AutoWebPerf provides a flexible and scalable framework for running web performance audits with arbitrary audit tools like WebPageTest and PageSpeedInsights. This library enables developers to collect metrics consistently and store metrics to a preferred data store such as local JSON files, Google Sheets, BigQuery, or an in-house SQL database.
Check out https://web.dev/autowebperf for introduction.
AutoWebPerf takes a list of Tests from an arbitrary data store platform, such as local JSONs, Google Sheets, BigQuery, or a self-hosted SQL database. With the list of Tests, it executes audits based on each Test config, collects metrics from individual data sources into a list of Results.
The process of running an audit through an measurement tool (e.g. WebPageTest) is defined in the individual Gatherer. The logic of reading and writing with a data platform (e.g. local JSON) is implemented in a Connector.
- A library of web audit automation that can be plugged-in to any platforms, like Google Sheets, GCP App Engine, or simply a cron job that writes to JSON file.
- Providing the ability to run recurring tests with customizable frequency (e.g. daily, weekly, monthly, etc), network conditions, and other audit configs, etc.
- Metric gatherers are designed as modules that are decoupled with the output data format and automation logic.
- Connector modules are designed to read Test list and write audit results to
specific data format or platforms. e.g. a connector for CSV files.
(See
src/connectors/csv-connector
for details)
AutoWebPerf serves as a performance audit aggregator that automates the process of performance audit and metrics collection through multiple speed measurement tools including WebPageTest, PageSpeedInsights, and Chrome UX Report. As each individual speed measurement tool provides audit metrics, AutoWebPerf aggregates the results and writes to any preferred data storage platform, such as local JSONs, cloud-based database, or GoogleSheets.
First, clone AWP repo and run npm install:
git clone https://github.com/GoogleChromeLabs/AutoWebPerf.git
npm install
Once finished, check the install by running a single test with the following command:
./awp run examples/tests.json output/results.json
This command uses the example file in examples/tests.json
and returns the results to output/results.json
.
To start recurring tests, you'll need to include a recurring.frequency
property in the test file and set the next trigger in the test file. To setup the next trigger time and to run a one-off test, use this command after adding the recurring.frequency
property to your tests:
./awp recurring examples/tests-recurring.json output/results.json
If this was successful, the trigger time will have updated base on your chosen frequency, and a result would have been written to output/results.json
.
Once the trigger time is correctly set, you can have your tests auto-run on the next triger time with the continue
command:
./awp continue examples/tests-recurring.json output/results.json
This will automatically run each test at the frequency specified. More information can be found below in the "Run recurring tests" section below.
Single URL: To test a single URL through PageSpeedInsights:
./awp run url:https://www.thinkwithgoogle.com/ json:output/results.json
Pick Gatherer: to test a single URL with a specific gatherer like PageSpeedInsights or WebPageTest:
./awp run --gatherers=psi url:https://web.dev json:output/results.json
CSV file: To run tests defined in a CSV file and write results to a JSON file:
./awp run csv:examples/tests.csv json:output/results.json
PageSpeedInsights API: To run PageSpeedInsights tests with an API Key:
PSI_APIKEY=SAMPLE_KEY ./awp run examples/tests.json output/results.json
WebPageTest API: To run WebPageTest tests:
WPT_APIKEY=SAMPLE_KEY ./awp run examples/tests-wpt.json output/results.json
Override vs. Append: To run tests and override existing results in the output file
./awp run examples/tests.json output/results.json --override-results
- WebPageTest - See docs/webpagetest.md for details.
- PageSpeed Insights - See docs/psi.md for details.
- Chrome UX Report API - See docs/cruxapi.md for details.
- Chrome UX Report BigQuery - See docs/cruxbigquery.md for details.
- JSON connector - reads or writes to local JSON files. This is the default connector if a conenctor name is not specified. For example:
./awp run examples/tests.json output/results.json
Alternatively, to specify using the JSON connector for the Tests
path and the Results
path:
./awp run json:/examples/tests.json json:output/results.json
- CSV connector - reads or writes to local CSV files.
To specify using the CSV connector for the
Tests
path and theResults
path:
./awp run csv:/examples/tests.csv csv:output/results.csv
- URL connector - generates just one
Test
with a specific URL for audit. To run an audit with just oneTest
with a specific URL:
./awp run url:https://example.com csv:output/results.csv
Please note that this connector only works with Tests
path, not for the Results
path.
- Google Sheets connector See docs/sheets-connector.md for detailed guidance.
You can run the following anytime for printing CLI usages:
./awp --help
To run tests, you can run the following CLI command with given Tests JSON, like
examples/tests.json
, which contains an array of tests. You can check out the
examples/tests.json
for the data structure of Tests.
./awp run examples/tests.json output/results.json
This will generate the result object(s) in the given path to results.json
.
By default, AWP will use JSON as the default connector for both reading tests
and writing results. Alternatively, you can specify a different connector in the
format of <connector>:<path>
.
E.g. to run tests defined in CSV and write results in JSON:
./awp run csv:examples/tests.csv json:output/results.csv
For some audit platforms like WebPageTest, each test may take a few minutes to fetch actual results. For these type of asynchronous audits, each Result will stay in "Submitted" status. You will need to explicitly retrieve results later.
Run the following to retrieve the final metrics of results in the
results.json
.
./awp retrieve examples/tests.json output/results.json
This will fetch metrics for all audit platforms and update to the Result object
in the output/results.json
. You can check out examples/results.json
for
details in Result objects.
If you'd like to set up recurring tests, you can define the recurring
object
that contains frequency
for that Test.
./awp recurring examples/tests-recurring.json output/results.json
This will generate the Result object in the results.json
and updates the next
trigger time to its original Test object in the tests.json
. E.g. the updated
Test object would look like the following, with the updated nextTriggerTimestamp
.
{
"label": "web.dev",
"url": "https://web.dev",
"recurring": {
"frequency": "Daily",
"nextTriggerTimestamp": 1599692305567,
"activatedFrequency": "Daily"
},
"psi": {
"settings": {
"locale": "en-GB",
"strategy": "mobile"
}
}
}
The nextTriggerTimestamp
will be updated to the next day based on the previous
timestamp. This is to prevent repeated runs with the same Test and to guarantee
that this Test is executed only once per day.
In most Unix-like operating system, you can set up a cron job to run the AWP CLI periodically.
For example, in macOS, you can run the following commands to set up a daily cron job with AWP:
# Edit the cronjob with a text editor.
EDITOR=nano crontab -e
Add the following line to the crontab for a daily run at 12:00 at noon. Note that this is based on the system time where it runs AWP.
0 12 * * * PSI_APIKEY=SAMPLE_KEY cd ~/workspace/awp && ./awp run examples/tests.json csv:output/results-recurring.csv
An extension is a module to assist AWP to run tests with additional process and
computation. For example, budgets
extension is able to add performance budgets
and compute the delta between the targets and the result metrics.
To run with extensions:
./awp run examples/tests.json output/results.json --extensions=budgets
The list of tests is simply an array of Tests objects, like the sample Tests
below. Or check out src/examples/tests.js
for a detailed example of Tests
list.
[{
"label": "Test-1",
"url": "example1.com",
"webpagetest": {
...
}
}, {
"label": "Test-2",
"url": "example2.com",
"psi": {
...
}
}]
Each Test
object defines which audits to run by defining gatherers
property.
For example, the first Test
has a webpagetest
property which defines the
configuration of running a WebPageTest audit. The second Test
has a psi
property that defines how to run PageSpeedInsights audit.
After running tests, a list of Results
is generated like below. Each Result
contains its corresponding metrics to the predefined gatherers
such as
WebPageTest and PageSpeedInsights. See the example below.
[{
"label": "Test-1",
"url": "example1.com",
"webpagetest": {
"metrics": {
FirstContentfulPaint: 900,
...
}
}
}, {
"label": "Test-2",
"url": "example2.com",
"psi": {
"metrics": {
FirstContentfulPaint: 900,
...
}
}
}]
Some conenctors or gatherers may require one or more environmental variables, such as API keys or the path to service account. For example, running with the CrUX API gatherer requires the CrUX API key.
To pass the environmental variables in the CLI, run the command with the regular usage of environment vars:
CRUX_APIKEY=<YOUR_KEY> ./awp run url:https://wev.dev/ json:output/results.json
AWP supports the following audit gatherers. Please check out the corresponding documentations for details.
The WebPageTest gatherer runs Tests through either the public WebPageTest endpoints or a custom private WebPageTest instance.
See docs/webpagetest.md for more details for the usage of WebPageTest gatherer.
The PageSpeed Insights gatherer runs Tests through the public PageSpeed Insights API.
See docs/psi.md for more details for the usage of PSI gatherer.
The CrUX API gatherer collects performance metrics through the Chrome UX Report API.
See docs/cruxapi.md for more details for the usage of CrUX API gatherer.
The CrUX BigQuery gatherer collects performance metrics through the Chrome UX Report with its public Google BigQuery project. Please noet that you would need set up a Google Cloud project in order to query the public BigQuery table.
See docs/cruxbigquery.md for more details for the usage of CrUX API gatherer.
AWP is designed with modules, including modules for running audits with WebPageTest, PageSpeedInsights, or other tools, and modules for reading/writing data from data platforms such as JSON, GoogleSheets or a Cloud service.
In a high-level view, there are three types of modules:
- Gatherer - A Gatherer runs audits and generates metrics.
- Connector - A Connector reads test configs from and writes results to a data platform, such as a local JSON file or with Google Sheets.
- Extension - An Extension adds additional metrics or information, either before or after running audits.
The AWP engine uses two major JavaScript Object structures for running audits and collecting metrics.
- Test - An object that contains the audit configuration for one test task,
such as URL, audit methods, or extension config. You can refer to
examples/tests.json
for an actual Test object. - Result - An object that contains audit configuration, metrics and overall
status. You can refer to
examples/results.json
for an actual Result object.
In order to deal with asynchronous audit tool like WebPageTest, AWP breaks the audit cycle into three steps of actions:
- Run - This action takes a list of
Tests
and generates a list ofResults
objects. - Recurring - Similar to Run, this action takes a list of
Tests
, generates a list ofResults
, and updates nextTriggerTimestamp for each recurringTest
. This action is useful when running with periodic or timer-based tasks such as cron job. - Retrieve - This action scans the list of Results, and collects metrics
when the results are not in
Retrieved
status.
To set up modules and their configurations, an overall AWP Config is required as a JavaScript Object.
AWP Config has the following required properties:
connector
: The name of connector.helper
: A helper for a specific connector, including API Handler and other helper functions, which will be used in gatherers and extensions.dataSources
: An array of audit sources, such aswebpagetest
orpsi
. Each of the data source needs to have a corresponding Gatherer file in thesrc/gatherers
folder.extensions
: An array of extensions. Each extension needs to have a corresponding Extension file insrc/extensions
.
Other optional properties:
verbose
: Whether to print verbose messages.debug
: Whether to print debug messages.
The following config example comes from the examples/awp-config.js
:
{
connector: 'JSON',
helper: 'Node',
dataSources: ['webpagetest'],
json: { // Config for JSON Connector.
tests: 'tests.json',
results: 'results.json',
},
extensions: [
'budgets',
],
budgets: { // Config for Budgets extension.
dataSource: 'webpagetest',
},
verbose: true,
debug: false,
}
With the example config above, it will use JSON
connector which reads and
writes Tests and Results as JSON files. See examples/tests.json
and
examples/results.json
for examples.
In addition to fundamental properties, there are a few additional properties used by modules:
json
property as the configuration for JSONConnector.budgets
property as the configuration for BudgetsExtension
Examples of creating a new instance of AWP:
let awp = new AutoWebPerf({
connector: 'JSON',
helper: 'Node',
dataSources: ['webpagetest'],
extensions: extensions,
json: { // Config for JSON connector.
tests: argv['tests'],
results: argv['results'],
},
verbose: verbose,
debug: debug,
});
To submit all tests:
awp.run();
To submit specific tests using filters: This will run the test which has id=1 and selected=true properties.
awp.run({
filters: ['id="1"', 'selected'],
});
To retrieve all pending results, filtering with status !== "Retrieved".
awp.retrieve({
filters: ['status!=="Retrieved"'],
});
- For more advanced usage of PatternFilter, please check out
src/utils/pattern-filter.js
with more examples.
To run recurring tests:
// This will run the actual audit and update the nextTriggerTimestamp.
awp.recurring();
To run tests with specific extensions:
// This will override the extension list defined in the awpConfig.
awp.run({
extensions: ['budgets']
})
A Gatherer class extends src/gatherers/gatherer.js
and overrides the
following methods:
-
constructor(config, apiHelper, options)
:config
: The config defined in a property with this gatherer's name in the AWP config. Some audit tools like WebPageTest or PageSpeedInsights require API keys. The API key for the gatherer is located in theconfig.apiKey
.options
: Additional settings likeverbose
anddebug
.
-
run(test, options)
:test
: ATest
object for this audit run. The data required for this gatherer (e.g. settings or metadata) will be in the property with the gatherer's name. E.g. the data for WebPageTest will be inwebpagetest
of thisTest
object.options
: Additional settings.
-
retrieve(result, options)
:result
: AResult
object to retrieve metrics with. The data required for this gatherer will be in the property with the gatherer's name. E.g. the data and metrics will be inwebpagetest
of thisResult
object.options
: Additional settings likeverbose
anddebug
.
A Connector class extends src/connectors/connector.js
and overrides the
following methods:
-
constructor(config, apiHandler)
:config
: The config defined in a property with this connector's name in the AWP config.apiHandler
: The API handler instance used for making API calls.
-
getConfig()
: The method to return the Connector's additional config object. This config object depends on where this Connector stores its additional settings including API keys for gatherers. For example, JSONConnector uses thetests.json
and reads additional settings from theconfig
property, including API keys for each gatherers. -
getTestList(options)
: The method to return the list ofTests
as an array. -
updateTestList(newTests, options)
: The method to updateTests
list, given the list of newTests
. -
getResultList(options)
: The method to return the list ofResults
as an array. -
appendResultList(newResults, options)
: The method to append newResults
to the end of the currentResults
list. -
updateResultList(newResults, options)
: The method to update existingResults
in the currentResults
list.
A Extension class extends src/extensions/extension.js
and overrides the
following methods:
constructor(config)
:config
: The config defined in a property with this extension's name in the AWP config.
beforeRun(context)
: The method before executing Run step for aTest
.context.test
: The correspondingTest
object.
afterRun(context)
: The method after executing Run step for aTest
.context.test
: The correspondingTest
object.context.result
: The correspondingResult
object.
beforeAllRuns(context)
: The method before executing Run step.context.tests
: AllTest
objects in this Run.
afterAllRuns(context)
: The method after executing Run step.context.tests
: AllTest
objects in this Run.context.results
: AllResult
objects in this Run.
beforeRetrieve(context)
: The method before executing Retrieve step for aResult
.context.result
: The correspondingResult
object.
afterRetrieve(context)
: The method after executing Retrieve step for aResult
.context.result
: The correspondingResult
object.
beforeAllRetrieves(context)
: The method before executing Retrieve step.context.result
: The correspondingResult
object.
afterAllRetrieves(context)
: The method after executing Retrieve step.context.result
: The correspondingResult
object.
A standard Test
object contains the following properties:
(You can refer to examples/tests.json
for an example.)
selected
: Whether to perform Run for thisTest
.label
: Name of thisTest
.url
: URL to audit.recurring
: Settings for recurring audit.frequency
: The frequency string defined insrc/common/frequency.js
. E.g. 'Daily', 'Weekly' or 'Monthly'.
Gatherer-specific settings will be in their own property with the Gatherer's name in lower case. For example, the settings for WebPageTests will be:
webpagetest
settings
: Setting object contains audit location, connection, etc.metadata
: Metadata object contains WebPageTests's ID, JSON URL, etc.
A standard Result
object contains the following properties:
selected
: Whether to perform Retrieve for thisResult
.id
: Auto-generated unique ID for thisResult
.type
:Single
orRecurring
audit.status
:Submitted
,Retrieved
orError
. Refer tosrc/common/status.js
for details.label
: String label for thisResult
. This label inherits from its originalTest
object.url
: Audited URL.createdTimestamp
: When thisResult
is created.modifiedTimestamp
: When thisResult
is last modified.
All metric names used in AWP are required to follow the names, case
sensitive. See the full list of standardized metrics in src/common/metrics.js
-
Timing metrics
TimeToFirstByte
FirstPaint
FirstMeaningfulPaint
FirstContentfulPaint
VisualComplete
SpeedIndex
DOMContentLoaded
LoadEvent
TimeToInteractive
TotalBlockingTime
FirstCPUIdle
FirstInputDelay
LargestContentfulPaint
-
Resource Size
HTML
Javascript
CSS
Fonts
Images
Videos
-
Resource Count
DOMElements
Connections
Requests
-
Resource Scores
Performance
ProgressiveWebApp
All source codes for major functions are located in src
folder. Files are
organized into the following subfolders:
common
: Common classes and definitions, such as Status, Frequency, Metrics, etc.connectors
: Connector classes.extensions
: Extension classes.gatherers
: Gatherer classes.utils
: Utilities and tools.
Run the following commands to run unit test:
npm test
To run individual test spec, you can install Jest NPM module to your local machine:
npm install -g jest
jest test/some-module.test.js
The Unit Test is based on Jest unit test framework. All
unit tests are located in the ./test
folder, and are organized into its own
corresponding subfolders, as the same structure as in the src
folder.