DevDuck

Inspiration

Rubber duck debugging is a beloved hacker tradition — just by explaining your code out loud, you often discover the solution. We asked ourselves: what if the duck could actually respond? With one of us passionate about hardware (Mechatronics Engineering) and the other passionate about AI/ML (Computer Science), we set out to bring a rubber duck to life.

Features

DevDuck is a sentiment-aware chatbot paired with a physical duck: -🗣️ Hackers chat with DevDuck like they would with a debugging buddy.

-🤖 DevDuck analyzes the tone and sentiment of the conversation.

-🦆 The physical duck reacts with movements that match the mood — encouragement if you’re frustrated, celebration if you’re excited, calm nods when things make sense.

-💬 On top of that, DevDuck can support general coding-related discussions and is designed to expand into deeper debugging assistance.

It’s not just a chatbot — it’s a companion that makes debugging feel less lonely and more interactive.

How we built it

-AI & Sentiment Analysis: Vapi for real-time LLM (OpenAI Real Time API) + voice capabilities, with sentiment extracted from conversations.

-Hardware: Arduino with servos and simple components (plus a bit of hot glue) to animate the duck.

-Integration: Utilised serial communication to send commands from python script to arduino for physical movement. Seperate chain invoked for sentiment analysis

Challenges we ran into

-Synchronizing chatbot output with hardware reactions in real time.

-Building smooth hardware–software communication under hackathon constraints.

-Managing Vapi integration and figuring out how to access conversational data at different points in the pipeline for other uses.

-Experimenting with RAG-based context injection so the duck could eventually answer coding-related questions more effectively.

Accomplishments that we're proud of

-Built a working prototype that brings AI + hardware + hacker culture together.

-Designed modular duck-control functions that simplify hardware expansion.

-Created a memorable hackathon project that puts a playful twist on debugging.

What we learned

-How to integrate Vapi for live, conversational AI experiences.

-Practical techniques for bridging LLM sentiment with physical responses.

-That playfulness and personality can make technical tools much more engaging.

What’s next for DevDuck

-🛠️ Add true codebase integration so DevDuck can directly analyze files.

-🎛️ Expand the duck’s behavior set for richer, more expressive reactions.

-☁️ Integrate with IDEs like VSCode so DevDuck can live alongside your workflow.

Built With

• aiofiles
• arduino
• c++
• css
• electron
• fastapi
• gpt-realtime
• hotglue
• html
• javascript
• ngrok
• openai
• pydantic
• pyserial
• python
• speechrecognition
• uvicorn
• vapi
• vapi-ai/web
• websockets