On-demand Builders
On-demand Builders are serverless functions used to generate web content as needed that’s automatically cached on Netlify’s Edge CDN. They enable you to build pages for your site when a user visits them for the first time and then cache them at the edge for subsequent visits.
Consider serverless functions with the durable cache instead
For better performance and fewer function invocations, consider using serverless functions with the durable
directive instead of On-demand Builders.
Key characteristics of On-demand Builders:
- They are Lambda-compatible serverless functions written in TypeScript or in JavaScript.
- They accept GET requests only.
- They don’t provide access to HTTP headers or query parameters from incoming requests.
- They are protected from repeated invocations by caching generated pages or assets.
- They use the request’s URL path to determine whether to create a new cache object or reuse an existing one. This approach can’t be customized with cache key variations.
- They use the response type to determine whether or not to cache the response.
- They support an optional time to live (TTL) pattern for configuring caching. They don’t support cache control headers.
- They buffer responses. They don’t support response streaming.
- A new site deploy invalidates the cached responses associated with builders in the same deploy context, such as the same Deploy Preview number or the same branch deploy.
On-demand Builders were designed to achieve Distributed Persistent Rendering (DPR) on Netlify. They also currently enable support for Incremental Static Regeneration (ISR) for Next.js sites using Netlify’s Next.js Runtime v4. You can join the conversation about DPR and ISR in our Support Forums.
# Cached responses
Successful responses, redirection responses, and Not Found responses are cached at the edge and served to subsequent requests. That is, any response with a status code of 2xx
(such as 200
or 201
), 301
, 302
, 307
, 308
, or 404
is cached.
If you use Time to Live (TTL), responses with the above status codes invalidate the cache after the TTL expires.
Responses with any other status codes, such as 500
, are not cached.
# Create On-demand Builders
To get started, follow our docs to create a Lambda-compatible function in TypeScript or in JavaScript. Use the synchronous function format, and return a response that can be served as a cached static asset. To convert the function to a builder, use the builder()
method from the @netlify/functions
package.
First, add the package to your site dependencies:
npm install -D @netlify/functions
Then, pass your function as a parameter to the builder()
method and export this method as the handler
from your function file.
import { builder, type Handler } from "@netlify/functions";
const myHandler: Handler = async (event, context) => {
// logic to generate the required content
};
const handler = builder(myHandler);
export { handler };
const { builder } = require("@netlify/functions")
async function handler(event, context) {
// logic to generate the required content
}
exports.handler = builder(handler);
For example, your builder could return a full HTML page or a processed image file. Here’s a “hello world” example:
import { builder, type Handler } from "@netlify/functions";
const myHandler: Handler = async (event, context) => {
return {
statusCode: 200,
headers: {
"Content-Type": "text/html",
},
body: `<!DOCTYPE html> <html> <body> Hello World </body> </html>`,
};
};
const handler = builder(myHandler);
export { handler };
const { builder } = require("@netlify/functions")
async function handler(event, context) {
return {
statusCode: 200,
headers: {
"Content-Type": "text/html",
},
body: `
<!DOCTYPE html>
<html>
<body>
Hello World
</body>
</html>
`,
};
}
exports.handler = builder(handler);
On success, this builder returns a 200
status code and an HTML page with a “Hello World” message. This successful response is cached at the edge and served to subsequent requests.
If the initial invocation of a builder fails with a status code that is not one of the cached response codes, for example because a third-party API is down and the response status is 500
, the erroring response is served but not cached. The next request will retry the builder to get a successful response.
For a working example of using On-demand builders with code to explore, you can visit this example site and its corresponding code repository.
# Time to live (TTL)
On-demand Builders support an optional time to live (TTL) pattern that allows you to set a fixed duration of time after which a cached builder response is invalidated. This allows you to force a refresh of a builder-generated response without a new deploy.
You can set a TTL for your builder by including a ttl
, in seconds, in your response. ttl
has a minimum value of 60.
import { builder, type Handler } from "@netlify/functions";
const originalResponse = {
body: ':thumbsup:',
statusCode: 200,
ttl: 3600,
};
const myHandler: Handler = async (event, context) => {
// logic to generate the required content
return originalResponse;
};
const handler = builder(myHandler);
export { handler };
const { builder } = require("@netlify/functions")
const originalResponse = {
body: ':thumbsup:',
statusCode: 200,
ttl: 3600,
}
async function handler(event, context) {
// logic to generate the required content
return originalResponse
}
exports.handler = builder(handler);
After the TTL has expired, the next request triggers a builder invocation to regenerate the content. The previously cached response is served for 60 seconds after a refresh is triggered. If the refresh is successful and returns with one of the cached response codes, the updated response is cached and served until the next successful refresh. If the refresh fails and returns with any other status code, for example because a third-party API is down, the previously cached successful response is served until the next successful refresh.
# Confirm request type
Incoming requests to On-demand Builders include a x-nf-builder-cache
header that indicates whether the request is an initial request for your content or a background refresh request. This gives you the opportunity to add conditional logic to your builder depending on the request type — for example, you might want to serve a loading page while a user waits for an initial request to complete.
To confirm the type of the incoming request, check the x-nf-builder-cache
header value in the headers
property of the event
object. If the header’s value is miss
, the request is an initial blocking request from the browser that needs to complete before a user can take next steps. If the header’s value is revalidate
, the request is an asynchronous TTL refresh request that runs in the background.
# Environment variables
On-demand Builders have access to environment variables in the runtime environment. If you have the option to set specific scopes for your environment variables, the scope must include Functions to be available to On-demand Builders during runtime.
Learn more about how to set and use environment variables with functions.
# Use On-demand Builders
Similar to other functions, On-demand Builders can be invoked at endpoints relative to the base URL of your site. Builders, however, use a unique builders
path for invocation:
/.netlify/builders/FUNCTION_NAME
.
Builder endpoints can be called directly like regular web URLs or can be executed by redirecting traffic from another URL.
For example, consider a news site with a large amount of infrequently visited, archived content. You can use a builder to generate those less-frequented pages on-demand when a user requests them. The request URL might be something like this:
https://myawesomenews.com/.netlify/builders/my-archive-builder/article/199x-news
The first call to the builder invokes it and returns the generated page, assuming that’s the logic in your function. Subsequent calls return the cached page content until the next deploy or until the time to live (TTL) has elapsed.
In another example, if you want to replace a static image src
with a custom-sized image generated using a builder, you can format an img
tag like this:
<img
alt="kittens."
src="/.netlify/builders/my-image-transformer/file/kittens.jpg/width/640"
/>
The first builder call processes and returns the image, and subsequent calls return the cached image until the next deploy or until the TTL has elapsed.
# Customize paths with redirects
In the example above, you might prefer to use a different URL instead of the default builder endpoint. You can do this by redirecting traffic from a URL to the builder using redirect rules in the _redirects
file or netlify.toml
Netlify configuration file.
Here are some examples with _redirects
file and netlify.toml
syntax.
/images/* /.netlify/builders/my-image-transformer 200
/archive /.netlify/builders/my-archive-builder 200
[[redirects]]
from = "/images/*"
to = "/.netlify/builders/my-image-transformer"
status = 200
[[redirects]]
from = "/archive"
to = "/.netlify/builders/my-archive-builder"
status = 200
The original path is passed to the builder as part of the event object. To continue an example from above, a call to /images/kittens.jpg/width/640
triggers the my-image-transformer
function with the following values passed to it as part of the event
:
{
"path": "/images/kittens.jpg/width/640",
"httpMethod": "GET"
...
}
For more information about the event
parameter and the Lambda-compatible Netlify Functions programming model, refer to our docs on synchronous function format for TypeScript or JavaScript.
# Redirect workaround for query parameters
On-demand Builders currently don’t process or cache query parameters from redirected paths. In most cases, it’s preferable to embed these parameters into the path directly, but in cases where you don’t control the path, you can re-shape it with an additional redirect rule to the path resolution. The first rule must have a 3XX
HTTP status code, as that ensures a new URL for our redirects engine to process the second rule, passing your new path to the builder. Here’s an example in _redirects
syntax:
/generated/image url=:url w=:w /img/:url/:w 301!
/img/:url/:w /.netlify/builders/image 200
# Shadowing
Redirect rules leverage shadowing, which means that they trigger only if there isn’t already an asset at the specified from
path. This can be useful if your prebuilt assets and content generated by your builder share the same path.
If you prefer to trigger the proxy regardless of the presence of a pre-built asset, you can append !
to the status code in the _redirects
file, or add force = true
in netlify.toml
.
# Local development
You can build and test logic like any other function using Netlify Dev. When testing locally, however, caching isn’t available.
# Secure On-demand Builders
You can use the same authentication and authorization strategy as the rest of your Netlify site to secure builders. We recommend using JSON Web Tokens (JWT), roles, and redirect rules for managing granular access. Refer to our docs on role-based access control to learn more.
# Limits
- The maximum response payload size for a builder is 6 MB.
- Pages or assets generated by On-demand Builders are limited to 10,000 per deploy. For a higher limit, please contact sales. If the limit is reached, additional generated pages or assets will have reduced caching and may be regenerated upon subsequent requests.
- There is a 30 second execution time limit that results in a timeout.
# Usage and billing
Requests to builders that return a previously non-cached response count towards your Netlify Functions usage for billing purposes. This includes requests per month and run time per month. If a cached response is available and served, no function invocation is made and no usage is incurred against your Netlify Functions allotment.
On-demand Builders execution doesn’t count towards build minutes.
Did you find this doc useful?
Your feedback helps us improve our docs.