🌟 Inspiration I wanted to build an AI agent that felt professional yet energetic. Luna isn't just a basic chatbot; she is a resilient specialist. My goal was to create a tool that bridges cultural gaps through translation while staying strictly within her professional scope.
🚀 What it does Luna translates complex phrases across 24 different languages, providing native scripts (like Hindi, Japanese, and Korean) and essential cultural context. She is built with strict guardrails to ensure she only performs translation tasks and nothing else.
🛠️ How I built it The project is powered by n8n for automation and Gemini 1.5 Flash for high-speed processing.
Custom Engineering: I manually engineered the JSON payloads to set a temperature of 0.3 for maximum accuracy.
Resilience: I wrote custom JavaScript logic to handle 429 API rate limit errors, ensuring Luna stays online even during heavy use.
Security: I implemented robust system instructions to act as safety guardrails.
💪 Challenges I overcame The biggest challenge was handling the API rate limits. By building a custom retry mechanism in n8n, I was able to make Luna much more stable than a standard out-of-the-box agent.
🎓 What I learned I learned how to combine UI/UX design thinking with deep technical automation to create a product that is both beautiful and highly functional.
I created this project and its accompanying video for the purposes of entering the Gemini Live Agent Challenge hackathon. #GeminiLiveAgentChallenge
Built With
- ai
- api
- gemini
- javascript
- json
- n8n
Log in or sign up for Devpost to join the conversation.