Calmly-Coach

Ever frozen in a difficult conversation? Calmly Coach uses AI to analyze your situation and suggest 3 emotionally intelligent responses (calm, empathetic, direct). Built with Raindrop + Vultr AI.

Comment

💡 Inspiration

Have you ever frozen in a difficult conversation, unsure what to say? Whether it's a tense workplace discussion, a relationship conflict, or a challenging friendship moment - we've all been there.

Calmly Coach helps you respond with confidence and emotional intelligence.

🎯 What It Does

Calmly Coach analyzes your communication challenges and generates three AI-powered response suggestions:

  1. 🧘 Calm & Professional - Composed and measured
  2. 💙 Empathetic - Warm and understanding
  3. 🎯 Direct but Kind - Clear and honest

Key Features:

  • Context Detection - Automatically identifies work vs relationship dynamics
  • Emotional Analysis - Detects frustration, anxiety, or uncertainty in your situation
  • Personalized Learning - Remembers your communication style preferences
  • Explanation Mode - Teaches WHY each response works
  • Real-time AI - Powered by Llama 3.3 70B (70 billion parameters)

🛠️ How We Built It

Raindrop Platform (LiquidMetal AI)

Built entirely on Raindrop using four SmartComponents:

1. SmartInference - AI response generation

const response = await env.ai.run('@cf/meta/llama-3.3-70b-instruct', {
  messages: [{ role: 'user', content: prompt }],
  max_tokens: 1000,
  temperature: 0.7
});

2. SmartMemory - Four-layer personalization system

  • Working memory: Current session
  • Episodic memory: Interaction history
  • Semantic memory: Learned preferences
  • Procedural memory: Communication templates typescript await env.USER_MEMORY.putMemory({ session_id: userId, content: JSON.stringify(interaction), key: "tone_preference" });

3. SmartBucket - Semantic search for scenario templates

const scenarios = await env.SCENARIO_LIBRARY.documentSearch({
  query: "team conflict attribution",
  threshold: 0.7
});

4. SQL Database - User data and analytics

Built with Claude Code: 100% of backend code generated using Claude Code AI assistant via Raindrop MCP Server

Vultr Integration

Production Database Infrastructure:

  • Service: Vultr Managed PostgreSQL 16
  • Region: Johannesburg
  • Features:
    • Automatic daily backups
    • Connection pooling (1000+ concurrent users)
    • SSL/TLS encryption
    • 99.99% uptime SLA
    • Real-time monitoring dashboard

Why Vultr:

  • Zero-ops database management
  • Production-grade reliability
  • Cost-effective scaling ($15/mo starter → enterprise)
  • Geographic distribution for low latency

Architecture:

  • Raindrop SmartMemory handles real-time AI personalization
  • Vultr PostgreSQL provides durable storage for user data
  • Hybrid approach balances AI capabilities with production reliability

Frontend

  • React 18 + TypeScript (strict mode)
  • TailwindCSS + Framer Motion animations
  • Vite for fast builds
  • Deployed on Netlify

🚧 Challenges We Ran Into

1. Prompt Engineering

  • Initial AI responses were generic and unhelpful
  • Solution: Iterative refinement with explicit structure requirements
  • Result: Contextual, actionable responses with explanations

2. Memory Architecture

  • Balancing immediate queries vs long-term learning
  • Solution: Implemented Raindrop's four-layer memory system
  • Result: <500ms retrieval with continuous learning

3. Context Detection

  • Distinguishing work vs relationship contexts reliably
  • Solution: Enhanced keyword detection + confidence scoring
  • Result: 85%+ accuracy in testing

4. Async Processing

  • Early version had UI lag from synchronous memory writes
  • Solution: Queue-based architecture with Observer pattern
  • Result: Sub-3s API responses

🏆 Accomplishments

Shipped in 7 days - Working product from concept to deployment Full Raindrop integration - SmartInference, SmartMemory, SmartBucket, SQL all working together Production architecture - Proper error handling, async processing, type safety Built with AI - 100% of backend code generated using Claude Code Beautiful UX - Responsive design with smooth animations Vultr infrastructure - Production-ready database with monitoring and backups

📚 What We Learned

Technical:

  • How to leverage SmartMemory's four-layer architecture
  • Effective prompt engineering for consistent AI behavior
  • Queue-based patterns for scalable microservices
  • Raindrop MCP's power for rapid development

Product:

  • People struggle with difficult conversations more than expected
  • Users want multiple options, not a single "best" answer
  • Explanations ("why this works") are as valuable as the responses
  • Context matters - workplace ≠ relationship communication

AI Development:

  • Claude Code can generate production-quality code
  • MCP servers eliminate months of infrastructure work
  • SmartComponents are game-changers for AI apps

🚀 What's Next

Immediate:

  • [ ] Voice mode with ElevenLabs (practice responses out loud)
  • [ ] Response refinement ("make it firmer/softer")
  • [ ] Mobile-responsive improvements

Near-term:

  • [ ] Chrome extension for Gmail/Slack
  • [ ] Team features for workplace coaching
  • [ ] Analytics dashboard

Long-term:

  • [ ] Multi-language support (Spanish, Mandarin)
  • [ ] Mobile apps (iOS/Android)
  • [ ] Enterprise plan for HR teams
  • [ ] Community scenario library

Vision: Make emotionally intelligent communication accessible to everyone.

🛠️ Built With

raindrop vultr llama-3.3-70b smartmemory smartbucket smartinference react typescript tailwindcss framer-motion claude-code postgresql netlify vite

🔗 Try It

👥 Team

Sharon - Solo Developer (Full-stack, AI integration, UI/UX)

Built with Claude Code AI assistant

📝 License

MIT - Open source and free to use

🙏 Thanks

LiquidMetal AI • Vultr • Anthropic • Open Source Community

#AI #Communication #EmotionalIntelligence #Raindrop #Vultr #Hackathon

### **What Problem Does This Solve?** (Short answer field)
```People freeze in difficult conversations - workplace conflicts, 
relationship issues, or challenging friendships. Calmly Coach 
provides AI-powered response suggestions with three different 
communication styles (calm, empathetic, direct) plus explanations 
of why each works, helping users respond with confidence and 
emotional intelligence.

Technologies Used (Select tags)

  • raindrop
  • vultr
  • artificial-intelligence
  • react
  • typescript
  • postgresql
  • machine-learning
  • natural-language-processing
  • llm

Links:

Website/Demo URL:

https://calmly-coach.netlify.app/

GitHub Repository:

https://github.com/Justsharon/AI-championship-Calmy-Coach

Video Demo:

https://youtu.be/sDOZlXZKsPI

Built With

Share this project:

Updates