MindfulAI - DevPost Submission

Inspiration

The inspiration for MindfulAI came from a perfect convergence of cutting-edge technology and a deeply personal mission. Initially, I was captivated by the emerging possibilities of conversational AI, particularly getting hands-on with powerful tools like Tavus and ElevenLabs. The agentic AI trend had me completely hooked, and I was determined to build something meaningful in the conversational AI space.

But then it hit me - why not combine this revolutionary agentic conversational AI with one of the most common yet often unspoken challenges we all face: mental health? Every individual, including myself, has something going on in their life. Sometimes all we need is someone who will truly listen to us and be there for us. That realization sparked the idea to create an AI companion that could provide that constant, understanding presence while leveraging the latest in conversational AI technology.

What it does

MindfulAI lives up to its name by being a truly mindful AI companion. The platform offers multiple therapeutic modalities including video calls with AI avatars, natural voice conversations, intelligent chat sessions, comprehensive journaling, and personalized meditation exercises.

But here's what makes it revolutionary: everything is unified by our advanced global memory system. This isn't just a collection of separate features - it's a cohesive ecosystem where every interaction, from the smallest mood check-in to lengthy therapy sessions, contributes to a comprehensive 5000-token memory profile unique to each user.

This means that no matter what activity a user engages in, no matter what day they return to the app, the AI knows exactly what's going on in their life and can pick up right where they left off. Whether they're switching from a morning meditation to an evening journal entry, or returning after weeks away, MindfulAI maintains complete therapeutic continuity and provides deeply personalized support that truly understands their journey.

Challenges we ran into

The development journey was filled with significant technical challenges that tested both patience and problem-solving skills. The first major hurdle was deciding on the tech stack and evaluating the available tools in my arsenal, particularly around database selection. I ended up changing databases three times before finding the right fit, which consumed considerable development time and required multiple refactoring sessions.

Authentication setup presented another significant challenge. Initially, I attempted to use NextAuth, but it felt overly complex and wasn't as streamlined as I needed for the project's scope. This led me to discover Convex Auth, which became a game-changer by allowing me to streamline both database operations and authentication under the Convex ecosystem - a decision that significantly simplified the architecture.

Beyond these major challenges, there were numerous smaller but equally frustrating obstacles. Linter issues proved particularly annoying, requiring constant configuration tweaks to maintain code quality standards. Perhaps the most surprising challenge was working with Bolt.new, which definitely needs more refinement on the Next.js side of things. It felt less optimized for Next.js applications, leading to some workflow friction.

Despite these challenges, the overall development experience was incredibly rewarding and taught me valuable lessons about perseverance in complex full-stack development.

Accomplishments that we're proud of

I'm genuinely proud of the entire platform that emerged from this development journey. The final result exceeded even my own expectations - there were moments during development where I was genuinely shocked at how well certain features turned out. I approached this project with a "leave no stone unturned" mentality and went all in on every single aspect.

While I'm proud of every component I built, the global memory feature stands out as the crown jewel of the entire platform. This feature doesn't just enhance the user experience - it fundamentally transforms it. The global memory system makes the AI feel truly alive and aware, creating a sense of continuity and understanding that bridges every interaction. It's the difference between talking to a static chatbot and having a conversation with an AI companion that genuinely knows and remembers you.

Watching the global memory system work in practice - seeing how it connects a user's morning mood check-in with their evening therapy session, or how it personalizes meditation scripts based on their journal entries from weeks ago - that's when I knew the platform had achieved something special. It's like the app came alive and developed its own understanding of each user's unique journey.

What we learned

This project was an incredible learning journey that expanded my skills across multiple dimensions. I dove deep into conversational AI technology and significantly improved my efficiency as both a prompter and a "vibe coder." The experience taught me how to truly maximize the potential of AI tools and how a small idea can come to life in a matter of minutes when you know how to leverage these technologies effectively.

I also gained valuable insights into code standards and became much more proficient at resolving TypeScript errors - skills that will serve me well in future projects. However, I firmly believe that the experience is exponentially better when you have a solid coding foundation yourself. Having that knowledge allows you to be truly in sync with the AI tools you're working with, creating a harmonious development workflow.

I'd like to highlight how incredibly Bolt.new has evolved since I first tried it. I remember experimenting with it early on, and the improvements have been remarkable - it's taken a huge leap forward in capability and user experience.

Most importantly, I learned the true essence of building something genuinely valuable while writing minimal to no code. This project demonstrated the power of modern AI-assisted development and how it can democratize the creation of meaningful applications. Now that's what I truly love about this new era of development.

What's next for Mindful AI

First and foremost, I sincerely hope my solo effort gets the recognition it deserves and achieves its full potential. This project represents countless hours of passion and dedication, and I believe it has the power to make a real difference.

I have serious plans to scale MindfulAI into something much bigger because it genuinely has the applications and potential to help people - and I know it can. Mental health issues affect the majority of people, yet so many don't receive proper care due to economic barriers, lack of awareness, or simply not having enough time. MindfulAI solves all these problems: it offers a nominal subscription cost, provides help to anyone who wants it, and remembers each user's unique journey no matter when they return.

On the technical side, I plan to implement system-level optimizations and develop more cost-efficient AI usage patterns. I want to add additional personas and potentially even allow users to create their own personalized AI companions, making the experience even more tailored to individual needs.

Beyond these immediate plans, I have countless ideas brewing in my mind for features, improvements, and expansions such as scheduling, binaural beats, asmr, songs etc. all with AI workflows. The possibilities are endless, and I'm excited about the journey ahead. Let's hope for the best and see how far we can take this vision of accessible, personalized mental health support.

Built With

  • claude
  • convexauth
  • convexdb
  • framer
  • googlegemini
  • nextjs
  • openrouter
  • razorpay
  • shadcnui
  • tavusai
  • tiptapeditor
Share this project:

Updates