The AI Enhanced Therapeutic System - Zax

Zax is an AI designed to support mental well-being. Engage in meaningful conversations and reimagine the hurdles life throws at you. Our primary goal is to ensure therapy is accessible to everyone.

Comment

Inspiration

Coming from the South Asian community, we are aware of the stigma that often surrounds therapy. One incident that stands out is the story of a close friend. The unnamed individual was deeply in need of therapy, yet they were scolded by their parents for even suggesting such an idea. This resulted in unnecessary self-deterioration and self-harm. Today, they are in a better place, but their experience planted a seed in our minds. How could we bridge the gap and ensure therapy is accessible for those who need it most? Thus, Zax was born—a statement to our commitment to making mental health more accessible.

What it does

In essence, Zax processes user speech input, interprets it, and delivers spoken responses. Zax's design is tailored to grant accessibility to communities like the visually impaired and individuals with typing limitations. Additionally, Zax currently supports all Latin-based languages, enhancing its reach and impact. Not to mention it is completely online, so anyone from anywhere can access it, as they please.

How we built it

AI Therapist For the AI therapist segment, we used OpenAI's GPT-3.5 model. Then through prompt engineering, we fine-tuned the AI to act like a therapist. Our speech input mechanism records a 10-second WAV file, which is later transcribed via OpenAI's Whisper model. For speech output, we used Google Cloud's Text-to-Speech API, complemented by Pygame for audio playback. To wrap it up, the conversation history is stored in a list, guaranteeing continuity of interactions.

Website In a 10-12 hour sprint, I created a website using Python and Streamlit. From planning the layout to integrating data and user features, each step unfolded swiftly: Concept to Blueprint: I sketched the website's structure and goals. Streamlit Immersion: I swiftly learned Streamlit's essentials from tutorials. Framework Formation: I shaped the website's layout with Python. Data Alchemy: I weaved the dataset into interactive visuals using Python. Engagement Enhancements: I introduced user-centric features like feedback. Final Flourishes: I polished visuals, ensuring cross-device functionality. The result? A Python-powered site in 10-12 hours, a testament to Streamlit's speed and adaptability for web creation.

Challenges we ran into

The time constraint of 36 hours emerged as a huge issue. The both of us both worked in our respective jobs during the hackathon, so this led to further issues. Therefore, some of our original ideas had to be compromised or appended to fit the tight schedule. Nevertheless, we still finished the AI segment and the website, which was a massive accomplishment, considering our limited experience with Python.

Another issue that came up was web development. As people versed in HTML, JavaScript and C#, coding a website in Python was out of our comfort zone. Streamlit's potential to transform data scripts into interactive web applications was interesting. To overcome this challenge, we had to immerse ourselves in various YouTube tutorials, Github pages, and the list goes on.

Accomplishments that we're proud of

Undoubtedly, one of our most significant achievements stands as our ability to showcase our innovation. This hackathon marked our debut, and along the journey, there were moments when all hope was lost. However, emerging triumphant from the various challenges shows our growth throughout the hackathon. To add, another achievement is the fact that we became better python users. This is very important to us, given Python's prominence within the modern job market. Mastering the intricacies of the versatile programming language is a feat in itself, demonstrating our journey of continuous learning and progress.

What's next for The AI Enhanced Therapeutic System - Zax

One of our goals is to create a 3D model for Zax, nurturing a stronger connection with the therapist. This new update would seek to enhance the engagement between the user and the AI. Furthermore, we have aspirations on transforming the website into a WIFI less accessible app. We aspire to make therapy more inclusive, this includes parts of the world that may not always have access to WIFI, such as our communities in South Asia. To end off our list of potential updates, we also want more professional input from licensed mental health workers in an attempt to elevate Zax's output.

Built With

  • google-cloud-texttospeech
  • openai
  • pycharm
  • python
  • streamlit
Share this project:

Updates