Tila AI

An all-in-one AI coding companion that teaches, reviews, converts, and optimizes code with a modern, integrated IDE

Comment

Tila AI

Inspiration

Learning to code can be overwhelming. Traditional tutorials lack interactivity, and finding personalized help is expensive. We envisioned an AI-powered platform that acts as a personal coding mentor - one that can speak with you, understand your curriculum, generate challenges tailored to your level, and help you write better code. The combination of LiquidMetal AI's Raindrop Platform and Vultr's infrastructure gave us the perfect foundation to build this vision.

What it does

Tila AI is a comprehensive AI-powered coding companion that transforms how developers learn and code:

  • 🎓 AI Tutor with Voice - Chat with an AI mentor using text or voice (ElevenLabs). Upload your syllabus and the AI tailors responses to your curriculum.

  • 💻 Smart Code Editor - Write code with AI assistance that can:

    • Auto-convert code when you switch languages
    • Complete code based on your logic
    • Generate optimized solutions
    • Create unit tests automatically
    • Analyze code quality and detect bugs

  • 📓 Interactive Notebooks - Create coding notebooks and export them as Markdown files

  • 🗺️ Learning Roadmaps - Generate personalized learning paths for any technology

  • 🏆 Coding Challenges - AI-generated challenges to test and improve your skills

  • 👥 Community Hub - Share code, discover solutions, and collaborate with other developers

How we built it

Frontend:

  • React 18 with TypeScript for type-safe, component-based UI
  • Vite for fast development and optimized builds
  • TailwindCSS for responsive, modern styling

Backend:

  • Hono.js - lightweight, fast web framework
  • LiquidMetal AI Raindrop Platform providing:
    • SmartMemory for context-aware AI conversations
    • SmartBucket for document storage
    • KV Cache for session management
    • Actors for stateful user management

AI Integration:

  • Google Gemini 1.5 Flash for code generation, analysis, and chat
  • ElevenLabs for voice-to-text and AI mentor voice conversations

Infrastructure:

  • Deployed on Vultr Cloud via Raindrop Platform
  • Auto-scaling and load balancing
  • Secure HTTPS endpoints

Challenges we ran into

  1. Windows CLI Compatibility - The Raindrop CLI had issues spawning npx on Windows. We had to patch the CLI bundle files to use npx.cmd instead.

  2. Environment Variable Access - Discovered that env parameter in Raindrop Service class was undefined. Had to use this.env to access environment bindings correctly.

  3. Real-time Voice Integration - Synchronizing ElevenLabs voice responses with the chat interface required careful state management to ensure smooth user experience.

  4. Context Management - Building a system where AI remembers syllabus content and previous conversations while staying within token limits was challenging.

Accomplishments that we're proud of

  • Seamless Voice Integration - Users can naturally speak to their AI mentor and receive voice responses, making learning more accessible and engaging.

  • Smart Language Conversion - The code editor intelligently converts code between languages, preserving logic and adapting syntax automatically.

  • Syllabus-Aware Learning - The AI tutor actually understands uploaded curriculum and provides contextually relevant explanations.

  • Full-Stack Raindrop Deployment - Successfully deployed a production-ready application using Raindrop's SmartMemory, SmartBucket, KV Cache, and Actors.

  • Comprehensive Feature Set - Built a complete platform with chat, editor, notebooks, roadmaps, challenges, and community features - all AI-powered.

What we learned

  • Raindrop Platform Power - SmartMemory and SmartBucket make building AI-aware applications significantly easier than traditional approaches.

  • Voice AI UX - Integrating voice requires careful attention to latency, feedback, and fallback mechanisms for the best user experience.

  • AI Prompt Engineering - Crafting prompts that produce consistent, high-quality code generation and analysis results is an art.

  • Serverless Architecture - Raindrop's serverless model with Actors provides elegant solutions for stateful operations without managing infrastructure.

What's next for Tila AI

  • 🔌 IDE Extensions - VS Code and JetBrains plugins to bring Tila AI directly into developers' existing workflows

  • 👥 Real-time Collaboration - Multiple users coding together with shared AI assistance

  • 📊 Learning Analytics - Track progress, identify weak areas, and get personalized improvement suggestions

  • 🌍 More Languages - Expand voice support to multiple languages for global accessibility

  • 🎮 Gamification - Badges, streaks, and leaderboards to make learning more engaging

  • 📱 Mobile App - Learn and code on the go with a dedicated mobile experience

  • 🔗 GitHub Integration - Connect repositories for AI-powered code reviews and suggestions

Built With

Share this project:

Updates