Trivia Wizard 🧙‍♂️

Summon your inner genius.

Daily trivia challenges, infinite customizable practice, and more. An AI trivia platform built with Next.js, Merit System's Echo billing integration, and the Vercel AI SDK.

Game Modes

• Daily Quiz - One deterministic daily challenge per day with category rotation (EST timezone)
• Freeplay Mode - AI-powered quizzes with infinite categories, replayability
• Faceoff- Challenge your friends to a trivia battle, facing the same 5 or 10 questions
• Survival - How many can you answer correctly in a row?
• Jeopardy - Our take on the most famous trivia game ever played... how high can you score?
• Campaign - Play through our entire array of categories at your own pace, aiding the Trivia Wizard in his battle against the Drift

Features

• Quiz History - Review up to 20 past quiz sessions with full answer details
• Results & Ranking - Earn performance titles (120+ unique ranks across 6 tiers) and share your results
• Answer Review - Detailed explanations for every question with correct answer display
• Recipe System - Deterministic quiz generation for preset categories (infinite variety via seed-based recipes)
• Custom Categories - Ask about any topic with simplified AI generation
• Performance Tracking - Category-bazsed statistics stored locally
• Share Feature - Native share with emoji grids and referral codes

Tech Stack

• Framework: Next.js 15 (App Router)
• Styling: Tailwind CSS + shadcn/ui
• AI: Vercel AI SDK + Echo LLM Provider
• Billing: Echo metered LLM usage
• Storage: IndexedDB (local quiz history)
• Database: Account Info, Stats, Achievements stored on Supabase
• Hosting: Vercel

Quiz Generation: Seed → Recipe → LLM Pipeline

Trivia Wizard uses two generation modes:

1. Recipe System (Preset Categories Only)

For the 400+ preset categories (History, Science, etc.), quizzes are generated deterministically:

Seed Generation:

• Daily Quizzes: SHA256(date + category) → Same recipe every time for that day/category
• Practice Quizzes: Random 32-byte hex seed → Unique recipe each time

Recipe Building: A recipe is deterministically generated from the seed and includes:

• Tone (6 options): scholarly, playful, cinematic, pub_quiz, deadpan, sports_banter
• Category Mix: 4-6 related categories selected from preset
• Question Types (2-3 types): multiple_choice, true_false, fill_blank, ordering
• Difficulty Curve (3 patterns): ramp (gradual), wave (alternating), valley (easy middle)
• Explanation Style: one_line_fact, compare_contrast, mini_story, why_wrong

LLM Generation:

• Claude receives the recipe as strict constraints in the prompt
• Generates questions matching the specified tone, era, and difficulty progression
• The LLM never decides config - it only fills content based on the deterministic recipe
• JSON schema validation with automatic repair if malformed

Permutation Space: 10M+ possible recipes = infinite variety

2. Simple Generation (Custom Categories)

For custom user-entered topics (e.g., "Quantum Physics" or "1980s Movies"):

• No recipe system - bypasses the complexity
• Direct prompt with user's category, difficulty, and question count
• Still uses JSON schema and validation
• Faster generation, less constrained

Why Two Modes?

• Preset categories: Reproducible daily quizzes + consistent quality + infinite variety
• Custom categories: Flexibility to quiz on any topic without recipe overhead

Categories

Preset Categories (~400 and growing!)

History, Science, Geography, Literature, Art & Culture, Music, Movies & TV, Sports, Technology, Food & Drink, Mythology, Politics, Nature & Animals, Mathematics, Philosophy, Video Games, Space & Astronomy, Medicine & Health, Economics, Language & Linguistics & More!

Note: Custom categories can be anything you want!

Question Configuration

Question Types

• Multiple Choice: 3-5 options, one correct answer
• True/False: Binary choice with explanation
• Short Answer: Text input with AI-powered fuzzy matching (≥85% threshold)
• Mixed: Random mix of all types

Difficulty Levels

• Easy: Accessible questions for casual players
• Medium: Moderate challenge
• Hard: Expert-level questions
• Mixed: Dynamic difficulty throughout quiz

Question Counts

• 5 questions: Quick quiz (~2-3 minutes)
• 10 questions: Standard quiz (~5-7 minutes)

Performance Ranking System

After each quiz, you earn a title based on your score percentage:

6 Performance Tiers (120+ unique titles):

• Disaster Zone (0-19%): "Complete Catastrophe", "Absolute Trainwreck", etc.
• Needs Work (20-49%): "Rough Start", "Room for Improvement", etc.
• Not Bad (50-69%): "Decent Effort", "Getting There", etc.
• Pretty Good (70-84%): "Well Done", "Impressive Performance", etc.
• Excellent (85-99%): "Outstanding", "Near Perfect", etc.
• Absolute Perfection (100%): "Flawless Victory", "Perfect Score", etc.

Titles are randomly selected from the tier pool to add variety to repeated plays.

Echo Integration

The app uses Echo (by Merit Systems) for authentication and metered AI billing.

Features:

• User Authentication: Echo React SDK with sign-in/sign-out
• Metered LLM Billing: Pay-per-use AI generation (transparent cost tracking)
• Balance Display: Real-time account balance in UI
• Top-Up: Quick links to add credits
• Referral System: User IDs included in share messages

AI Model:

• Claude Sonnet 4 (claude-sonnet-4-20250514)
• Accessed via Echo's anthropic() provider
• Wrapped with Vercel AI SDK's generateText()

Echo Endpoints:

• Quiz Generation (/api/trivia/generate): Metered through Echo
• Answer Evaluation (/api/trivia/evaluate): Metered through Echo (fuzzy matching only)

All LLM calls are tracked and billed through Echo's platform.

Roadmap

Future enhancements being considered:

• Image-based questions
• Audio/video question support
• Timed speedrun mode
• Quiz templates & presets
• Export/Import quiz JSON
• Leveling Sytem

Adaptive Difficulty Engine

• Quizzes learn from player performance and dynamically adjust question difficulty.

Category Mastery Paths

• Mini “skill trees” where players unlock harder tiers in topics as they excel.

“Boss Questions”

• End-of-quiz ultra-hard question that gives bonus XP or a special badge.

Progressive Hints

• Tiered hint system: word reveal → elimination → full explanation.

Profile Badges

• Achievement-based icons (e.g., “History Master”, “Perfectionist 10×”).

Community Packs/Pacts

• Curated pacts made by influencers/experts (e.g., “Physicist Pack”).

LLM Cross-Verification

• A second pass to verify answer accuracy using another model or endpoint.

Explain-Like-I’m-5 Mode

• Auto-generate simpler explanations for accessibility.

“Continue the Lesson” Mode

• After a quiz, users get a short mini-lesson expanding on the questions.

Trivia Tutor

• Chat mode where the AI can drill you on weak areas or generate follow-ups.

Full Player Profile System

• XP, levels, ranks, badges, and category mastery metrics.

Pack Store (Free)

• Add thematic quiz packs:
• 90s Nostalgia
• World Mythologies
• NBA Legends
• Startups & Tech
• US History AP Level

Expert Packs (Paid or Referral Unlocks)

• Unlock premium or special packs with referrals or Echo credits.

Curriculum Mode:

• AP/IB-style structured practice sets (English, History, Bio)

Offline Mode

• Caches recipes and quizzes for offline play in PWA mode.

License

MIT