Project Configuration

A project's configuration is constructed by Nx from three sources:

  1. Tasks inferred by Nx plugins from tooling configuration
  2. Workspace targetDefaults defined in the nx.json file
  3. Individual project level configuration files (package.json and project.json)

Each source will overwrite the previous source. That means targetDefaults will overwrite inferred tasks and project level configuration will overwrite both targetDefaults and inferred tasks. The combined project configuration can be viewed in the project details view by using Nx Console in your IDE or by running:

โฏ

nx show project myproject --web

Project Details View

demo

Root: packages/demo

Type: Application

Targets

  • dev

    vite dev

  • build

    vite build

The project details view also shows where each setting is defined so that you know where to change it.

Project Level Configuration Files

If you need to edit your project settings or modify an inferred task, you can do so in either package.json or project.json files. The examples on this page show both styles, and the only functional difference is that tasks that use executors must be defined in a project.json. Nx merges the two files to get each project's configuration. The full machine readable schema is available on GitHub.

The following configuration creates build and test targets for Nx.

package.json
1{ 2 "name": "mylib", 3 "scripts": { 4 "test": "jest", 5 "build": "tsc -p tsconfig.lib.json" // the actual command here is arbitrary 6 } 7} 8

You can invoke nx build mylib or nx test mylib without any extra configuration.

Below are some more complete examples of project configuration files. For a more intuitive understanding of the roles of each option, you can highlight the options in the excerpt below that relate to different categories. Orchestration settings control the way Nx runs tasks. Execution settings control the actual task that is run. Caching settings control when Nx caches a task and what is actually cached.

package.json
1{ 2 "name": "mylib", 3 "scripts": { 4 "test": "jest", 5 "build": "tsc -p tsconfig.lib.json", // the actual command here is arbitrary 6 "ignored": "exit 1" 7 }, 8 "nx": { 9 "namedInputs": { 10 "default": ["{projectRoot}/**/*"], 11 "production": ["!{projectRoot}/**/*.spec.tsx"] 12 }, 13 "targets": { 14 "build": { 15 "inputs": ["production", "^production"], 16 "outputs": ["{workspaceRoot}/dist/libs/mylib"], 17 "dependsOn": ["^build"] 18 }, 19 "test": { 20 "inputs": ["default", "^production"], 21 "outputs": [], 22 "dependsOn": ["build"] 23 } 24 }, 25 "includedScripts": ["test", "build"] // If you want to limit the scripts Nx sees, you can specify a list here. 26 } 27} 28

Task Definitions (Targets)

A large portion of project configuration is related to defining the tasks for the project. In addition, to defining what the task actually does, a task definition also has properties that define the way that Nx should run that task. Those properties are described in detail below.

Cache

In Nx 17 and higher, caching is configured by specifying "cache": true in a target's configuration. This will tell Nx that it's ok to cache the results of a given target. For instance, if you have a target that runs tests, you can specify "cache": true in the target default configuration for test and Nx will cache the results of running tests.

project.json
1{ 2 "targets": { 3 "test": { 4 "cache": true 5 } 6 } 7} 8
Per Project Caching + Distribution

If you are using distributed task execution and disable caching for a given target, you will not be able to use distributed task execution for that target. This is because distributed task execution requires caching to be enabled. This means that the target you have disabled caching for, and any targets which depend on that target will fail the pipeline if you try to run them with Nx Agents enabled.

Inputs and Named Inputs

Each cacheable task needs to define inputs which determine whether the task outputs can be retrieved from the cache or the task needs to be re-run. The namedInputs defined in nx.json or project level configuration are sets of reusable input definitions.

A typical set of inputs may look like this:

1{ 2 "namedInputs": { 3 "default": ["{projectRoot}/**/*"], // every file in the project 4 "production": ["default", "!{projectRoot}/**/*.spec.tsx"] // except test files 5 }, 6 "targets": { 7 "build": { 8 "inputs": [ 9 "production", // this project's production files 10 { "externalDependencies": ["vite"] } // the version of the vite package 11 ] 12 } 13 } 14} 15

Outputs

Targets may define outputs to tell Nx where the target is going to create file artifacts that Nx should cache. "outputs": ["{workspaceRoot}/dist/libs/mylib"] tells Nx where the build target is going to create file artifacts.

This configuration is usually not needed. Nx comes with reasonable defaults (imported in nx.json) which implement the configuration above.

Specifically, by default, the following locations are cached for builds:

  • {workspaceRoot}/dist/{projectRoot},
  • {projectRoot}/build,
  • {projectRoot}/dist,
  • {projectRoot}/public

Read the configure outputs for task caching recipe for helpful tips for setting outputs.

Basic Example

Usually, a target writes to a specific directory or a file. The following instructs Nx to cache dist/libs/mylib and build/libs/mylib/main.js:

1{ 2 "targets": { 3 "build": { 4 "outputs": [ 5 "{workspaceRoot}/dist/libs/mylib", 6 "{workspaceRoot}/build/libs/mylib/main.js" 7 ] 8 } 9 } 10} 11

Specifying Globs

Sometimes, multiple targets might write to the same directory. When possible it is recommended to direct these targets into separate directories.

1{ 2 "targets": { 3 "build-js": { 4 "outputs": ["{workspaceRoot}/dist/libs/mylib/js"] 5 }, 6 "build-css": { 7 "outputs": ["{workspaceRoot}/dist/libs/mylib/css"] 8 } 9 } 10} 11

But if the above is not possible, globs (parsed by the GlobSet Rust library) can be specified as outputs to only cache a set of files rather than the whole directory.

1{ 2 "targets": { 3 "build-js": { 4 "outputs": ["{workspaceRoot}/dist/libs/mylib/**/*.{js,map}"] 5 }, 6 "build-css": { 7 "outputs": ["{workspaceRoot}/dist/libs/mylib/**/*.css"] 8 } 9 } 10} 11

More advanced patterns can be used to exclude files and folders in a single line

1{ 2 "targets": { 3 "build-js": { 4 "outputs": ["{workspaceRoot}/dist/libs/!(cache|.next)/**/*.{js,map}"] 5 }, 6 "build-css": { 7 "outputs": ["{workspaceRoot}/dist/libs/mylib/**/!(secondary).css"] 8 } 9 } 10} 11

dependsOn

Targets can depend on other targets. This is the relevant portion of the configuration file:

1{ 2 "targets": { 3 "build": { 4 "dependsOn": ["^build"] 5 }, 6 "test": { 7 "dependsOn": ["build"] 8 } 9 } 10} 11

A common scenario is having to build dependencies of a project first before building the project. This is what the "dependsOn": ["^build"] property of the build target configures. It tells Nx that before it can build mylib it needs to make sure that mylib's dependencies are built as well. This doesn't mean Nx is going to rerun those builds. If the right artifacts are already in the right place, Nx will do nothing. If they aren't in the right place, but they are available in the cache, Nx will retrieve them from the cache.

Another common scenario is for a target to depend on another target of the same project. For instance, "dependsOn": ["build"] of the test target tells Nx that before it can test mylib it needs to make sure that mylib is built, which will result in mylib's dependencies being built as well.

You can also express task dependencies with an object syntax:

1{ 2 "targets": { 3 "build": { 4 "dependsOn": [ 5 { 6 "target": "build", // target name 7 "params": "ignore" // "forward" or "ignore", defaults to "ignore" 8 } 9 ] 10 } 11 } 12} 13

Examples

You can write the shorthand configuration above in the object syntax like this:

1{ 2 "targets": { 3 "build": { 4 "dependsOn": [{ "dependencies": true, "target": "build" }] // Run build on my dependencies first 5 }, 6 "test": { 7 "dependsOn": [{ "target": "build" }] // Run build on myself first 8 } 9 } 10} 11

With the expanded syntax, you also have a third option available to configure how to handle the params passed to the target. You can either forward them or you can ignore them (default).

1{ 2 "targets": { 3 "build": { 4 // forward params passed to this target to the dependency targets 5 "dependsOn": [ 6 { "projects": "{dependencies}", "target": "build", "params": "forward" } 7 ] 8 }, 9 "test": { 10 // ignore params passed to this target, won't be forwarded to the dependency targets 11 "dependsOn": [ 12 { "projects": "{dependencies}", "target": "build", "params": "ignore" } 13 ] 14 }, 15 "lint": { 16 // ignore params passed to this target, won't be forwarded to the dependency targets 17 "dependsOn": [{ "projects": "{dependencies}", "target": "build" }] 18 } 19 } 20} 21

This also works when defining a relation for the target of the project itself using "projects": "self":

1{ 2 "targets": { 3 "build": { 4 // forward params passed to this target to the project target 5 "dependsOn": [{ "target": "pre-build", "params": "forward" }] 6 } 7 } 8} 9

Additionally, when using the expanded object syntax, you can specify individual projects in version 16 or greater.

1{ 2 "targets": { 3 "build": { 4 // Run is-even:pre-build and is-odd:pre-build before this target 5 "dependsOn": [ 6 { "projects": ["is-even", "is-odd"], "target": "pre-build" } 7 ] 8 } 9 } 10} 11

This configuration is usually not needed. Nx comes with reasonable defaults (imported in nx.json) which implement the configuration above.

Project Metadata

The following properties describe the project as a whole.

tags

You can annotate your projects with tags as follows:

package.json
1{ 2 "name": "mylib", 3 "nx": { 4 "tags": ["scope:myteam"] 5 } 6} 7

You can configure lint rules using these tags to, for instance, ensure that libraries belonging to myteam are not depended on by libraries belong to theirteam.

implicitDependencies

Nx uses powerful source-code analysis to figure out your workspace's project graph. Some dependencies cannot be deduced statically, so you can set them manually like this. The implicitDependencies property is parsed with the minimatch library, so you can review that syntax for more advanced use cases.

package.json
1{ 2 "name": "mylib", 3 "nx": { 4 "implicitDependencies": ["anotherlib"] 5 } 6} 7

You can also remove a dependency as follows:

package.json
1{ 2 "name": "mylib", 3 "nx": { 4 "implicitDependencies": ["!anotherlib"] # regardless of what Nx thinks, "mylib" doesn't depend on "anotherlib" 5 } 6} 7

An implicit dependency could also be a glob pattern:

package.json
1{ 2 "name": "mylib", 3 "nx": { 4 "implicitDependencies": ["shop-*"] # "mylib" depends on all projects beginning with "shop-" 5 } 6} 7

Including package.json files as projects in the graph

Any package.json file that is referenced by the workspaces property in the root package.json file will be included as a project in the graph. If you are using Lerna, projects defined in lerna.json will be included. If you are using pnpm, projects defined in pnpm-workspace.yml will be included.

If you want to ignore a particular package.json file, exclude it from those tools. For example, you can add !packages/myproject to the workspaces property.

Ignoring package.json scripts

Nx merges package.json scripts with any targets defined in project.json. If you only wish for some scripts to be used as Nx targets, you can specify them in the includedScripts property of the project's package.json.

1{ 2 "name": "my-library", 3 "version": "0.0.1", 4 "scripts": { 5 "build": "tsc", 6 "postinstall": "node ./tasks/postinstall" 7 }, 8 "nx": { 9 "includedScripts": ["build"] 10 } 11} 12