Soul-Support.ai

Imagine a trusted friend in your pocket—listening, learning, and acting to keep you emotionally balanced. That’s Soul‑Support.ai: an avatar‑led companion powered by Gemini and agentic AI.

Comment 2

Soul‑Support.ai Project Story

About the Project

Soul‑Support.ai is an avatar‑led AI companion designed to support mental wellness. By harnessing advanced generative AI and a robust agentic backbone, the platform listens to your feelings, analyzes sentiment and emotion intensity, stores this data, and proactively guides you through breathing exercises, journaling prompts, and personalized insights. When stress levels spike, it autonomously creates concise journal summaries—ensuring your emotional journey is safe, supported, and consistently evolving.

Inspiration

  • A Need for Always‑On Support
    Anxiety or low moods can strike at any moment, often when reaching out for help feels difficult. We envisioned an empathetic companion that’s available 24/7, ready to listen and respond with warmth.

  • Advances in Generative AI
    The rise of context‑aware models inspired us to build more than a chatbot—an AI that understands nuance, remembers history, and takes initiative on behalf of users.

  • Agentic AI Potential
    We wanted an AI that doesn’t just reply, but acts: nudging healthy habits, escalating in crisis, and summarizing long conversations into digestible journal entries.

What We Learned

  • Designing Precise AI Prompts
    Crafting instructions that reliably produce structured sentiment scores, emotion categories, and intensity metrics taught us the value of clarity and iterative refinement.

  • Implementing Agentic Behavior
    Building autonomous checks—when to nudge, when to escalate, when to summarize—reinforced the importance of ethical guardrails and user‑override options.

  • Building Full‑Stack Workflows
    Integrating backend logic, a NoSQL database, and a React‑based frontend with charts and interactive elements deepened our skills in end‑to‑end data flows.

How We Built It

  1. Emotion Analysis & Storage

    • Incoming user messages are sent to the AI for sentiment and emotion analysis.
    • Results (score, category, intensity, timestamp) are persisted in a database.
    • When intensity crosses a threshold, the AI processes the entire conversation history into a brief journal summary and saves it.

  2. Conversational Companion

    • Users chat with the avatar‑powered assistant, receiving supportive responses and micro‑tasks (breathing exercises, gratitude prompts) based on their current state.

  3. Insights & Summaries

    • On demand or when high intensity is detected, the AI generates two‑sentence insights or 2–3 sentence journal entries, offering reflections and gentle guidance.

  4. Dashboard & Visualization

    • A web interface displays interactive charts of emotion frequency and intensity over time.
    • An emoji grid highlights which feelings have been experienced in a given period.
    • A dedicated journal view surfaces AI‑written summaries as styled cards.

Challenges Faced

  • Ensuring Consistent AI Output
    Getting the model to return valid JSON every time required careful prompt engineering and cleaning of extraneous formatting.

  • Balancing Proactivity with Sensitivity
    Deciding when the AI should take initiative—offering tips, escalating to crisis resources, or summarizing chats—meant setting thoughtful thresholds and always giving users control.

  • Real‑Time Data Handling
    Optimizing data fetching and state updates in the dashboard ensured charts stay smooth and up-to-date without overwhelming the UI.

  • Designing for Empathy
    Choosing language, colors, and interactions that feel comforting (rather than clinical) involved user feedback loops and accessibility checks to create a true “friend in your pocket.”

From initial concept to an operational agentic companion, Soul‑Support.ai embodies how AI can actively care—listening, learning, and acting to foster better mental wellness.

Built With

Share this project:

Updates