> ## Documentation Index
> Fetch the complete documentation index at: https://docs.diffraction.sh/llms.txt
> Use this file to discover all available pages before exploring further.

# Existing Playwright tests

> Discover and select existing tests for isolated execution with native recordings.

## Select standing flows

1. Open **Repos → repository → Tests** as a Pro workspace administrator.
2. Enable discovery for the next review. Discovery is off by default.
3. After a catalog is available, select up to ten important flows.
4. Confirm the dedicated synthetic-data requirement for raw native media and save.

No discovered test is selected implicitly. Organization policy can restrict native
tests. Settings and the selected catalog identities are versioned and captured for
each review; changed or mismatched identities fail explicitly rather than silently
substituting another test.

## What runs

Selected tests use their original native Playwright projects, fixtures, assertions,
setup, retries, and repeats. This path does not rewrite tests into AI-generated
flows or upgrade repository dependencies. The pinned runtime must support the
repository's runner version. The current supported Playwright versions are
1.58.2 and 1.62.1; other versions require a compatible runtime before execution.

Standing native coverage begins before AI impact classification and planning, so
it can still provide results when visual review is skipped. Selected flows replace
the legacy `review.e2eCommand` for that review, avoiding duplicate execution.
Native tests do not themselves invoke an AI planner or enable video analysis.

Each native execution has a three-minute execution limit, a five-minute combined
source/bootstrap/execution bound, two workers, and a 40 MB retained-media limit.
Unsupported setup, exhausted bounds, and omitted media remain visible.

## Inspect results

Read test and browser-project identity, status, and recorded steps separately
from the generated UI journey. Native videos and traces are private artifacts.
Native steps do not claim exact video timestamps, and native traces are separate
from the semantic replay viewer. Expired artifacts are unavailable.

Raw native media require dedicated synthetic data. Semantic rrweb masks do not
sanitize native video or trace bytes. Production credentials/session content and
general authenticated customer recording are outside the current verified gate.

Production fixture tests have produced verified videos and traces in Chromium
and WebKit. That evidence does not establish compatibility with every Playwright
suite or complete signed-in production viewer verification.
