TypeShi

• Timed typing tests with configurable durations (15, 30, 60, 120 seconds)
• Real-time visual feedback: correct characters highlight green, incorrect highlight red
• WPM and accuracy calculation
• Guest access, tests without an account or login
• User authentication via Auth0 (email/password or Google OAuth)
• Automatic result saving for authenticated users
• Profile dashboard: average WPM, best WPM, total tests taken, average accuracy
• Paginated test history with the ability to delete individual results
• Account management: editable username, password reset via Auth0

Request flow

1. User visits the Vercel URL, browser loads the React SPA
2. Frontend fetches random words from /api/words (no auth required)
3. User types; all keystroke logic runs locally in the browser with zero network calls
4. Timer ends; WPM and accuracy calculated client-side
5. If authenticated, frontend POSTs results to /api/tests with JWT in the Authorization header
6. Backend middleware validates the JWT against Auth0's public keys (JWKS)
7. Prisma inserts the test result into PostgreSQL
8. Profile page fetches stats and history via two parallel GET requests

Authentication flow

1. User clicks login; browser redirects to Auth0's Universal Login page
2. User enters credentials on Auth0's domain; the app never sees the password
3. Auth0 validates and issues a signed JWT (RS256)
4. Browser redirects back to TypeShi with the token
5. Frontend stores the token in memory, not localStorage
6. Every authenticated API call includes the token as Authorization: Bearer <token>
7. Backend validates the signature using Auth0's cached public keys

20 source files total.

users id (UUID, PK), auth0_id (unique, links to Auth0 account), email (unique), username (unique), created_at, updated_at

typing_tests id (UUID, PK), user_id (FK to users, cascade delete), wpm and accuracy (Decimal), duration (seconds), total_chars, correct_chars (Integer), created_at. Indexed on [user_id, created_at DESC] for fast history queries.

word_bank id (auto-increment), word (unique), ~200 common English words seeded on setup.

Words Per Minute:

The standard "5 characters = 1 word" convention normalizes for varying word lengths. Only correctly typed characters count toward WPM (net WPM, not gross).

Accuracy:

Rounded to one decimal place. Defaults to 0 if no characters were typed.

Edge cases:

• Typing through all available text before the timer ends ends the test early; WPM uses actual elapsed time
• Backspace resets a character's status to "pending" and adjusts both correct and total counts
• Zero characters typed results in 0 WPM and 0% accuracy, with no division by zero

Satisfies all functional requirements R1–R10 from the project requirements document: