Skip to main content

Quiz App Prompt

Logo Quiz App using Google ADK​

An infinite real‑time quiz: display company logos, let users identify them, update scores live, and continue indefinitely while users can join or leave anytime.

Tech Stack:

  • Backend: Python with Google ADK (Agents) + FastAPI
  • Frontend: React (UI)
  • Database: PostgreSQL
  • Communication: WebSockets for live leaderboard
  • Deployment: Docker (backend, frontend, database)

1. Core Features​

  1. User Experience

    • User enters a nickname (no authentication).
    • Presented with a random logo and four multiple-choice answers.
    • Scores: +10 for correct, -5 for incorrect.
    • Continuous gameplay (no end screen).
    • Leaderboard updates live for all connected players.

  2. Logos

    • Seeded dataset in PostgreSQL or pulled via external API (e.g., Clearbit Logos).
    • Logos stored as URLs.

  3. Leaderboard

    • Real-time updates via WebSockets.
    • Shows top 10 players with live ranks.

  4. Agent-Driven Orchestration

    • LogoQuizAgent (built with Google ADK) manages:
      • Fetching logos and building questions.
      • Validating answers.
      • Updating scores.
      • Broadcasting leaderboard changes.

2. Architecture Overview​

ADK Agent (Python)​

  • Primary Agent: LogoQuizAgent
    • Built with LlmAgent from google-adk.
    • Uses tools for:
      • Fetching questions.
      • Checking answers.
      • Updating and fetching leaderboard.
    • Orchestrated with Runner and SessionService for user sessions.
    • Uses LoopAgent or similar for continuous game cycles.

Example Agent:

from google.adk.agents import LlmAgent
from google.adk.runners import Runner
from my_logo_tools import get_next_question, check_answer, fetch_leaderboard

logo_agent = LlmAgent(
    name="logo_quiz_agent",
    model="gemini-2.0-flash",
    instruction="""Run infinite quiz rounds: fetch a question, validate user answers, update scores, return next question.""",
    tools=[get_next_question, check_answer, fetch_leaderboard],
)

runner = Runner(agent=logo_agent)

Backend (FastAPI + ADK)​

  • Endpoints:
    • POST /join: Create user session via ADK SessionService.
    • GET /question: Invoke get_next_question.
    • POST /answer: Invoke check_answer and update score.
    • GET /leaderboard: Fetch leaderboard snapshot.
    • WS /ws/leaderboard: Push leaderboard changes in real time.

Frontend (React)​

  • Pages:

    • Landing: Nickname input.
    • Quiz: Displays logo, options, and score.
    • Leaderboard: Side panel with live updates.

  • Uses REST for initial state and WebSocket for leaderboard.

  • State managed by Redux or Zustand.

Database (PostgreSQL)​

Tables:

  1. users (id, nickname, score, joined_at)
  2. logos (id, company_name, logo_url)
  3. questions (id, logo_id)
  4. answers_log (id, user_id, question_id, is_correct, timestamp)

Seed logos or fetch dynamically.

Dockerization​

Services:

  • backend: FastAPI + ADK.
  • frontend: React build served by Nginx.
  • db: PostgreSQL with seeded data.
  • Optional: pgadmin for DB inspection.

3. Game Loop Workflow​

  1. User joins → Backend registers session.
  2. Backend requests a question via get_next_question (ADK tool).
  3. Question sent to frontend.
  4. User answers → Backend invokes check_answer (ADK tool).
  5. Score updated in DB, leaderboard recalculated via fetch_leaderboard.
  6. WebSocket broadcasts new leaderboard.
  7. New question sent automatically; loop continues.

4. Tool Functions (Python)​

# my_logo_tools.py

def get_next_question(session_state) -> dict:
    # Fetch random logo, build choices, track in session state
    return {
        "question_id": 123,
        "logo_url": "https://cdn.example.com/google.png",
        "options": ["Google", "Amazon", "Netflix", "Microsoft"]
    }

def check_answer(session_state, question_id, answer) -> dict:
    # Verify answer, update score, log entry
    correct = answer == "Google"  # Example
    session_state["score"] += 10 if correct else -5
    return {"correct": correct, "new_score": session_state["score"]}

def fetch_leaderboard(session_state) -> list:
    # Query DB for top 10 users
    return [{"nickname": "Alice", "score": 120}, {"nickname": "Bob", "score": 95}]

Wrapped as ADK tools for LogoQuizAgent.

5. API Contracts​

REST​

  • POST /join

    { "nickname": "Deepak" }

    Response:

    { "user_id": 1, "nickname": "Deepak", "score": 0 }

  • GET /question

    { "question_id": 12, "logo_url": "https://cdn.example.com/apple.png", "options": ["Apple", "Samsung", "Sony", "Intel"] }

  • POST /answer

    { "user_id": 1, "question_id": 12, "answer": "Apple" }

    Response:

    { "correct": true, "new_score": 40 }

  • GET /leaderboard

    [ { "nickname": "Alice", "score": 120 }, { "nickname": "Bob", "score": 95 } ]

WebSocket​

ws://backend/ws/leaderboard

  • Broadcast: { "leaderboard": [ { "nickname": "Alice", "score": 130 }, { "nickname": "Bob", "score": 105 } ] }

6. Folder Structure​

quiz-app/
  backend/
    app/
      main.py
      agents/
        logo_agent.py
      tools/
        my_logo_tools.py
      routes/
        users.py
        quiz.py
        leaderboard.py
      websocket.py
    Dockerfile
  frontend/
    src/
      components/
      pages/
    Dockerfile
  db/
    init.sql
  docker-compose.yml

7. Deployment Steps​

  1. Build Docker images: docker-compose build
  2. Start the stack: docker-compose up
  3. Access frontend at http://localhost:3000 and backend API at http://localhost:8000.

8. Optional Enhancements​

  • Time limit per question (e.g., 10 seconds).
  • Daily/weekly leaderboard resets.
  • Admin dashboard for uploading logos.
  • Multi-agent setup (e.g., separate ScoreAgent, QuestionAgent using ADK).