Playwright reporter configuration

Every setting is an option of the reporter in playwright.config, and most have a PROBARA_* environment variable too. Keep the token in the environment, the project in the config (or the environment), and set the rest where it fits your pipeline.

Quick path

1. Create an app token from the Playwright card in Workspace › Integrations (get started) and store it as a CI secret named PROBARA_API_TOKEN.
2. Set the Project ID: projectId in the reporter options, or PROBARA_PROJECT. It is a capital letter followed by capital letters or digits (SHOP, E2E); any other value turns reporting off with an error.
3. Leave the rest to its default. The run is named after the CI build, and its branch, commit and build URL come from the CI.

playwright.config.ts

import { defineConfig } from '@playwright/test';

export default defineConfig({
  reporter: [['list'], ['@probara/playwright-reporter', { projectId: 'SHOP' }]],
});

apiToken exists for code that reads the token from its own secret store. Do not write the token into playwright.config: the file is committed.

Sources and precedence

The first source that sets a value wins:

Order	Source	Example
1	An option in playwright.config	{ run: { name: 'Release 2.4' } }
2	A PROBARA_* variable	PROBARA_RUN_NAME=Nightly
3	CI detection	CI #42 from the workflow and run number
4	The default	Automated run 2026-09-29 14:05 UTC

• An option set to undefined never hides its variable, and a blank variable counts as unset. To let a variable override an option, leave the option out (or pass process.env.X ?? default).
• Booleans accept true, 1, yes, on and false, 0, no, off. Lists are comma-separated in variables (PROBARA_RUN_TAGS=nightly,smoke), and records are key=value pairs (PROBARA_RUN_ULIDS=WEB=<ulid>,API=<ulid>).
• CI detection only fills the run’s name and its source (branch, commit, build URL, and in GitHub Actions the workflow run that groups parallel jobs): see what each CI fills in.

Options and variables can be mixed. This config sets the project and the tags, and the pipeline adds the environment with PROBARA_ENVIRONMENT=staging:

reporter: [
  ['list'],
  ['@probara/playwright-reporter', { projectId: 'SHOP', run: { tags: ['e2e', 'nightly'] } }],
],

Options

Option	Variable	Type	Default	What it does
enabled	PROBARA_ENABLED	boolean	on	false turns reporting off: nothing is sent, and nothing is logged
apiToken	PROBARA_API_TOKEN	string	none (required)	The app token. Keep it in the environment
projectId	PROBARA_PROJECT	string	none (required)	The Project ID (capital letters and digits), such as SHOP
baseUrl	PROBARA_BASE_URL	string	https://app.probara.net	The Probara host to report to (base URL)
run.ulid	PROBARA_RUN_ULID	string	none: a new run	Report into an existing run (sharding)
run.name	PROBARA_RUN_NAME	string	the CI build name, else Automated run <date>	Name of a new run
run.description	PROBARA_RUN_DESCRIPTION	string	none	Description of a new run
run.environment	PROBARA_ENVIRONMENT	string	none	Environment of a new run by name, created when none matches (run options)
run.environmentId	PROBARA_ENVIRONMENT_ID	string	none	Environment of a new run by ULID
run.milestone	PROBARA_MILESTONE	string	none	Milestone of a new run by display id (M-3) or name
run.milestoneId	PROBARA_MILESTONE_ID	string	none	Milestone of a new run by ULID
run.plan	PROBARA_PLAN	string	none	Test plan of a new run by display id (PLAN-2) or name: the run starts with its cases
run.configurations	PROBARA_CONFIGURATIONS	{ group, name }[]	none	Configurations of a new run by name (Browser=Chrome,OS=Linux), each group once
run.configurationUlids	PROBARA_CONFIGURATION_ULIDS	string[]	none	Configurations of a new run by ULID
run.tags	PROBARA_RUN_TAGS	string[]	none	Tags of a new run
run.ulids	PROBARA_RUN_ULIDS	Record<string, string>	none	The run to reuse in each project (WEB=<ulid>,API=<ulid>) (several projects)
projects	PROBARA_PROJECTS	string[]	none	Other projects results may go to (WEB,API) (several projects)
source	—	false | object (GitHub fields below)	detected from the CI	false sends no source; an object overrides the detected fields one by one
source.branch	PROBARA_BRANCH	string	detected from the CI	Branch of the run source
source.commit	PROBARA_COMMIT	string	detected from the CI	Commit of the run source
source.buildUrl	PROBARA_BUILD_URL	string	detected from the CI	CI build URL of the run source
createMissingCases	PROBARA_CREATE_MISSING_CASES	boolean	true	Create a case for a test that matches none (missing cases)
suiteUlid	PROBARA_SUITE_ULID	string	the project root	Suite that created cases go under
closeRun	PROBARA_CLOSE_RUN	boolean	true for a new run, false for a reused one	Close the run at the end
closeRuns	—	Record<string, boolean>	none	Close the run of each project ({ SHOP: true, WEB: false }) when closeRun is not set
uploadAttachments	PROBARA_UPLOAD_ATTACHMENTS	boolean	true	false uploads no file (attachments)
captureOutput	PROBARA_CAPTURE_OUTPUT	boolean	false	Attach each attempt’s stdout and stderr as stdout.log and stderr.log
statusMapping	PROBARA_STATUS_MAPPING	{ passed, failed, skipped, blocked }	none	Send a status as another (failed=blocked) (statuses)
statusFilter	PROBARA_STATUS_FILTER	string[]	none	Send no result with these statuses, after the mapping (skipped)
resultsFile	PROBARA_RESULTS_FILE	string	none	JSON file for what could not be sent (results file)
assignFailedTo	PROBARA_ASSIGN_FAILED_TO	string[]	none	Assign each failed result without an assignee to one of these members, in turn (assigning failures)
debug	PROBARA_DEBUG	boolean	false	Log every request (debugging)
rootDir	—	string	Playwright’s rootDir	Directory the file paths of automation keys are relative to (keys)
chunkSize	—	number	500	Results per report request, 1 to 500
timeoutMs	—	number	30000	Timeout of one HTTP attempt, in milliseconds, 1 to 600000 (network)
maxRetries	—	number	4	Retries of a failed request, 0 to 10
attachmentConcurrency	—	number	2	Results whose attachments upload at the same time, 1 to 8

The reporter sends its own name and version in the User-Agent (such as probara-playwright-reporter/0.1.0 probara-core/0.2.0 node/22.12.0); there is no option for it.

In GitHub Actions the source also gets these fields, detected from GITHUB_* variables:

Option	Variable	Type	Default	What it does
source.provider	—	string	detected from GitHub Actions	github_actions (detected) or other
source.repository	—	string	detected from GitHub Actions	Repository of the run source, owner/name
source.commitAuthor	—	string	detected from GitHub Actions	Who triggered the build
source.pullRequestNumber	—	number	detected from GitHub Actions	Pull request of the build
source.workflow	—	string	detected from GitHub Actions	CI workflow of the run source
source.ciRunId	—	string	detected from GitHub Actions	CI run id: the jobs of one GitHub Actions run share one run
source.ciRunAttempt	—	number	detected from GitHub Actions	Attempt of the CI run: a re-run reports into a run of its own
source.ciJob	—	string	detected from GitHub Actions	CI job of the run source

Runtime options

For code and tests, the reporter also takes the runtime options of @probara/core, the library it is built on. They have no variable, and a playwright.config rarely needs them:

Option	Type	Default	What it does
logger	{ debug, info, warn, error }	[probara] lines on stderr	Where the reporter logs
env	Record<string, string>	process.env	Where PROBARA_* and CI variables are read from
fetch	function	the global fetch	The HTTP client
sleep	function	setTimeout	Waits between retries
random	function	Math.random	Jitter of the retry backoff
now	function	the current time	Clock of the default run name and of Retry-After dates

What happens with each setup

Setup	What the reporter does
No token and no project, or enabled: false	Stays off and quiet: nothing is sent, nothing is logged (except at debug).
Only one of them, or an invalid value	Reporting is off; each problem is logged, and the tests run as usual.
A token and a project, every value valid	Reports.

A value of the wrong type, such as run.tags: 'nightly' instead of a list, is an invalid value:

reporter: [['@probara/playwright-reporter', { projectId: 'SHOP', run: { tags: 'nightly' } }]],

$ npx playwright test
[probara] Probara reporting is off: run.tags must be a list of strings

Problems name the option or the variable at fault, never its value, so a token never shows up in the log.

Turn reporting off

Set enabled: false or PROBARA_ENABLED=false, or leave the token out: the reporter sends nothing and stays quiet. Builds that get no secrets, such as pull requests from forks on most CIs, report nothing on their own. To keep what a run would have sent, even with reporting off, set a results file.

Base URL

baseUrl (PROBARA_BASE_URL) is the Probara host the reporter talks to: the URL of the app, without a path. It defaults to https://app.probara.net, and requests go to that host and nowhere else. Leave it unset unless Probara gives you another host.

Network

The reporter talks HTTPS to one host, the base URL, from Playwright’s main process.

Timeouts

Setting	Default	Range	What it bounds
timeoutMs	30000 ms	1 to 600000	One attempt of a request, the response body included
Uploads	120000 ms, or timeoutMs if larger	—	One attempt of an attachment upload (a large file on a slow link)

A timed-out attempt is retried like a network error.

Retries and backoff

Answer	What happens
Network error, timeout, 408, 429, 500, 502, 503, 504	Retried
409 with a Retry-After (the same request is still in flight)	Retried
A 201 whose body never arrived	Retried; the idempotency key replays it
Any other answer (400, 401, 403, 404, 409, 413, 422…)	Not retried: the request fails

• maxRetries (0 to 10, default 4) sets the retries after the first attempt: 5 attempts in all by default.
• The wait before retry n is 1 s × 2ⁿ⁻¹, at most 30 s, plus up to 20 % of random jitter.
• A Retry-After header (seconds or an HTTP date) replaces that wait, capped at 120 s.
• Each retry is logged, with the attempt and the wait.

With the defaults, a request that keeps failing gives up after about 15 s of waiting (1 + 2 + 4 + 8 s, up to 18 s with the jitter), plus the attempts themselves, each at most timeoutMs. The worst case is a Retry-After on every answer: 120 s before each retry, so 8 minutes of waiting with the default 4 retries (20 minutes with maxRetries: 10). That is per request, and a failed report stops the later ones of its project. Playwright waits for the reporter before it exits, so lower maxRetries or timeoutMs to bound a CI step.

Rate limit

Probara limits API requests per organization (60 requests per minute on Free, 1,200 on Pro, and 3,000 on Enterprise), for every token of the organization together. What a run costs:

Request	How many
Report	1 per chunk of up to 500 results (chunkSize)
Attachment stage	1 per result with files (per 20 files or 64 MiB)
Attachment commit	1 per result with files
Close	1, when attachments were queued (otherwise the last report closes the run)

A run of 2000 attempts without files costs 4 requests; one where 100 attempts have a trace costs about 200 more. 429 answers are retried after their Retry-After. To spread uploads, lower attachmentConcurrency; to send fewer files, keep traces and videos for failures (retain-on-failure) or set uploadAttachments: false.

Chunks and ordering

• Results go out in chunks of chunkSize (1 to 500, default 500), one after another, in the order the attempts ended. A run case keeps the last outcome recorded, so the order matters. A chunk also stays within Probara’s totals of steps and tags per report.
• The first chunk creates the run (unless run.ulid names one); the others reuse it.
• When a chunk fails after its retries, the later chunks of that project are not sent and the run is left open; a results file keeps them.
• Uploads start once their report is recorded, attachmentConcurrency results at a time (1 to 8, default 2), and the run is closed after the uploads.

Idempotency

Every report, run creation, run close and attachment commit carries an Idempotency-Key, the same on each retry of that request. A retried report whose first attempt was recorded after all is replayed by Probara, never recorded twice (idempotency). Running playwright test again is a new run, with every result recorded again.

Proxies

The reporter uses Node’s built-in fetch and has no proxy option of its own:

Node.js	Proxy support
22.21 or later, 24 or later	Set NODE_USE_ENV_PROXY=1 with HTTPS_PROXY (and NO_PROXY); fetch goes through it
22.12 to 22.20	Not supported: fetch ignores HTTPS_PROXY and HTTP_PROXY

NODE_USE_ENV_PROXY=1 HTTPS_PROXY=http://proxy.example.com:8080 npx playwright test

The setting applies to the whole Node process (Node’s docs): put localhost in NO_PROXY if your tests reach a local server through fetch. On older Node versions, upgrade Node or allow app.probara.net through the firewall.

Custom certificate authorities

When your network inspects TLS with a private certificate authority (a corporate proxy, say), give Node the CA certificate (PEM) when it starts:

NODE_EXTRA_CA_CERTS=/etc/ssl/certs/company-ca.pem npx playwright test

Node reads NODE_EXTRA_CA_CERTS once at startup, so set it in the environment of the step. On Node 22.19 and later, NODE_USE_SYSTEM_CA=1 trusts the operating system’s certificate store instead. There is no option to turn TLS verification off.

Next: link your tests to cases, or see the run options.