@diffdeckai/playwright-reporter

A Playwright reporter that uploads each test's video recording — together with a per-step timeline — to DiffDeck. DiffDeck hosts the videos and renders the step timeline over them so you can scrub a failing run step by step.

The reporter does not run or change your tests. It hooks into Playwright's reporting lifecycle: on each test end it finds the recorded video attachment, packages the step timing it captured via onStepBegin/onStepEnd, and POSTs both to your DiffDeck repository.

Install

npm install --save-dev @diffdeckai/playwright-reporter

@playwright/test is a peer dependency (Playwright >=1.30).

Configure

Add the reporter to playwright.config.ts and make sure Playwright is recording video (the reporter has nothing to upload otherwise):

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

export default defineConfig({
  use: {
    video: "on", // or "retain-on-failure"
  },
  reporter: [
    ["list"],
    ["@diffdeckai/playwright-reporter"],
  ],
});

That's the whole integration. The only thing you must supply is the project token via DIFFDECK_TOKEN (see below) — host is optional and defaults to https://diffdeck.ai (production), and branch/commit auto-resolve from CI env vars. Point at a self-hosted / non-default instance only if you need to:

reporter: [["list"], ["@diffdeckai/playwright-reporter", { host: "https://diffdeck.example" }]],

Options

All options are optional and fall back to environment variables.

Option	Type	Env fallback	Default	Description
host	string	DIFFDECK_HOST	https://diffdeck.ai	DiffDeck base URL. Recordings are POSTed to <host>/api/products/ui-review/recordings.
token	string	DIFFDECK_TOKEN	—	Per-repo project token, sent as the X-UI-Review-Token header. Required to upload.
branch	string	DIFFDECK_BRANCH, GITHUB_HEAD_REF, GITHUB_REF_NAME, CI_COMMIT_REF_NAME, GIT_BRANCH	—	Git branch tagged on each recording.
commitSha	string	DIFFDECK_COMMIT, GITHUB_SHA, CI_COMMIT_SHA, GIT_COMMIT	—	Git commit SHA tagged on each recording.
stepCategories	string[]	—	all categories	If set, only steps with these Playwright categories are captured (e.g. ["test.step","expect"]).
quiet	boolean	—	false	Silence the reporter's console output.
mode	"upload" | "write"	DIFFDECK_MODE	"upload"	"upload" (default) posts each recording during the run. "write" instead writes a metadata sidecar (see below).

Environment variables

The only thing CI must provide is the token (and usually the host):

export DIFFDECK_TOKEN="your-repo-project-token"
export DIFFDECK_HOST="https://diffdeck.ai"   # optional; this is the default

If no token is found, the reporter logs a one-time warning and uploads nothing — your test run is never failed by a missing token.

What gets uploaded

For each completed test that has a video, the reporter sends a multipart/form-data request:

Field	Source
video	the recorded video blob (video/webm or video/mp4)
videoType	the video MIME type
testTitle	test.title
testFile	spec file, relative to the Playwright rootDir
testId	test.id
status	passed | failed | timedOut | skipped | interrupted
durationMs	result.duration
retries	result.retry
branch	resolved branch (see options)
commitSha	resolved commit (see options)
metadata	JSON string — the step-timeline document (below)

Authentication is the X-UI-Review-Token header.

If the repository does not have the DiffDeck recordings add-on enabled, the route responds 402. The reporter treats this as a soft signal: it logs once and stops uploading for the rest of the run, without failing the test run.

Deferred upload (mode: "write")

By default the reporter uploads during the run — that's the recommended integration and needs no extra CI step. If you'd rather upload after the run (for example to keep the test job offline, or to batch the upload in a separate step), set mode: "write" (or DIFFDECK_MODE=write). Instead of uploading, the reporter writes a <video>.json metadata sidecar next to each recorded video, carrying the same fields it would have uploaded (test title, file, id, status, duration, retries, branch, commit, and the step timeline). No token is needed in write mode.

A later step then uploads them — e.g. the diffdeck CLI / GitHub Action, which reads those sidecars so nothing has to be derived from file paths:

- uses: diffdeck/diffdeck-action@v1
  with:
    command: recording
    token: ${{ secrets.DIFFDECK_TOKEN }}
    videos: test-results   # dir of videos + their .json sidecars

Most projects should stay on the default "upload" mode.

Step-timeline metadata shape

The metadata field is a JSON document describing the per-step timeline. This is the data the DiffDeck player renders over the video. All step times are in milliseconds relative to the start of the recording, where the recording start equals the test's startTime (recordingStartTime).

{
  "schemaVersion": 1,
  "reporter": "@diffdeckai/playwright-reporter",
  "reporterVersion": "0.1.0",
  "recordingStartTime": "2026-06-30T12:00:00.000Z", // ISO; timeline zero point
  "testDurationMs": 1500,
  "steps": [
    {
      "title": "sign in",            // step label
      "category": "test.step",       // test.step | expect | pw:api | hook | fixture | ...
      "startMs": 100,                // offset from recordingStartTime (>= 0)
      "endMs": 500,                  // offset from recordingStartTime (>= startMs)
      "durationMs": 400,             // endMs - startMs
      "depth": 0,                    // nesting depth (0 = top level)
      "location": "e2e/login.spec.ts:12:5", // optional source location
      "error": "locator.click: timeout"     // optional; present only on failed steps
    },
    {
      "title": "expect(page).toHaveURL(\"/dashboard\")",
      "category": "expect",
      "startMs": 300,
      "endMs": 350,
      "durationMs": 50,
      "depth": 1
    }
  ]
}

TypeScript types for this document are exported from the package:

import type { RecordingMetadata, RecordingStep } from "@diffdeckai/playwright-reporter";
import { RECORDING_METADATA_SCHEMA_VERSION } from "@diffdeckai/playwright-reporter";

Notes on alignment

• startMs/endMs are computed from each step's startTime minus the test's startTime, then clamped to >= 0. The test start time is the best alignment to the video that Playwright exposes; sub-frame drift between the first painted frame and startTime is expected and small.
• Steps are ordered by startMs (ties keep capture order), and depth reflects nesting so a consumer can render the timeline hierarchically.

Development

npm install
npm run build       # tsc -> dist/
npm test            # node:test unit tests (mocked fetch)
npm run typecheck

License

MIT