Frameworks API
Web development frameworks can use the Frameworks API to define how to deploy a site to Netlify. Use the API to provision and configure platform primitives so that developers don’t have to, allowing for a seamless integration with the framework and a zero-config setup.
The API is file-based: as part of the build command, the framework should write files to specific directories, with a well-defined format. Netlify then reads and further processes these files to create a deployment.
This page is for authors and collaborators of web frameworks who are building framework adapters for Netlify. It contains the full reference of the Frameworks API.
# Features
# .netlify/v1/config.json
Use the config.json
file to configure how Netlify builds and deploys a site. It’s a subset of the user-facing file-based configuration (using JSON instead of TOML) and it supports the following properties.
# edge_functions
Accepts edge function declarations with the same properties as the user-facing configuration. While edge functions can be configured in the same file as the function code, this property is useful if you would like to declare multiple edge functions to run on the same path and customize the order they run in.
{
"edge_functions": [
{
"function": "auth",
"path": "/admin"
},
{
"cache": "manual",
"function": "injector",
"path": "/admin"
},
{
"function": "auth",
"path": "/blog/*"
},
{
"function": "rewriter",
"path": "/blog/*"
},
{
"excludedPattern": "/products/things/(.*)",
"function": "highlight",
"pattern": "/products/(.*)"
},
{
"excludedPath": "/img/*",
"function": "common",
"path": "/*"
}
]
}
Entries of the edge_functions
array in the config.json
file only take a single path per edge function. This means that if you want to configure the edge function to run on /path1
and /path2
, you need to create two separate entries. This has the advantage of letting you configure the exact order of precedence of each edge function for a given path.
{
"edge_functions": [
{
"function": "my-framework-edge-function-1",
"path": "/path1"
},
{
"function": "my-framework-edge-function-2",
"path": "/path1"
},
{
"function": "my-framework-edge-function-2",
"path": "/path2"
},
{
"function": "my-framework-edge-function-1",
"path": "/path2"
}
]
}
# functions
Accepts function configuration options, including any property from the inline configuration options. When you define the properties, we prefer snake case — for example, use included_files
instead of includedFiles
.
{
"functions": {
"directory": "myfunctions/",
"included_files": ["files/*.md"]
}
}
Optionally, you can apply these settings to only a subset of a site’s functions. For example, if a framework prefixes the functions it generates with my_framework_
, it can target them within the functions
object.
{
"functions": {
"my_framework_*": {
"included_files": ["files/*.md"]
}
}
}
# headers
Specifies custom headers that Netlify serves with the site when a client makes a request. It follows the same capabilities as the headers
array of the user-facing file-based configuration.
{
"headers": [
{
"for": "/*",
"values": {
"Basic-Auth": "someuser:somepassword anotheruser:anotherpassword",
"X-Frame-Options": "DENY",
"X-XSS-Protection": "1; mode=block",
"cache-control": "max-age=0,no-cache,no-store,must-revalidate"
}
}
]
}
# images
Configures the Netlify Image CDN. You can specify allowed domains for remote image transformations.
The remote_images
property accepts an array of regular expressions, which can represent specific subdomains or directories.
Note that slashes need to be double escaped (once due to the regular expression syntax, and a second time due to the JSON syntax). For example, here’s how you could configure the Image CDN to allow any images under https://example.com/
.
{
"images": {
"remote_images": ["https:\\/\\/example.com\\/.*"]
}
}
# redirects
Specifies redirect and rewrite rules that Netlify evaluates when a client makes a request. It follows the same syntax as the redirects
array of the user-facing file-based configuration.
{
"redirects": [
{
"force": false,
"from": "/old-path",
"status": 301,
"to": "/new-path",
"conditions": {
"Country": ["US"],
"Language": ["en", "es"]
},
"query": {
"path": ":path"
}
},
{
"from": "/news",
"to": "/blog"
}
]
}
These rules are appended to any rules defined by the user in the _redirects
file or the netlify.toml
file. Since Netlify reads rules from top to bottom and processes the first matching rule it finds, this means that the user-defined rules take precedence over rules defined in the Frameworks API.
If you want to change this behavior and ensure that the rules defined by the framework are evaluated first, you can use the redirects!
property instead (with an exclamation mark). Please note that this takes precedence over any redirect rules defined by framework users, so use it sparingly and make sure to communicate to your framework users how this might affect their workflows.
{
"redirects!": [
{
"from": "/some-path",
"to": "/new-path"
}
]
}
# .netlify/v1/blobs
Upload blobs to the deploy-specific store. This is useful if you want the ability to generate files at build time and then modify them at runtime throughout the lifecycle of the deploy.
Our build system traverses the .netlify/v1/blobs/deploy
directory and generates a blob for each blob
file it finds. The blob key is the path of the file relative to the .netlify/v1/blobs/deploy
directory (without a leading slash).
.netlify/
└── v1/
└── blobs/
└── deploy/
├── example.com/
│ └── blob
└── netlify.com/
├── blob
└── blob.meta.json
For example, the directory tree above would generate two blobs:
- a blob with the key
example.com
, holding the contents of the file at.netlify/v1/blobs/deploy/example.com/blob
- a blob with the key
netlify.com
, holding the contents of the file at.netlify/v1/blobs/deploy/netlify.com/blob
Optionally, you can include a blob.meta.json
file that contains an object with arbitrary metadata, encoded as JSON, which you can then retrieve with the getMetadata
and getWithMetadata
client methods.
Root key
You can’t place a blob
file directly under the .netlify/v1/blobs/deploy
directory, because that would lead to an empty string as the relative path, which isn’t a valid blob key. If the concept of a root key exists in your specific use case (for example, naming keys after URL paths), you can place all entries under another sub-directory, like my-cache
, and treat my-cache
as /
and my-cache/child
as /child
.
Let’s imagine that your framework implements a cache for HTTP requests. You could use Netlify Blobs to store cached responses, and make your generated functions and edge functions check whether there’s a blob for a given URL before making an HTTP call.
// .netlify/v1/functions/my-framework-cache.ts
import { getDeployStore } from "@netlify/blobs";
import type { Config, Context } from "@netlify/functions";
export default async (req: Request, context: Context) => {
const { domain } = context.params;
const cache = getDeployStore();
const cachedResponse = await cache.get(domain, { type: "stream" });
if (cachedResponse !== null) {
return new Response(cachedResponse);
}
const response = await fetch(`https://${domain}`);
const body = await response.blob();
await cache.set(domain, body);
return response;
}
export const config: Config = {
// Accepts requests on paths like "/cache/example.com".
path: "/cache/:domain"
}
To pre-warm the cache with a response for https://example.com
, you could fetch it at build time and write the response to a file at .netlify/v1/blobs/deploy/example.com/blob
.
Additionally, you could modify the example above to also persist the headers of cached responses. To write a metadata object for the same example.com
key, write a JSON object to .netlify/v1/blobs/deploy/example.com/blob.meta.json
with a payload like:
{
"headers": {
"Content-Type": "text/html; charset=UTF-8",
"Date": "Wed, 12 Jun 2024 09:14:11 GMT"
}
}
# .netlify/v1/edge-functions
Generate edge functions for a site and configure the URL paths on which they run.
Edge Functions let you intercept requests at the very beginning of the request chain, and intercept responses just before they are delivered to the client. In both cases, you have the option to modify the payloads in any way you like.
For example, you could generate an edge function that intercepts requests for JSON (by inspecting the Accept
header) and rewrite them to a path that implements your framework’s API route. If a client requests https://example.com/products/123
and asks for a JSON response, the edge function would serve the contents of the https://example.com/api/products/123
path.
To create this edge function, your framework would write the code below to a file at .netlify/v1/edge-functions/my-framework-api-route-handler.ts
.
import type { Config, Context } from "@netlify/edge-functions";
export default async (req: Request, context: Context) => {
if (req.headers.get("accept") === "application/json") {
const { pathname } = new URL(req.url)
return new URL(`/api${pathname}`, req.url)
}
}
export const config: Config = {
// Configures the paths on which the edge function runs.
// The value below lets you intercept requests for any path.
path: "/*",
// Sometimes it's useful to exclude certain paths. In our
// example, we don't need to intercept requests that are
// already targeting an `/api/` path.
excludedPath: ["/api/*"]
};
# Configuration options
To configure generated edge functions from within the function file, export a config
object. It accepts the following properties:
path
(string
orstring[]
): The URL path or set of paths for which the edge function is invoked. It accepts a string for a single path, or an array of strings for multiple paths. It supports wildcards and named parameters using theURLPattern
web standard syntax.excludedPath
(string
orstring[]
): A URL path or set of paths for which the edge function is never invoked. It accepts a string for a single path, or an array of strings for multiple paths. It supports wildcards and named parameters using theURLPattern
web standard syntax.method
(string
orstring[]
): A method or set of methods for which the edge function is invoked. Supported HTTP methods are GET, POST, PUT, PATCH, DELETE, and OPTIONS. If not specified, the edge function is invoked for all methods.cache
(string
): Opts in to response caching.name
(string
): A human-friendly name for the edge function. It is shown instead of the filename in build logs and in the UI.generator
(string
): Lets you tag an edge function with the name and the version of your framework. While this has no direct effect on any user-facing areas, it lets us internally correlate requests for this edge function with a specific framework version to identify and troubleshoot possible issues with a release.onError
(string
): Defines the behavior for when the edge function terminates unexpectedly. You can choose to serve a generic error page (fail
, the default value), skip the erroring edge function and continue the request chain (bypass
), or provide a URL path that requests are rewritten to in the event of an error (for example,/my-framework-error-page
).rateLimit
(object
): Sets custom rate limiting rules for the edge function.
Static values only
The config
object must be defined in the edge function’s main file and it must not be renamed or re-exported. It can only use static values, which means that using constants or variables is not allowed.
# Bundling
When an edge function is created, either directly by a user or by a framework using the Frameworks API, Netlify automatically handles the bundling process. This involves collecting all the edge function’s dependencies, including local imports and third-party modules, into a single file.
Optionally, you can choose to handle this process yourself, using tools like esbuild or Vite to generate a self-contained bundle with all the dependencies inlined. When you do this, the Netlify bundling process is essentially a no-op, and we use your generated bundle as is.
There are some things to consider if you choose to do this:
- It’s important to note that edge functions run on Deno, not Node.js. This has the following implications:
- You shouldn’t rely on Node.js built-ins.
- You should always generate ESM code (not CommonJS) with the latest syntax.
- If you’re bundling your code using esbuild, set the
platform
property toneutral
and settarget
toesnext
. - For Vite, set
target
towebworker
and mark all Node.js built-ins as external.
# Import maps
You can customize the resolution of module specifiers in edge functions using import maps. To do this, place a file following the import map syntax at .netlify/v1/edge-functions/import_map.json
.
For example, the file below would let you rewrite the bare specifier html-rewriter
to a specific remote URL, and point utils
to a local directory so you can import utils/ab-test.ts
instead of ./nested/directory/utils/ab-test.ts
.
{
"imports": {
"html-rewriter": "https://ghuc.cc/worker-tools/html-rewriter/index.ts",
"utils/": "./nested/directory/utils"
}
}
# .netlify/v1/functions
Generate serverless functions for a site and configure the URL paths on which they run. It accepts the same naming conventions and uses the same syntax as user-defined functions.
Functions are a way to dynamically render parts of your application, like API routes. You can also use them to power your framework’s server-side rendering capabilities.
For example, you could generate an SSR function by writing the following code to .netlify/v1/functions/my-framework-ssr.ts
.
import type { Config, Context } from "@netlify/functions";
import { renderPage } from "./your-framework-code.ts";
export default async (req: Request, context: Context) => {
const html = await renderPage(req);
return new Response(html, {
headers: {
"content-type": "text/html"
}
});
};
export const config: Config = {
path: "/*"
};
# Configuration options
Generated functions can be configured from within the function file using an exported config
object. It accepts the following properties:
path
(string
orstring[]
): The URL path or set of paths for which the function is invoked. It supports wildcards and named parameters using theURLPattern
web standard syntax.preferStatic
(boolean
): By default, functions run for any requests to the configured paths, regardless of whether or not they also match static files. To prevent the function from shadowing files on the CDN, setpreferStatic: true
.displayName
(string
): A human-friendly name for the function. It is shown instead of the filename in build logs and in the UI.generator
(string
): Lets you tag a function with the name and the version of your framework. While this has no direct effect on any user-facing areas, it lets us internally correlate requests for this function with a specific framework version to identify and troubleshoot possible issues with a release.includedFiles
(string[]
): List of additional paths to include in the function bundle. Although Netlify includes statically referenced files (likerequire("./some-file.js")
) by default, you can specify additional files or directories and reference them dynamically in function code. You can use*
to match any character or prefix an entry with!
to exclude files. Paths are absolute paths relative to the base directory.nodeBundler
(string
): Changes the default bundling mechanism used to collect all function dependencies (including local files and third-party modules) into a deployable package. You can set it toesbuild
if you want to use esbuild to bundle everything into a single file, which usually leads to a smaller payload; you can set it tonone
if you want to use your own bundling process.externalNodeModules
(string[]
): When settingnodeBundler
toesbuild
, this property lets you provide a list of npm packages that are marked as external, which means their original structure is kept in anode_modules
directory as opposed to being inlined with the function code.ignoredNodeModules
(string[]
): When settingnodeBundler
toesbuild
, this property lets you provide a list of npm packages that are marked as external but whose source is not preserved in thenode_modules
directory. This should only be used for packages that you know for a fact are not used at runtime, otherwise it can lead to an error when the function is invoked.nodeVersion
(string
): Sets the major version of Node.js to run the function with. It takes values like20
to use Node.js 20.x, for example.schedule
(string
): When defined, turns the function into a scheduled function and defines the schedule on which it runs.rateLimit
(object
): Sets custom rate limiting rules for the function.
Static values only
The config
object must be defined in the function’s main file and it must not be renamed or re-exported. It can only use static values, which means that using constants or variables is not allowed.
# Bundling
When a function is created, either directly by a user or by a framework using the Frameworks API, Netlify automatically handles the bundling process by default. This means collecting all the function’s dependencies, including local files and third-party modules, into a deployable package.
Optionally, you can choose to take control over this process. This may be useful if your framework already includes its own bundling pipeline. To opt out of the default bundling mechanism, you must set two configuration properties (nodeBundler
and includedFiles
) in the function’s config
object.
import type { Config, Context } from "@netlify/functions";
export default async (req: Request, context: Context) => {
return new Response("Hello world");
}
export const config: Config = {
// Setting `nodeBundler` to `none` lets you opt out of the
// automatic bundling process.
nodeBundler: "none",
// When using a custom bundling process, you must provide the
// full list of files to include in the deployable package.
// You can use the `includedFiles` property, which supports globs.
includedFiles: ["my-framework-bundle-output/**"]
};
# Merging configurations
All Netlify platform primitives can be provisioned and configured directly by developers — for example, a developer can create a redirect rule by writing a _redirects
file or deploy a function by writing a file to the functions directory.
The Frameworks API was designed to extend these configuration options, not override them. This means that any user-defined configurations are respected, and the following rules apply when those configurations are merged with the Frameworks API:
- For properties that are “additive” (like the allowed domains for the Netlify Image CDN or the list of serverless functions), any entries from the Frameworks API are appended to the ones defined by the user; if order is relevant, user-defined configuration takes precedence
- For properties that are mutually-exclusive (like the functions directory), any user-defined configuration takes precedence
Avoid collisions with user-defined functions
To help make sure your framework-defined functions aren’t overridden by user-defined functions, we recommend that you include your framework’s name in your function names. For example,
my-framework-ssr.ts
.
# Versioning
The filesystem is the only interface to the Frameworks API; there are no client libraries for programmatic access. This lowers the barrier of entry to the absolute minimum, allowing any framework, of any technology stack, to take advantage of the Netlify platform by writing files to disk.
But even without pulling in a library, there is still an intrinsic dependency between the Frameworks API and any framework that uses it. As we update the API surface with new features or changes to existing ones, we will always do so in a backward-compatible way, to ensure that all framework versions continue to operate without any disruption.
To this end, all features have a version identifier in their file paths (for example, .netlify/v1/config.json
). A breaking change to that feature would be released as .netlify/v2/config.json
, with no changes to the previous path, giving framework authors the choice to update their implementation at their own schedule.
Did you find this doc useful?
Your feedback helps us improve our docs.