Most playlist tools react. Playlistify learns.
Playlistify AI is not a playlist generator. It's a personal music agent that learns your emotional patterns and intent over time, then proactively creates and evolves Spotify playlists to support focus, mood, and daily routines.
🔗 Live Application: playlistify.up.railway.app
🚀 Pro Version (Apify): apify.com/viverun/playlistfy
| Traditional Playlist Generators | Playlistify AI |
|---|---|
| Enter mood → get playlist | Understands emotional intent behind your words |
| Forgets you after each use | Remembers your preferences and evolves |
| Waits for commands | Proactively suggests based on patterns |
| Dumps raw tracks | Explains why it chose each track |
| Static output | Playlists that evolve based on feedback |
"This week you've been stressed. I kept energy low and vocals minimal."
Music is not trivial. Music is memory, emotion, routine, identity.
We built Playlistify AI to be emotionally intelligent — a system that doesn't just respond to "play happy music," but understands:
- "I need something to survive a 2am debugging session"
- "Music that feels like quiet confidence before a presentation"
- "Background vibes that won't distract me while I write"
| Pillar | Description |
|---|---|
| 🎯 Intent > Mood | We parse emotional context, not just keywords |
| 🧠 Agent Memory | Learns what you like at different times and adapts |
| ⚡ Agentic Behavior | Suggests playlists without being asked |
| 💬 Explainability | Every playlist comes with reasoning |
| 🔄 Evolution | Playlists improve based on your feedback |
| 🔒 Privacy-First | We store derived preferences, never raw data |
| Team Member | Role | Connect |
|---|---|---|
| Avanish Kasar | Lead Developer |   |
| Jamil | Co-Developer |
Not just "happy music" — we understand context:
Input: "Late night coding session, low stress"
Parsed Intent:
├── Emotional State: focused, calm
├── Activity: coding
├── Time Context: late-night
├── Target Energy: medium
├── BPM Range: 100-130
├── Vocal Preference: instrumental
└── Suggested Genres: electronic, lo-fi, ambient
Playlistify AI remembers you:
- What worked: Tracks you liked, playlists you loved
- What didn't: Songs you skipped, vibes that missed
- When you listen: Morning vs. late-night preferences
- How you evolve: Your taste fingerprint develops over time
// Every interaction teaches the agent
"Last time you liked lo-fi + low BPM at night. I kept that energy."The agent acts without being asked:
"You usually listen around this time. Want a fresh focus playlist?"
Based on:
- Habitual listening times
- Day-of-week patterns
- Recent mood trends
Every playlist explains why:
"I picked tracks that match your 'focused' mood for coding, at late-night. I kept the energy medium and chose low-vocal tracks based on your past preferences."
Playlists are systems, not outputs:
- Like/Skip buttons on every track
- Feedback influences future generations
- Weekly refresh suggestions
- One-click modifications ("more energy", "less vocals")
- We store only derived preferences, never raw listening history
- Local JSON storage (no external DB required)
- One-click data export (GDPR compliance)
- Clear all data anytime
Click above to watch a complete walkthrough of Playlistify AI
┌─────────────────────────────────────────────────────────────────────┐
│ PLAYLISTIFY AI │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────────────┐ │
│ │ Frontend │───▶│ Express.js │───▶│ Agentic Engine │ │
│ │ (Liquid UI) │ │ API │ │ (Brain of the app) │ │
│ └──────────────┘ └──────────────┘ └──────────────────────┘ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────┐ ┌──────────────────────┐ │
│ │ Spotify │ │ Intent Engine │ │
│ │ Handler │ │ (NLP Understanding) │ │
│ └──────────────┘ └──────────────────────┘ │
│ │ │ │
│ ▼ ▼ │
│ ┌──────────────┐ ┌──────────────────────┐ │
│ │ Spotify │ │ Agent Memory │ │
│ │ Web API │ │ (Learning System) │ │
│ └──────────────┘ └──────────────────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────┐ │
│ │ Local JSON Store │ │
│ │ (Privacy-First) │ │
│ └──────────────────────┘ │
│ │
└─────────────────────────────────────────────────────────────────────┘
| Component | File | Purpose |
|---|---|---|
| Agent Memory | src/agentMemory.ts |
Persistent learning system |
| Intent Engine | src/intentEngine.ts |
Natural language understanding |
| Agentic Engine | src/agenticEngine.ts |
Proactive behavior & orchestration |
| Spotify Handler | src/spotifyHandler.ts |
Spotify API integration |
| Main API | src/main.ts |
Express server & endpoints |
| Category | Technology |
|---|---|
| Runtime | Node.js 20+ |
| Language | TypeScript |
| Backend | Express.js |
| Frontend | Vanilla HTML/CSS/JS with iOS Liquid Glass Theme |
| API Integration | Spotify Web API |
| NLP | Custom intent engine with 50+ emotional concepts |
| Memory | JSON-based persistent storage |
| Containerization | Docker |
| Hosting | Railway |
| Pro Version | Apify Actor with MCP Tools |
- Input: Describe your ideal playlist in natural language
- Analysis: NLP engine extracts mood, genre, tempo, and era
- Search: Dual-strategy approach finds the perfect tracks
- Curation: Smart filtering removes duplicates and optimizes flow
- Output: Get a polished playlist ready to enjoy
Visit playlistify.up.railway.app and start creating playlists instantly!
-
Clone the repository:
git clone https://github.com/avanishkasar/Playlistify-AI.git cd Playlistify-AI -
Install dependencies:
npm install
-
Configure environment (optional for testing):
The app includes default credentials for quick testing. For production, create a
.envfile:SPOTIFY_CLIENT_ID=your_client_id SPOTIFY_CLIENT_SECRET=your_client_secret SPOTIFY_REFRESH_TOKEN=your_refresh_token PORT=3001
-
Start the development server:
npm run dev
Open
http://localhost:3001in your browser.
Endpoint: POST /api/generate-playlist
The brain of the operation. Parses intent, generates recommendations, creates playlist, explains choices, and updates memory.
Request:
{
"prompt": "I need focus music for a late night coding session",
"userId": "user-123"
}Response:
{
"playlist": {
"id": "...",
"name": "Late Night Coding Focus",
"uri": "spotify:playlist:..."
},
"tracks": [...],
"explanation": {
"summary": "I picked tracks that match your 'focused' mood for coding...",
"factors": [
{ "factor": "Emotional State", "value": "focused, calm", "reasoning": "..." }
],
"personalizations": ["Based on your preference for lo-fi at night..."]
},
"modifications": [
{ "action": "increase_energy", "label": "More energy" },
{ "action": "decrease_vocals", "label": "Less vocals" }
],
"memoryId": "playlist-uuid"
}Endpoint: GET /api/proactive-suggestion?userId=user-123
Check if the agent has a proactive suggestion based on user patterns.
Response:
{
"hasSuggestion": true,
"suggestion": {
"type": "habitual_time",
"message": "You usually listen around this time. Want a fresh focus playlist?",
"suggestedPrompt": "focus music",
"confidence": 0.85
}
}Endpoint: POST /api/parse-intent
Preview how the system understands your input before generating.
Request:
{
"prompt": "melancholic indie for a rainy evening",
"userId": "user-123"
}Response:
{
"emotionalState": {
"primary": "melancholic",
"secondary": ["nostalgic", "reflective"],
"intensity": 0.7
},
"audioParams": {
"energy": 0.4,
"bpmRange": [70, 100],
"preferredGenres": ["indie", "folk"],
"vocalPreference": "soft-vocals"
},
"reasoning": [...]
}Endpoint: POST /api/feedback/track
Record like/skip actions on individual tracks.
Request:
{
"userId": "user-123",
"memoryId": "playlist-uuid",
"trackId": "spotify:track:...",
"action": "like"
}Endpoint: POST /api/feedback/playlist
Rate an entire playlist (1-5 stars).
Request:
{
"userId": "user-123",
"memoryId": "playlist-uuid",
"rating": 4,
"keepFavorites": ["track1", "track2"]
}Endpoint: GET /api/taste-profile?userId=user-123
Get the user's learned taste fingerprint.
Response:
{
"profile": {
"dominantMoods": ["focused", "calm", "energetic"],
"genreAffinities": {
"lo-fi": 0.8,
"electronic": 0.6,
"indie": 0.5
},
"timePreferences": {
"morning": { "mood": "energetic", "genre": "pop" },
"night": { "mood": "calm", "genre": "lo-fi" }
}
}
}Endpoint: POST /api/evolve-playlist
Refresh/evolve a playlist based on feedback and new preferences.
Request:
{
"userId": "user-123",
"memoryId": "playlist-uuid",
"keepFavorites": true,
"direction": "more_energy"
}Endpoint: DELETE /api/memory?userId=user-123
Delete all stored preferences for a user.
Endpoint: GET /api/export-data?userId=user-123
Export all user data in JSON format.
Endpoint: POST /mcp
Request:
{
"prompt": "Upbeat jazz for cooking"
}Response:
{
"tracks": [
{
"name": "Track Name",
"artist": "Artist Name",
"uri": "spotify:track:..."
}
]
}This project is optimized for Railway deployment:
- Fork this repository
- Sign up at Railway
- Create New Project → Deploy from GitHub
- Select your forked repo
- Railway auto-detects the Dockerfile and deploys
Railway provides:
- ✅ Automatic HTTPS
- ✅ Environment variable management
- ✅ Auto-deploy on git push
- ✅ Free tier available
The agent understands intent, not just keywords:
| What You Say | What We Understand |
|---|---|
| "I need to survive a 2am debugging session" | Late-night + focus + low energy + instrumental |
| "Music that feels like quiet confidence" | Calm assertiveness + medium tempo + minimal lyrics |
| "Background vibes that won't distract me" | Ambient + low vocals + steady rhythm |
| "Getting pumped for my presentation" | Pre-performance energy + motivational + building intensity |
- "Energetic workout songs from the 2000s"
- "Relaxing acoustic guitar for Sunday morning"
- "Dark electronic music for late night coding"
- "Happy pop songs for a road trip"
- "Melancholic indie rock for introspection"
- "खुश गाने सुबह के लिए" (Happy songs for morning - Hindi)
- "இரவு நேரத்திற்கான மெதுவான பாடல்கள்" (Slow songs for night - Tamil)
- "నా జిమ్ వర్కౌట్ కోసం ఎనర్జెటిక్ పాటలు" (Energetic songs for gym - Telugu)
1. You generate a playlist
↓
2. Agent stores the intent + characteristics
↓
3. You interact (like/skip tracks, rate playlist)
↓
4. Agent updates your taste profile
↓
5. Next time: more personalized recommendations
✅ We store:
- Derived preferences (e.g., "prefers lo-fi at night")
- Aggregated feedback patterns
- Time-based listening tendencies
❌ We never store:
- Raw listening history
- Personal information
- Track-by-track logs
Over time, the agent builds a profile like:
{
"dominantMoods": ["focused", "calm"],
"genreAffinities": { "lo-fi": 0.8, "electronic": 0.6 },
"timePreferences": {
"night": { "mood": "calm", "genre": "ambient" },
"morning": { "mood": "energetic", "genre": "pop" }
},
"avoidPatterns": ["heavy-metal", "aggressive-vocals"]
}For developers and power users, we offer a Pro version on Apify with MCP (Model Context Protocol) tools:
MCP Tools Available:
search-track- Search Spotify tracks by queryrecommend- Get AI-powered track recommendationscreate-playlist- Create playlists directly on Spotify
Pricing: Pay per event (API call) - perfect for automation and integrations!
Contributions are welcome! Feel free to:
- 🐛 Report bugs
- 💡 Suggest new features
- 🔧 Submit pull requests
This project is licensed under the MIT License - see the LICENSE file for details.
- Built with Spotify Web API
- Deployed on Railway
- Pro Version on Apify
- Powered by TypeScript and Node.js
- Built for Hack This Fall 2025 🏆
"Most AI tools react. This one learns. This one remembers. This one feels human."
| Criteria | Our Answer |
|---|---|
| Innovation | First playlist tool that combines agent memory + emotional intent + proactive behavior |
| Technical Depth | 3 new cognitive modules, 50+ emotional concepts, persistent learning |
| User Impact | Music that matches your emotional state, not just keywords |
| Privacy | Derived preferences only, full data export, one-click deletion |
| Scalability | JSON storage = no DB required, works anywhere |
Made with ❤️ by Team DDoxer
"Music is not trivial. Music is memory, emotion, routine, identity."
Try it Now | Pro Version | Report Issue | Request Feature
...