Playwright trace viewer

mnml-test-playwright is a terminal viewer for Playwright trace.zip files — the artifact each failing (or trace: 'on') Playwright test drops in test-results/<test>/. Browse the per-action timeline, console messages, errors, and stdio with type-toggle filters. Runs standalone in any terminal.

┌─ trace ──────────────────────────────────────────────────────────┐
│ checkout.spec.ts:24 · filters:  actions  console  errors  stdio  │
└──────────────────────────────────────────────────────────────────┘
┌─ events (47/89) ─────────────────────────────────────────────────┐
│ ▸ ▶    0.012s  page.goto https://shop.example.com                │
│   ▶    1.203s  page.fill #email = [email protected]               │
│   ▶    1.412s  page.click [data-testid="submit"]                 │
│   ●    1.518s  Error: Timeout 30000ms exceeded                   │
│   …                                                              │
└──────────────────────────────────────────────────────────────────┘
  ↑↓/jk · a/c/e/s · E errors-only · R show-all · r reload · q quit

What stays in mnml core

The Playwright test runner (tests.run, tests.run_file, tests.run_cursor, the flaky-tests history view, the “heal with AI” flow) stays in mnml core. Those need tight editor integration — run the test for the buffer you have open, jump to the failing assertion’s source line, edit the spec from the same window.

What lives in this sibling is the trace viewer — the read-only side. Useful when:

• You’re inspecting a trace.zip from CI without running mnml.
• You want a dedicated pane for trace browsing rather than the inline trace render mnml ships.
• You’re sharing a trace with a collaborator who doesn’t run mnml — they install one binary, get a real viewer.

This split landed 2026-06-06; the in-tree Pane::Trace was removed.

Install

cargo install --git https://github.com/chris-mclennan/mnml-test-playwright mnml-test-playwright

Usage

# Open a single trace.zip
mnml-test-playwright path/to/trace.zip

# Print version and exit (no auth or config to validate — trace files are self-contained)
mnml-test-playwright --check

The path is positional — no config file in v0.1. Trace files are self-contained ZIPs (test name, action timeline, console messages, network requests, screenshots, snapshots) so there’s nothing to set up.

Each Playwright test that runs with trace: 'retain-on-failure' (or 'on') drops a trace.zip under test-results/<test>/; point this viewer at that file.

Layout

One main column — a scrollable event list. The header shows the test file + line and the currently-active filter chips.

Each event row carries:

• A type chip — ▶ action / 📜 console / ● error / › stdio
• The relative timestamp from test start (0.012s, 1.518s, etc.)
• A summary — for actions, the method + args; for console, the message; for errors, the stack frame’s top line.

Filtering is non-destructive — toggling a chip hides those rows from the list (events (47/89) — 47 visible of 89 total) but doesn’t drop them from memory. Toggle back on to see them again.

Filter presets

The four single-key toggles plus two presets:

Key	Action
a	Toggle Actions
c	Toggle Console
e	Toggle Errors
s	Toggle Stdio
E	Preset — errors only (everything else off)
R	Preset — show all kinds

Default on open: all four kinds visible.

Keys

Chord	Action
↑ / k	Move selection up
↓ / j	Move selection down
PgUp / PgDn	Page up / down
g / G	Top / bottom
a / c / e / s	Toggle Actions / Console / Errors / Stdio
E	Errors-only preset
R	Show-all preset
r	Reload trace from disk (re-opens the ZIP)
q / Esc / Ctrl+C	Quit

r is useful when the trace file is being overwritten by a re-run — reload to pick up the new events without restarting the binary.

Two run modes

Standalone

mnml-test-playwright path/to/trace.zip

The TUI takes over until you q.

Hosted as a mnml Pty pane

:term mnml-test-playwright path/to/trace.zip

mnml spawns it in a Pty pane — splittable, focusable, key-routed like any other pane.

The positional trace.zip path is passed through verbatim — useful for wiring it into a palette command that opens a fresh trace on demand.

Wire it into mnml’s left rail

If you want a one-click chip in mnml’s rail that opens the last failing trace, drop this into your ~/.config/mnml/config.toml:

[[ui.integration_icon]]
id       = "playwright_trace"
glyph    = "\U000F0668"            # nf-md-play_circle (TOML 8-digit form)
fallback = "P"
command  = ":term mnml-test-playwright last-failure-trace.zip"
color    = "purple"
tooltip  = "Open last Playwright trace"

Setting [[ui.integration_icon]] replaces the built-in defaults, so copy the defaults from src/config.rs into your config first if you want to extend rather than replace. See the launcher-icon strips for the field reference.

You can also point command at a shell-resolved path via :term mnml-test-playwright $(find test-results -name trace.zip | head -1) — though that requires the shell to expand before the ex-cmdline runs, which mnml’s parser doesn’t do today. For dynamic paths, the easier route is a thin wrapper script that calls mnml-test-playwright with the discovered path.

Status

v0.1 (this release) — Trace viewer only. No flaky-test history view, no test runner, no detail panel for individual events (errors render their first stack line; the full trace is in the ZIP). v0.2 may grow:

• A flaky-tests history view that reads mnml’s per-workspace .mnml/test-history.json (predicated on a real use case for “I want this without running mnml”).
• A detail panel for the focused event — full stack on errors, full args on actions, full network request/response on network rows.
• Snapshot / screenshot rendering inline (currently the trace ZIP carries snapshot HTML that the viewer doesn’t unpack).

Source

The viewer lives in its own sibling repo: github.com/chris-mclennan/mnml-test-playwright. MIT-licensed. See Building integrations for the anatomy of an integration, or Community integrations for the directory of siblings.