perplexitree - AI-Powered Knowledge Tree Game

• Live at: https://perplexitree.app/
• Live demo: https://www.youtube.com/watch?v=P5U9jo97fCk

A beautiful, interactive web game that combines the meditative puzzle mechanics of "Prune" with AI-powered knowledge exploration. Players grow knowledge trees by pruning branches while leveraging the Perplexity API to discover unique, real-time information on any topic. Each growth session returns fresh, non-redundant search results to build comprehensive understanding.

Project Summary

Purpose: perplexitree transforms learning into an engaging, visual experience where users cultivate knowledge trees through strategic pruning and AI-enhanced exploration. The game addresses information overload by presenting unique, curated insights through an intuitive tree metaphor.

Technical Approach: Built with a modular architecture featuring FastAPI backend and vanilla JavaScript frontend, the application integrates Perplexity's real-time search capabilities to generate unique knowledge areas for each growth session. The system ensures non-redundant information retrieval through intelligent query generation and structured output parsing.

Perplexity API Integration: The game leverages Perplexity's reasoning and retrieval capabilities through a two-phase approach:

• Initial Search: Uses Perplexity's sonar-pro chat completion model with structured JSON output to generate 5 unique knowledge areas from the initial query
• Subsequent Growth: Uses Perplexity's search model with negative prompting to find unique web results for each new branch
• Flashcard Generation: Uses sonar-pro chat completion to create study materials from search results
• Unique Results: Negative prompting system ensures each growth session returns fresh, non-redundant information

The integration ensures players receive diverse, comprehensive information while maintaining the game's meditative, focused learning experience.

Features

• AI-Powered Knowledge Tree: Grow branches that represent real search results from Perplexity API
• Interactive Tools: 7 different tools for managing your knowledge tree
• Unique Search Results: Each growth session returns fresh, non-redundant information
• Study System: Create flashcards from search results for enhanced learning
• Visual Progression: Transform knowledge into flowers and fruits
• Game State Persistence: Save and load your knowledge trees

Game Mechanics

Tools Available

1. Growth Tool: Click on nodes to grow new branches with search results
2. Cut Tool: Click and drag to prune unwanted branches
3. Leaves Tool: Create flashcards from search results on branches
4. Flower Tool: Add flowers to branch ends (knowledge blossoming)
5. Fruit Tool: Transform flowers into apples (fruit of labor)
6. Reposition Tool: Drag branch ends to move and resize branches
7. Study Tool: Hover over nodes to view search result details
8. Pan Tool: Drag to move around the view

Core Gameplay

1. Start: Enter a search query to begin growing your knowledge tree
2. Grow: Use the growth tool to expand branches with AI-powered search results
3. Prune: Cut away branches that don't contribute to your learning goals
4. Study: Create flashcards and study materials from your search results
5. Progress: Watch your knowledge tree evolve with flowers and fruits

Installation & Setup

Prerequisites

• Python 3.8+
• Virtual environment (recommended)

Quick Start

1. Navigate to the backend directory:

cd backend

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Run the server:

python main.py

5. Open your browser and go to: http://localhost:8001

Environment Variables

Create a .env file in the backend/ directory:

PERPLEXITY_API_KEY=your_perplexity_api_key_here

Technical Implementation

Architecture

• Backend: FastAPI server with SQLite database
• Frontend: Vanilla JavaScript with HTML5 Canvas rendering
• AI Integration: Perplexity API for real-time knowledge generation

Key API Endpoints

POST /api/search - Initial Knowledge Tree Creation

• Uses Perplexity sonar-pro with structured JSON output
• Generates exactly 5 unique knowledge areas from user query

POST /api/web-search - Branch Growth with Negative Prompting

• Key Feature: Uses negative prompting to exclude existing results
• Ensures each growth session returns fresh, non-redundant information
• Constructs queries like: "machine learning -"existing result 1" -"existing result 2"

POST /api/save-game-state - Public Game Storage

• Saves complete game state (public saves - all games are shareable)
• Stores branches, search results, flashcards, and visual elements

POST /api/create-flashcards - AI Study Material Generation

• Uses Perplexity to create flashcards from search content
• Links flashcards to specific tree nodes with difficulty ratings

Perplexity API Integration

Two-Phase Approach

1. Initial Search: sonar-pro chat completion for structured knowledge areas
2. Subsequent Growth: Search API with negative prompting for unique results

Negative Prompting Strategy

• Purpose: Prevents redundant information across growth sessions
• Implementation: Excludes existing search result titles and snippets
• Result: Each branch growth returns fresh, unique content

Database Design

• Relational Model: GameSession → SearchResult → Branch hierarchy
• Public Access: All saved games are publicly accessible
• Cascade Deletion: Automatic cleanup when sessions are deleted

Key Technologies

Backend: FastAPI, SQLAlchemy, Perplexity API
Frontend: HTML5 Canvas, Vanilla JavaScript
Database: SQLite with relational modeling

License

Created for Perplexity Hackathon. Tree growth/prune mechanic inspired by Prune game by Joel McDonald.