading elements only trigger when the scroll increment is smaller than this, however, e.g. 400px.

scroll_delay

When Urlbox decides to split a screenshot into multiple sections, the scroll delay is the time to wait between taking the screenshots of each individual section, in milliseconds. While Urlbox does detect animations, and attempts to wait for them before taking a screenshot, this option could be used to force Urlbox to wait for a certain amount of time after scrolling to the next section, to wait for things like animations to finish.

Highlighting Options

Options for highlighting a given string on the page. These are useful for either highlighting or hiding (if you use the same foreground and background) words or a given set of characters.

highlight

Specify a string to highlight on the page before capturing a screenshot or PDF. To highlight multiple words, separate words with a pipe character e.g. Hello|World

highlight Example

highlightfg

default: white

Specify the text color of the highlighted word.

highlightfg Examples

highlightbg

default: red

Specify the background color of the highlighted word.

highlightbg Examples

Geolocation Options

Options for the geolocation API.

latitude

Sets the latitude used to emulate the Geolocation API.

latitude Example

longitude

Sets the longitude used to emulate the Geolocation API.

longitude Example

accuracy

Sets the accurate of the Geolocation API in metres.

accuracy Examples

Side Render Options

Options for generating side renders: extra artifacts captured from the same page load as your main render, such as the page's HTML, an MHTML snapshot, a markdown conversion, page metadata, extracted content and thumbnails. Side renders don't trigger a second render: everything is produced while the page is already loaded, and each artifact is saved alongside the main render and returned as an extra URL field in the JSON response, so make sure you are using the API or response_type=json to see them. See the side renders guide for a walkthrough.

Side render and thumbnail file sizes count toward your render's output size, which is billed at one render credit per 5MB of file size over the included allowance.

save_markdown

default: false

Converts the rendered page to markdown and saves it alongside the main render, so you can capture a screenshot and an LLM-friendly text version of the page in one request. The JSON response gains a markdownUrl field linking to the saved .md file. To get markdown as the main render itself, use format=md instead.

save_markdown Example

save_html

default: false

Saves the page's rendered HTML alongside the main render. The HTML is captured from the live DOM after the page has finished loading, so it includes any changes made by javascript. The JSON response gains an htmlUrl field linking to the saved .html file.

save_html Example

save_metadata

default: false

Extracts the page's metadata (title, description, author, canonical URL, open graph tags, twitter card tags and other meta tags, along with the requested and resolved URLs) and saves it as a JSON file alongside the main render. The JSON response gains a metadataUrl field linking to the saved file, and the metadata is also returned inline in the metadata field.

save_metadata Example

save_headings

default: false

Extracts every heading (h1-h6) from the page and saves them as a JSON side render. Each entry contains the heading level, its text, and its id when present - useful for building a table of contents or checking a page's structure. The JSON response gains a headingsUrl field linking to the saved file.

save_headings Example

save_tables

default: false

Extracts every HTML <table> on the page and saves them as a JSON side render. Each table contains its cells organised as rows and columns, a flag per row indicating whether it is a header row, and a markdown rendering of the whole table. The JSON response gains a tablesUrl field linking to the saved file.

save_tables Example

save_structured_data

default: false

Extracts structured data embedded in the page, such as JSON-LD / schema.org markup, and saves it as a JSON side render. Each entry contains the data type, the raw JSON and, when detected, the schema type (e.g. Product or Article). The JSON response gains a structuredDataUrl field linking to the saved file.

save_structured_data Example

save_links

default: false

Extracts every link on the page - resolved to absolute URLs, along with the link text - and adds them to the page metadata under links. The metadata is returned in the metadata field of the JSON response and saved as a JSON side render linked from metadataUrl.

save_links Example

save_clicks

default: false

Records the elements Urlbox clicked automatically during the render - for example the cookie-banner buttons clicked by click_accept - and adds them to the page metadata under clickedOn, including each clicked element's selector, tag name, classes and text. Useful for auditing exactly what was dismissed before your screenshot was taken. The metadata is returned in the metadata field of the JSON response and saved as a JSON side render linked from metadataUrl.

save_clicks Example

save_cookies

default: false

Saves the cookies set in the browser by the end of the render into the page metadata, as an array of cookie strings under cookies plus a cookiesSaved count. The metadata is returned in the metadata field of the JSON response and saved as a JSON side render linked from metadataUrl.

save_cookies Example

save_headers

default: false

Includes the target page's HTTP response headers in the JSON response. By default the response object contains the status code and requested/resolved URLs only; setting save_headers=true adds a headers object containing the headers the target site responded with.

save_headers Example

save_http_headers

default: false

Alias of save_headers - includes the target page's HTTP response headers in the response.headers field of the JSON response.

save_http_headers Example

save_resource_info

default: false

Saves a JSON debug artifact listing every resource the page requested during the render - each entry records the URL, HTTP method, resource type, content type, status code, transferred and inflated byte sizes, and encoding. This is the data Urlbox uses to calculate the render's bandwidth. The file is saved to storage alongside the render; its location is not returned in the standard render response, but is visible in the render's details in your dashboard, making this primarily a debugging aid.

save_resource_info Example

save_mhtml

default: false

Saves an MHTML snapshot of the page alongside the main render. MHTML bundles the page and its resources into a single file that can be opened in a browser for offline viewing. The JSON response gains an mhtmlUrl field linking to the saved .mhtml file.

save_mhtml Example

thumbnails

Generates up to 5 thumbnails of the main render, resized from the same captured image - no extra render is performed. Applies to image formats (png, jpeg, webp, avif); each thumbnail keeps the main render's format. Because the value is an array of objects, use it with JSON POST requests to the render endpoints.

Each thumbnail object accepts:

• key - a name for the thumbnail, up to 10 characters. Used as the file suffix and as the thumbnail's key in the response. When omitted, a key is derived from the preset or dimensions (e.g. 400, 400x300 or sm).
• preset - a named size relative to the main render's dimensions: xs (10%), sm (20%), md (30%), lg (40%), xl (50%), 2xl (60%), 3xl (70%), 4xl (80%), 5xl (90%), or the fractions 1/4, 1/2 and 3/4.
• size - sets both the width and height, in pixels. Must be between 10 and 2000.
• width / height - individual pixel dimensions, each between 10 and 2000. Ignored on an axis where a preset applies.
• fit - how the image should be resized to fit the dimensions: cover, contain, fill, inside or outside (see img_fit for what each does). Defaults to the request's img_fit, or cover.
• position - how the image is positioned when fit is cover or contain (see img_position).
• bg - background colour used when fit leaves empty space, e.g. with contain. Defaults to the request's img_bg or bg_color, then black.
• suffix - overrides the file name suffix used for the thumbnail in storage.
• presigned_url - an S3 presigned URL to upload this thumbnail to, if you want it delivered straight into your own bucket.

Each thumbnail must resolve to a unique key. The JSON response gains a thumbnails array with one { "key": ..., "location": ..., "size": ... } entry per thumbnail - set thumbnails_object to get them keyed by name instead. Thumbnail file sizes count toward the render's output size for billing, like other side renders.

thumbnails Examples

thumbnails_object

default: false

Returns the thumbnails results as an object keyed by each thumbnail's key, instead of an array - so "thumbnails": [{ "key": "small", "location": ..., "size": ... }] becomes "thumbnails": { "small": { "location": ..., "size": ... } }. Useful when you want to look thumbnails up by name rather than position.

thumbnails_object Example

Storage Options

Options related to storing renders in your own S3-compatible or Azure Blob Storage bucket.

use_s3

Only available on: hifi and above

default: false

Save the render directly to the S3 (or S3-Compatible) bucket configured on your account.

Mutually exclusive with use_azure - a request setting both fails.

s3_path

Only available on: hifi and above

Sets the S3 path, including subdirectories and the filename, to use when saving the render in your S3-compatible bucket.

no_suffix

Only available on: hifi and above

By default, urlbox adds the file extension (e.g. .png, .jpg, .pdf etc) to the s3_path (or azure_path).

If no_suffix=true, the file extension will NOT be added to the path.

use_azure

Only available on: hifi and above

default: false

Save the render directly to the Azure Blob Storage container configured on your project. Azure is not S3-compatible, so it has its own credentials (storage account, container and SAS token) and options - see the Azure Blob Storage guide for setup.

Mutually exclusive with use_s3 - a request setting both fails.

azure_path

Only available on: hifi and above

Sets the blob path, including subdirectories and the filename, to use when saving the render in your Azure container - the Azure equivalent of s3_path. Defaults to renders/{year}/{month}/{day}/{renderId}.

s3_bucket

Only available on: hifi and above

Overrides the configured bucket to use when saving the render.

s3_endpoint

Only available on: hifi and above

You can change the endpoint URL to use an S3 compatible storage provider e.g. DigitalOcean Spaces, Minio, Wasabi, Cloudflare R2 and more.

s3_region

Only available on: hifi and above

Override the configured S3 region when saving the render.

cdn_host

Only available on: hifi and above

If your custom bucket is fronted by a CDN, you can set the host name here.

s3_storageclass

Only available on: hifi and above

default: standard

Sets the s3 storage class.

The available values are:

standard standard_ia reduced_redundancy onezone_ia intelligent_tiering glacier deep_archive outposts

LLM Options

Options related to LLM usage.

use_llm

Only available on: ultra and above

Use the LLM configuration setup in your project settings, or those passed into the request.

llm_prompt

Only available on: ultra and above

The prompt to give the LLM EG "Analyse this image and its associated HTML, giving me back a summary of what the website contents are and a list of all of the links it has "

llm_system_prompt

Only available on: ultra and above

This can be used to provide more overall context for the AI's response.

llm_key

Only available on: ultra and above

The API access key for the given LLM provider.

llm_provider

Only available on: ultra and above

The LLM provider to use.

The available values are:

anthropic openai google azure mistral cohere amazon-bedrock google-vertex groq xai deepseek perplexity togetherai fireworks cerebras openrouter

llm_model

Only available on: ultra and above

The LLM model to use for the given provider. E.g. 'gpt-5.1' for OpenAI, 'claude-sonnet-4-5-20250929' for Anthropic.

llm_temperature

Only available on: ultra and above

The temperature (creativity) for the LLM prompt. Defaults to 0 for less creative responses.

llm_max_tokens

Only available on: ultra and above

The max number of output tokens the LLM can generate a response with. Defaults to 1000.

llm_height

Only available on: ultra and above

The height of the thumbnail image that is sent to the LLM. Defaults to 512.

llm_width

Only available on: ultra and above

The width of the thumbnail image that is sent to the LLM. Defaults to 512.

llm_base_url

Only available on: ultra and above

Override the default API endpoint for the LLM provider. Useful for proxies, self-hosted models, or custom endpoints.

llm_azure_resource_name

Azure OpenAI resource name. Used to construct the endpoint URL: https://{resourceName}.openai.azure.com/

llm_azure_api_version

Azure OpenAI API version. Required when using deployment URLs. E.g. '2024-02-15-preview'.

llm_azure_use_deployment_urls

Use legacy Azure deployment URL format. Useful for compatibility with certain Azure OpenAI models or deployments that require the legacy endpoint format.

llm_aws_region

AWS region for Amazon Bedrock. Defaults to 'us-east-1'.

llm_aws_access_key_id

AWS access key ID for Amazon Bedrock authentication.

llm_aws_secret_access_key

AWS secret access key for Amazon Bedrock authentication.

llm_aws_session_token

Optional AWS session token for temporary credentials with Amazon Bedrock.

llm_gcp_project

Google Cloud project ID for Google Vertex AI.

llm_gcp_location

Google Cloud region for Vertex AI. Defaults to 'us-central1'.

llm_gcp_service_account_json

Google Cloud service account JSON key for Vertex AI authentication. Pass the entire JSON key file contents as a string.

llm_full_response

Only available on: ultra and above

Return the full LLM response including usage statistics and metadata, rather than just the text content.

llm_output

Only available on: ultra and above

You can provide a structured output type and schema to Urlbox, and we will prompt your LLM to give back that response structure. By passing an llm_schema without this option, it will default to responding with an object specified by your JSON Schema. If you choose Array, the response will be an array of your provided JSON Schema. If you choose enum, you can provide an array of strings as your schema, and your LLM will respond only with a value from that enum.

The available values are:

object array enum

llm_schema

Only available on: ultra and above

This is the JSON schema or string[] provided which we will pass over to your LLM. Your LLM provider will respond with a structured output (if supported by your provider) according to that schema. Please take a look at the various resources on the JSON Schema website for more information on designing and validating a JSON schema.

llm_schema Examples