Kai - Your AI Coach

An immersive, voice-interactive AI coach. Kai uses advanced Neuro-Linguistic Programming coaching models to help you find your own solutions and achieve your goals!

Comment

Inspiration

Our team was fascinated by the idea of making personal development more accessible. We know that professional coaching is an incredibly powerful tool, but it's often out of reach for many due to cost, scheduling, and the simple difficulty of finding the right coach. We asked ourselves: could we use the power of modern AI to create a supportive, insightful coaching experience that was available to anyone, anytime? We were inspired by the potential to build something that didn't just answer questions, but could actually help people reflect and grow.

What it does

Kai is your personal AI coach. It's an immersive, voice-interactive web application that provides a safe and private space to explore your goals and challenges. Kai is available 24/7 with no booking required.

Through natural conversation, Kai guides you using proven NLP frameworks to help you overcome limiting beliefs, clarify your objectives, and commit to actionable steps. At the end of a session, it generates a structured summary of your conversation in either Markdown or PDF format, giving you a tangible record of your progress. It even features a "Hands-Free" mode for a truly immersive, natural conversational flow.

How we built it

We built Kai over the weekend with a modern, full-stack approach.

  • The Frontend: We chose SvelteKit for its speed and simplicity, allowing us to build a reactive and animated UI very quickly. We used Tailwind CSS for styling and the Web Speech API for real-time transcription. The immersive, orb-centric design was a key focus for us.

  • The Backend: We used Python with FastAPI to build a robust and high-performance server. This backend handles all the heavy lifting: managing the conversation state and integrating with the external AI services.

  • The AI "Brain": This is the heart of our project. We used the Requesty API Gateway to access the incredible Gemini 1.5 Flash model. We spent a significant amount of time on "prompt engineering," designing Kai's personality by combining the conversational GROW model with the deep, analytical questions from the NLP Well-Formed Outcome pattern. This is what makes Kai a true coach, not just a chatbot.

  • The Voice: Kai's warm and empathetic voice is generated in real-time using the ElevenLabs API.

  • Our AI Partner: Throughout the entire process, the Roocode / Kilo Code AI assistant was an essential member of our team. We used it for everything from architecting the initial application to writing boilerplate code, and most importantly, for helping us hunt down the most complex and stubborn bugs.

Challenges we ran into

Our biggest challenge was the marathon of debugging required to get all the different APIs talking to each other correctly. We faced a series of cascading issues, starting with a mysterious 404 Not Found error from the Requesty API. We methodically debugged our code, only to realize the issue was an incorrect API endpoint specific to the hackathon. After fixing that, we had to solve a series of subtle bugs with the ElevenLabs API, correcting function names and parameters one by one. It was a true test of our persistence and debugging skills, but each error we solved taught us something new about how these complex systems interact.

Accomplishments that we're proud of

We are incredibly proud of building a fully functional, end-to-end AI application in a single weekend. But more than that, we're proud of the quality of the final experience. The "Hands-Free" conversational mode, the subtle UI animations, and the advanced prompt engineering that gives Kai its coaching ability are all features we believe elevate this project from a simple prototype to a genuinely useful and engaging tool. We're also proud of the final PDF summary feature, which provides a professional and valuable takeaway for the user.

What we learned

This hackathon was a masterclass in persistence and the power of modern AI tools. We learned that debugging a full-stack AI application requires a holistic approach, looking at everything from frontend state and backend logic to the specific error codes from external APIs. Most importantly, we learned that the quality of an AI's interaction is not just about the model you use, but about the care and thought you put into designing its personality and its purpose through detailed prompt engineering.

What's next for Kai - Your AI Coach

We are incredibly excited about Kai's potential. The next logical step would be to add a secure database (like Firebase or Supabase) to allow users to save their session history and track their progress over time. We'd also love to expand Kai's "coaching toolkit" with more specific NLP patterns and guided exercises for things like goal-setting and overcoming specific fears. We believe Kai has the potential to become a truly valuable companion for anyone on a journey of personal growth.

Built With

  • css3
  • elevenlabs-api
  • fastapi
  • github
  • google-gemini-1.5-flash
  • html5
  • javascript
  • python
  • reportlab
  • requesty-api
  • roocode
  • sveltekit
  • tailwind-css
  • vercel
  • web-audio-api
  • web-speech-api
Share this project:

Updates