Stamina AI - AI-Powered Mental Health Companion

Technologies:

AI / LLM (3)
Flutter (4)
PostgreSQL (27)
Auto Testing (25)
Django (23)
Linux (23)
Vue.js / Nuxt.js (12)

Domains:

AI Solutions (6)
Mobile Applications (4)
Personal Development (4)

Project Goals

The Stamina AI project aimed to develop a comprehensive mental health support application that leverages AI-driven solutions to help users cope with issues like burnout, anxiety, and procrastination. The application provides personalized guidance based on Large Language Models (LLMs), delivering evidence-based psychological support that is accessible anytime, anywhere.

Functional Capabilities

  • Personalized Mental Health Support: Users can receive personalized sessions based on their unique needs, with support for over 1000 customized sessions tailored to their mental health journey.
  • AI-Backed Therapy: The application is designed using LLMs, providing real-time conversational therapy that is comparable to working with a professional psychologist.
  • Mobile-First Development: The application is available both as a mobile application developed using Flutter and as a web application, ensuring that users can access support on the go.
  • Integration of Multiple LLM Models: The ExoChat engine powers Stamina AI, addressing challenges such as loss of context and multi-model task usage to ensure consistent, high-quality conversational experiences.
  • Dynamic Content Loading and Offline Functionality: The app supports dynamically loaded content, including paid subscriptions, and offers offline functionality with automatic synchronization upon reconnection.
  • Evidence-Based Techniques: The application incorporates Cognitive Behavioral Therapy (CBT) techniques to provide effective intervention for various mental health issues, promoting behavioral change by targeting thought patterns.

Solution Concept

The Stamina AI mental health companion was developed to provide users with accessible, personalized mental health support based on cutting-edge AI technology. By combining psychological expertise with the flexibility of a mobile application, Stamina AI serves as a digital ally for self-improvement and emotional well-being.

A significant challenge during development was to address issues related to AI context loss, slow iteration, and the complexity of utilizing multiple models for different functions. To overcome these challenges, the development of ExoChat—a specialized engine for creating LLM-based products—enabled the creation of Stamina AI and positioned it as a foundation for future solutions in the LLM space.

The collaboration between psychologists, prompt engineers, developers, and AI specialists was essential to ensure the product's efficacy and user satisfaction. This interdisciplinary approach led to the development of a solution that not only aids users in their personal development but also maintains a high level of security and confidentiality.

The back-end is developed using Python (Django), while PostgreSQL is used for database management. Flutter was chosen for the mobile front-end, and Vue.js was used for the web interface. The application also integrates with popular AI APIs like OpenAI and Gemini API.

Results

  • Successful Product Launch: The Stamina AI application has been launched and continues to serve users as an AI-powered mental health companion, providing reliable support for dealing with burnout, procrastination, and other challenges.
  • Increased User Engagement: Users have reported higher levels of engagement, thanks to the app's personalized approach to mental health support and the use of evidence-based techniques like CBT.
  • Improved Efficiency in Development: The development of ExoChat as a separate product not only enhanced the Stamina AI experience but also enabled faster iteration and easier integration of multiple LLMs for different use cases.
  • Revenue Generation: The app has become a successful source of revenue, offering users both free and subscription-based services, making professional mental health support accessible at a fraction of the cost of traditional therapy.

Technologies and Architecture

  • Back-End Development:
    • Python (Django): Utilized for server-side logic and managing requests, ensuring the secure and efficient handling of data.
    • PostgreSQL: Used for storing user data, subscription information, and session logs, with a focus on maintaining data integrity and security.
  • Front-End Development:
    • Flutter Mobile: Used for developing a responsive mobile application with offline functionality, providing a consistent experience across devices.
    • Vue.js: Implemented for the web application to provide users with easy access to mental health resources.
  • AI Integration:
    • OpenAI API: Provides the language model that powers real-time conversational therapy.
    • Gemini API and Llama: Additional LLMs used for various purposes, allowing the application to offer nuanced responses and handle different types of user interactions.
  • Data Synchronization and Security:
    • REST APIs: Integrated for smooth communication between front-end and back-end components, ensuring secure data transfer.
    • ExoChat Engine: Custom-built engine to overcome LLM-related challenges, manage context, and facilitate seamless use of multiple models.

User Cases

  • Users Seeking Psychological Support: Users can interact with Stamina AI for personalized mental health sessions, helping them to manage stress, burnout, and procrastination.
  • Mental Health Professionals: Professional psychologists can use Stamina AI as an auxiliary tool to support their clients, offering a digital extension to their regular sessions.
  • AI Developers: The development of ExoChat has opened opportunities for other developers to create products utilizing LLMs, paving the way for advancements in AI-based conversational tools.

Integration and Development Process

  • Research and Requirements Gathering: The development began with extensive research and consultations with psychologists to identify the best approaches for implementing AI therapy. Challenges related to LLMs, such as context retention and multi-model integration, were analyzed.
  • Team Formation: A cross-functional team including psychologists, prompt engineers, developers, and AI specialists was formed to ensure that both the technical and therapeutic aspects of the project were effectively addressed.
  • Iterative Development: Agile Scrum methodology was adopted, allowing continuous testing and feedback, particularly in refining the LLM interactions and improving response quality.
  • Launch and Iteration: After launching the product, feedback from users was collected, leading to further iterations and enhancements of the application to improve usability, response accuracy, and overall user satisfaction.

Client Benefits

  • Affordable Mental Health Support: Stamina AI offers an affordable alternative to traditional therapy, providing users with the benefits of CBT and AI-guided sessions at a fraction of the cost.
  • 24/7 Availability: Unlike traditional mental health services, Stamina AI is available 24/7, giving users the flexibility to seek help whenever they need it, improving accessibility and convenience.
  • Increased User Empowerment: By providing personalized sessions and real-time guidance, Stamina AI empowers users to take proactive steps towards improving their mental health, reducing the stigma associated with seeking help.
  • Scalable Solution: The development of ExoChat not only supported the Stamina AI application but also allowed for scaling the solution to support additional AI-based mental health initiatives in the future.

Technologies Used

  • Backend: Python (Django), PostgreSQL
  • Frontend: Flutter (Mobile), Vue.js (Web)
  • AI Integration: OpenAI API, Gemini API, Llama
  • Other Technologies: Linux, REST, Docker, Kanban