Featured project 6 · Python · API + UI

Python pytest + Playwright Framework

API, UI and hybrid test automation in Python with pytest, Playwright, requests and pydantic: 46 tests of a notes app in about 1 m 45 s with 4 parallel workers, on Chromium, Firefox and WebKit in GitHub Actions.

View source code View documentation View latest CI runs View test report

Purpose

Shows a Python test stack for one application with a web UI and a REST API on the same backend: the public Expand Testing Notes app. The API carries most of the checks, the browser checks what a user sees, and hybrid tests check that the two sides agree.

Technology stack

Area	Tools
Language	Python 3.12
Test runner	pytest 9 (fixtures, markers, parametrize, hooks)
Browser automation	Playwright for Python 1.63 with pytest-playwright (Chromium, Firefox, WebKit)
API	requests 2.34 (one Session per client)
Contracts	pydantic 2 models with extra fields forbidden
Parallel runs and report	pytest-xdist, pytest-html (one self-contained file), Playwright traces and videos
Code quality and CI	ruff (lint + format), GitHub Actions: lint gate, API job, browser matrix, nightly run

Architecture overview

tests/api, tests/ui, tests/e2e        what is checked
      │  fixtures (tests/conftest.py): throwaway user per test, logged-in client,
      │  page with ads blocked, page already logged in, report hooks
      │
framework/api        NotesApiClient (requests) + pydantic models
framework/pages      LoginPage, NotesPage (data-testid locators, no assertions)
framework/data       unique users and notes (uuid)
      │
framework/config     URLs and timeouts from environment variables

Key engineering decisions

Every test registers its own user through the API and deletes it in teardown, so tests are independent and safe with pytest-xdist on a shared public site
UI tests log in by putting the API token into localStorage before the page loads; the login form itself is tested once
The API client returns raw responses and never asserts, so the same client serves positive and negative tests
pydantic models with extra="forbid": an added or renamed field fails a test at once instead of passing silently
Page objects hold data-testid locators and actions; tests assert with Playwright's retrying expect, with no sleeps anywhere
Demo tests that fail, skip and xfail on purpose carry their own marker and stay out of normal runs

Implemented features

32 API tests: registration and validation, login, the same 401 for a wrong password and an unknown user, token checks, profile, password change, logout, notes CRUD with PUT and PATCH, bad ids, one user cannot see another user's notes
9 UI tests: login form, adding, completing, deleting and searching notes
5 hybrid tests: data created through the API shown in the app, changes made in the app checked through the API
HTML report with an embedded screenshot for every failed UI test; Playwright screenshot, video and trace kept for failures
Framework design notes and 86 interview questions with answers linked to the code (Markdown + PDF)

Setup and execution

python -m venv .venv && .venv\Scripts\activate
pip install -r requirements.txt
playwright install chromium firefox webkit
pytest -n 4                              # all 46 tests
pytest -m api                            # API only
pytest -m "ui or e2e" --browser firefox  # browser tests on Firefox
pytest -m demo                           # report demo: 2 failed, 1 skipped, 1 xfailed

Sample test results

pytest-html report: 46 tests passed — pytest-html report of a full run: 46 passed.

pytest-html report of the demo run with failures, a skip and an xfail — The report demo: a UI failure with its screenshot, an API failure with the API's message, a strict xfail and a skip.

CI workflow

On every push, pull request and night: ruff lint and format check first, then the API tests with 4 workers and the UI + hybrid tests on Chromium, Firefox and WebKit as three parallel jobs. Each job writes a pass/fail table to the run summary and keeps its HTML report and traces as an artifact.