Most things can be controlled via environment variables. Some things are specific to the client and some are specific to the builder etc.
This document attempts to describe each environment variable.
This is a general OS, optional, environment variable and it's needed if you want the ability to go from viewing a rendered document to opening the source in your preferred editor. It's needed for the "Edit this page in your editor" button to work.
For the builder, a lot of environment variables can be overridden with CLI arguments.
The general rule is that environment variables specific to the builder are all
prefixed with CONTENT_
. E.g. CONTENT_ROOT
Default: ../content/files
Path to the content files, cloned from https://github.com/mdn/content.
Default: ../translated-content/files
Path to the translated content files, cloned from https://github.com/mdn/translated-content.
Default: ../mdn-contributor-spotlight/contributors
Path to the contributor spotlight content, cloned from https://github.com/mdn/mdn-contributor-spotlight.
Default: `` (meaning, none)
Example: web/css,web/html
Applicable if you run batches of builds but want to limit it to only the folders you care about. When doing a batch build, it can be very time-consuming so just doing one or two sub-folders will speed things up.
Default: []
A comma or newline separated list of file paths. Can be absolute or relative to
the CONTENT_ROOT
.
Default: client/build
Location into which things should be built.
Default: *:warn
To set this you need to know what all the possible flaw identifiers are called.
The complete set can be found in content/src/constants.js
in
VALID_FLAW_CHECKS
. And the level values are error
, warn
, or ignore
.
The value of this environment variable is a comma separated string. Any identifier or level value that is not recognized will throw an error.
You can mix and match like this for example:
macros:error, broken_links: warn, bad_bcd_queries: ignore
Anything not mentioned defaults to ignore
, so the above example is
equivalent to:
macros:error, broken_links: warn
When a flaw is discovered, if the level is error
it will, halt the build and
throw an error. It will halt on the first flaw error encountered. If the level
is warn
it will inject the flaw message into the built index.json
file which
you can view when rendering the document on http://localhost:3000/
.
Default: false
Whether fixable flaws should be fixed when building documents.
Default: false
When set to true (with the BUILD_FIX_FLAWS
flag) it will only print out
information about fixable flaws instead of actually fixing it on disk.
Default: false
Default: https://mdn.mozillademos.org
When generating live samples <iframe>
tags, the src
attribute gets this set
as a prefix. The ultimate reason why it's meant to be different is because the
security of the iframe
's content has not been audited as carefully as the rest
of the site.
When doing local development, it's recommended to set this to
http://localhost:5042
in your personal .env
.
Default: https://interactive-examples.mdn.mozilla.net
The base URL used in the Interactive Example iframes.
Default: ''
If set, the rendered HTML will have a Google Analytics snippet. For example, to
test use: export BUILD_GOOGLE_ANALYTICS_ACCOUNT=UA-00000000-0
. By default it's
disabled (empty string).
Default: false
If true, and when BUILD_GOOGLE_ANALYTICS_ACCOUNT
is truthy, when it injects
the Google Analytics script tag it will use
<script src="https://www.google-analytics.com/analytics_debug.js"></script>
instead which triggers additional console logging which is useful for
developers.
Default: false
This exists so we can forcibly always include
<meta name="robots" content="noindex, nofollow">
into the HTML no matter what.
For example, on our stage or dev builds, we never want robots.
The only place where we want robots is in prod. That's explicitly always set in
prod-build.yml
.
We use this to make absolutely sure that no dev or stage build ever gets into
the Google index. Thankfully we always used a canonical URL
(<link rel="canonical" href="https://developer.mozilla.org/$uri">
) as a
"second line of defense" for dev/stage URLs that are public.
Default: https://hacks.mozilla.org/feed/
Which RSS feed URL to parse for displaying feed entries on the home page.
Default: 5
How many RSS feed entries to display on the home page.
Default: broken_links
Flaw types to be fixed when running fix-flaws
.
Default: 5042
Usually the server
module is started with foreman
(the nf
command) and
this is the default port.
Default: 8080
This is the port for the WebSocket server, which is started when you run
yarn start
.
Default ../client/build
If you want to serve static files some a completely different directory.
NOTE! Due to a quirk of how we build the client, anything REACT_APP_*
environment variable that you want to be available in the production-grade built
JS code that gets used when use localhost:5042
can not only be set in the root
/.env
file. Either use export REACT_APP_*=...
or write it once in /.env
and once in /client/.env
.
Default: localhost
Suggested value: HOST=localhost.org
For browser cookies to work as expected, you need to use the same host name in
Yari and in Kuma. The port numbers can be different. That means that any cookies
you picked up (e.g. sessionid
) over on http://localhost.org:8000
will be
automatically included in XHR calls on http://localhost.org:3000
.
Note that even if you set this, you can still continue to use
http://localhost:3000
.
Default: https://bcd.developer.allizom.org
The base URL (without trailing slash) used to fetch data for BCD tables.
If you want to use local BCD data, set REACT_APP_BCD_BASE_URL=""
.
Default: not set
When doing local development in Yari, the "Sign in" URL depends on this. If
you're running the dev server (i.e. localhost:3000
or localhost.org:3000
)
the link to sign in with Kuma needs this to be set.
The suggested value is to set this to set
REACT_APP_KUMA_HOST=localhost.org:8000
and now sign in links go to that host
name instead. That means you can log in from Yari with a single click.
Default: false
This removes sign-in and whoami
XHR fetching. Useful when using Yari purely
for content editing as authentication is then not required.
Default: NODE_ENV==='development'
Basically, these are the optional, lazy-loaded features of the app that only make sense when you're working on authoring the content. For example the Toolbar bar appears based on this.
It defaults to NODE_ENV==='development'
if not set which means that it's
enable by default when doing development with the localhost:3000
dev server.
Default: localhost, localhost.org, 127.0.0.1
Only applicable if REACT_APP_CRUD_MODE
is truthy. Essentially you can disable
certain CRUD mode features depending on the hostname you use. So if you built
the static assets (the React code) with REACT_APP_CRUD_MODE=true
it might
disable certain features if you use a window.location.hostname
that is not
in this list.
The use case for this is when you build the site in a pull request and want
flaws to appear but without the "Fix fixable flaws" link or the "Open in your
editor" button. We use this for previewing PR builds on the content site. Those
pages are built with flaw detection set to warn, but since you might be viewing
the pages on a remote domain (e.g. pr123.dev.content.mozit.cloud
) it doesn't
make sense to present the "Fix fixable flaws" button for example.
Default: NODE_ENV==='development'
Determines if the MDN++ SPA should be reachable or not.
Default: United States
If the /api/v1/whoami
does not include a geo.country
value, fall back on
this. Setting this allows you to pretend the XHR request to /api/v1/whoami
included this value for geo.country
.
Default: "dev"
- This routes glean analytics to either prod/staging in the backend. See Glean docs
Default: false
- Setting this to true logs glean pings to the console to help debug.
Default: false
- Disables/Enables glean upload. This should be set to
true
in prod and staging to send glean telemetry to the server- Be aware of flipping this between true/false as any persisted metrics, events and pings (other than first_run_date and first_run_hour) are cleared. More info