TherapyAI

TherapyAI helps you stay present in every session—capturing conversations, revealing patterns and summarizing therapy sessions—all while guiding your personal growth with mindful journaling prompts.

Comment

Our Project

Inspiration

We were inspired by the pressing need for better mental health support and the experience of wanting to reference topics and advice from previous therapy sessions. Patients don't often take note on what their therapists say, and therapists can forget details when managing 20+ patients on a weekly basis. We knew there had to be a better way to help all parties involved. And know there is!

How We Built It

We used Next.js for both the backend and frontend, allowing our team to keep development within a single framework for API routes and UI components. Firestore serves as our NoSQL database, providing flexibility for storing session details, messages, and summary data. We integrated the LiveKit API to enable video and audio sessions, which can be recorded during therapy. Recorded MP4 sessions are stored securely in an AWS S3 bucket. We utilize the Whisper API (OpenAI) to convert recorded MP4 sessions into text. GPT-4 then summarizes these transcripts, generating concise key points for easy reference.

Features

Patients and therapists have unique dashboards, tailored to the needs of the respective parties. Patients can see the AI-generated information from all of their previous sessions, manage and fill out the associated journaling prompts, schedule appointments, message their therapist, see emerging patterns from their previous sessions, and of course, join video sessions with their therapist.

Challenges

  • Dashboard Implementation
    Presenting messages, session recordings, and summarized data in a clean UI was tricky. We spent significant time ensuring each piece of data was properly associated with the correct patient or therapist.
  • Associating Data
    Mapping each session’s notes and transcripts to the correct user profiles required careful database structuring and security rules.
  • API Keys & Integration
    Coordinating multiple APIs (LiveKit, Whisper, GPT-4, Firebase) under secure and consistent configurations took the most time. The end-to-end pipeline—from video recording to transcription to summary—was especially challenging.

Accomplishments That We're Proud Of

  • Seamless Multi-Service Integration
    We successfully combined LiveKit, Firestore, S3, Whisper, and GPT-4 into one functioning ecosystem within a tight hackathon timeframe.
  • User-Centric Design
    We focused on creating an intuitive dashboard that makes it easy for therapists and patients to track sessions, messages, and summaries without feeling overwhelmed.
  • Real-Time & Asynchronous Support
    Balancing real-time video and messaging with asynchronous tasks such as transcription and summarization was a big technical win for us.

What We Learned

  • Balancing Privacy & Transparency
    We discovered the importance of clearly communicating how notes are recorded and used so that both therapist and patient feel comfortable.
  • Handling Diverse Data Flows
    Designing for real-time interactions (video, chat) alongside background processes (transcriptions, summarizations) gave us new insights into system architecture.
  • Prioritizing UI/UX
    We learned that even the most powerful features can feel daunting if they’re not presented in a straightforward, friendly way.
  • Database Structuring
    Ensuring that all data—notes, transcripts, session details, user profiles—linked together without conflicts demanded thoughtful design from the ground up.

What's Next for TherapyAI

  • Enhanced Analytics
    We plan to integrate more robust data analytics for therapists to measure session progress and outcomes over time.
  • More AI Capabilities
    From deeper sentiment analysis to personalized follow-up recommendations, there’s huge potential to leverage AI in further refining therapy insights.
  • Expanded Customization
    Allowing therapists to customize the summarization parameters—e.g., focus on certain keywords or topics—could make sessions even more tailored.
  • Mobile Application
    Building a dedicated mobile app would improve accessibility and convenience, letting users participate from anywhere.

Built With

Share this project:

Updates