When your mind feels heavy, you don’t have to carry it alone ♡,Mindora : AI Mental Wellness Companion

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

Meet Mindora — AI Mental Wellness Companion, a voice-first, privacy-focused daily emotional reflection and stress-decompression web app powered by local Google Gemma (via Ollama) and ElevenLabs voice synthesis.

The Problem It Solves

Students, freelancers, and remote workers frequently experience late-night burnout, academic overwhelm, and decision paralysis. However:

1. Traditional therapy is expensive and not always accessible at 2:00 AM when exam or deadline anxiety hits.
2. Generic cloud AI chatbots often feel cold, dump overwhelming walls of text, or raise serious privacy concerns when users share personal struggles.

Mindora bridges the gap as a non-clinical mental-wellness companion (explicitly designed not to replace a psychiatrist or diagnose disorders). Instead of a blank chat box, Mindora guides users through a structured, soothing 4-Step Decompression Journey in a warm pastel-green and cream interface:

Step	Stage Name	What Happens	User Value
Step 1	Welcome & Setup	User selects language (English, Hindi, or Hinglish) and mode (Voice + Text or Quiet Text).	Establishes clear non-clinical boundaries & 1-click session privacy.
Step 2	Emotional Check-In	6 gentle, skippable cards assessing Mood, Stress Frequency, Triggers, Sleep, Daily Impact, and Support Goal.	Produces a qualitative reflection (never fake clinical stress scores).
Step 3	Gemma + ElevenLabs Voice Session	User speaks or types their concern. Gemma crafts a warm, grounded reply and ElevenLabs speaks it aloud via a pulsing Voice Orb.	Validates feelings, asks one gentle question at a time, and avoids text overload.
Step 4	Closure & Grounding	Interactive 4-4-4 Box Breathing Pacer, 1 manageable next step, and an optional closing journal prompt.	Users leave the session calmer with a single concrete action item.

Who Would Pay For It? (The Side Hustle & Monetization Model)

Mindora is designed as a sustainable B2C Freemium & B2B Campus/Workspace Wellness Micro-SaaS:

Tier	Target Audience	Pricing	Key Features Included
Free Tier	Students & Everyday Users	$0 / mo	Daily 6-card emotional check-in, local Gemma reflection, browser TTS voice, & 4-4-4 box breathing tool.
Mindora Calm Pro	High-Stress Professionals & Creators	$7 / mo	Unlimited ElevenLabs ultra-realistic soothing voices, multilingual voice reflection (English/Hindi/Hinglish), & private encrypted local journal exports.
Mindora Campus / Team	Universities, Bootcamps & Remote Startups	$199 / mo	Anonymous campus/team stress-trigger heatmaps (zero personal chat logs stored) + custom local helpline integrations during exam/launch weeks.

---

Demo

• Live Demo / Video Walkthrough:

Key Screens & Experience

• Soothing Non-Clinical Aesthetic: Designed with Alabaster Cream (#FAF8F5) and Sage Green (#6B9080) to avoid the cold, sterile feel of hospital-white apps.
• Built-In Crisis Safety Pathway: Before any message reaches the AI model, FastAPI screens for urgent distress or self-harm indicators. If detected, normal AI advice pauses immediately and surfaces a 24/7 Human Crisis Helpline Banner (including India's Tele-MANAS 14416, Vandrevala Foundation, and 988).

🌿 How to Use Mindora (Visual Representation)

Interface	Steps
	Step 1: Welcome & Preferences
	Step 2: Emotional Check-In
	Step 3: Share & Voice Companion
	Step 4: Summary & Relaxation

🌿 How to Use Mindora (Quick User Guide)

Step / Feature	What You Do	How Mindora Helps Your Feelings
Step 1: Welcome & Preferences	Select your language (English, Hindi, Hinglish) and mode (Voice + Text or Quiet Text), then click Begin Emotional Check-In.	Sets a safe, private, non-clinical space in the language and style you feel most comfortable with.
Step 2: Emotional Check-In	Tap through 6 skippable cards (Mood, Stress Frequency, Triggers, Sleep, Daily Impact, and Support Goal).	Helps you name your emotions and what's bothering you without forcing you to type a long paragraph.
Step 3: Share & Voice Companion	Type, click a starter chip, or tap the Mic (🎙️) to share what's on your mind, then click Wrap Up Session → when ready.	Gemma validates your feelings and suggests one small strategy at a time while ElevenLabs speaks it aloud in a calm voice.
Step 4: Summary & Relaxation	Read your Gentle Session Reflection, check off 1 Manageable Next Step, and optionally write in the Closing Journal Prompt.	Turns overwhelming stress into one tiny, realistic action item so you leave with a clearer mind.
Right Panel: 4-4-4 Calm Breathing	Click Start/Pause Breathing Exercise and follow the expanding circle (Inhale 4s → Hold 4s → Exhale 4s).	Physically slows your heart rate and grounds you at any moment during your session.
Top & Left Controls: Privacy & Safety	Click Clear Session to wipe your data instantly, or click Need Urgent Helpline Numbers? if in severe distress.	Gives you 1-click privacy control and immediate access to 24/7 human crisis helplines.

---

Code

AdibaNaushad / mindora_health

🌿 Mindora — AI Mental Wellness Companion

A comforting, voice-first AI reflection space powered by Google Gemma (via Ollama) and ElevenLabs.

---

✨ Overview

Mindora is a supportive AI mental-wellness companion designed to help people unpack everyday stress—such as academic pressure, career uncertainty, relationship friction, or low confidence—through a gentle, structured journey and comforting spoken voice conversation.

Instead of acting like a cold chatbot or making clinical diagnoses, Mindora guides users through a complete 4-step emotional grounding flow:

1. Welcome & Session Preferences (Language & Voice/Text mode)
2. Non-Diagnostic Emotional Check-In (6 skippable reflection cards)
3. Empathetic Voice Conversation (Powered by Gemma for reasoning + ElevenLabs for realistic human speech)
4. Session Closure & Grounding (Interactive 4-4-4 box breathing, 1 manageable next step, and a closing journal prompt)

---

🛡️ Ethical Boundaries & Safety Guardrails

Mindora is intentionally built as a mental-wellness companion, NOT a psychiatrist, therapist, or medical device.

• No Fake Clinical Scores: Instead…

View on GitHub

(Or view the repository directly at: https://github.com/AdibaNaushad/mindora_health.git)

---

How I Built It

Mindora is built around a clean, modular full-stack architecture where every technology has one clear responsibility:

Layer	Technology Used	Role in Mindora
AI Brain (Open-Weight LLM)	Google Gemma (gemma2:2b) via Ollama	Runs locally to interpret the user's 6 check-in answers + chat history, validate emotions, and suggest 1 small, realistic action step.
AI Voice (Text-to-Speech)	ElevenLabs API (eleven_flash_v2_5)	Converts Gemma's written empathy into warm, human-like spoken audio with ~75ms ultra-low latency (with Web Speech API fallback).
Backend Coordinator	Python + FastAPI + HTTPX + Pydantic	Exposes /session/check-in and /session/chat, runs crisis keyword screening, formats Gemma system prompts, and keeps API keys secure.
Frontend Interface	React 18 + Vite + Tailwind CSS + Lucide	Modular component architecture (WelcomeStep, CheckInStep, ChatStep, SummaryStep, SidebarRight breathing orb).

How Gemma & ElevenLabs Work Together

1. Context-Rich Check-In (/session/check-in): When a user finishes the 6 check-in cards (e.g., Mood: Overwhelmed, Trigger: Studies & Exams, Sleep: Staying up too late, Goal: Help breaking down a problem), FastAPI packages these parameters into a structured context header for Gemma.
2. Empathetic Prompt Engineering (/session/chat): Gemma is guided by a strict system prompt:
   • Never diagnose medical conditions or act like a doctor.
   • Keep responses concise (2–4 warm sentences) so they sound natural when spoken aloud.
   • Validate the user's feelings first, then ask only ONE gentle follow-up question or offer one tiny next step.
3. Instant Spoken Comfort: FastAPI sends Gemma's response text directly to the ElevenLabs TTS endpoint (eleven_flash_v2_5 with voice EXAVITQu4vr4xnSDxMaL), base64-encodes the audio stream, and returns both text and audio to React so the Voice Orb speaks immediately.

---

Why Does Open Innovation Matter?

1. Having Someone to Listen When No One Else Is Around

Mental health struggles rarely follow a convenient 9-to-5 schedule. The heaviest emotional friction—exam panic, deadline burnout, relationship heartache, or the quiet ache of loneliness—almost always arrives at 2:00 AM when everyone else is asleep and there is nobody to call.

In those moments, staring at a traditional, silent AI chatbot feels transactional and clinical. It feels like talking into an empty void.

By combining Google Gemma's empathetic reasoning with ElevenLabs' natural voice synthesis, Mindora shifts the entire dynamic. It transforms a cold text window into a presence. When someone is spiraling, hearing a calm, soft voice acknowledge their pain ("I hear how exhausted you are right now, and it makes complete sense that this feels heavy") activates our natural auditory grounding system. It no longer feels like a synthetic chat; it feels like having a compassionate, non-judgmental companion who is genuinely listening, understanding, and gently speaking back. That auditory warmth breaks the paralysis of isolation when someone needs it most.

When building a mental-wellness application, trust and privacy are everything. Users will not honestly share their academic fears, family conflicts, or insecurities if they worry their most vulnerable thoughts are being logged to train a proprietary corporate model.

Open innovation—specifically Google's open-weight Gemma models running locally via Ollama—made three critical things possible that a closed API alone could not:

1. Local-First Emotional Privacy: The core reasoning engine (gemma2:2b) runs 100% locally on the user's or campus's own hardware with zero third-party data retention.
2. Zero-Cost Experimentation & Accessibility: Students and indie builders can run unlimited empathetic reflection sessions locally without worrying about per-token LLM billing spiking during exam season.
3. Predictable Safety Guardrails: By pairing an open-weight local model with a transparent Python FastAPI safety layer, we have deterministic control over crisis keyword interception and system instructions.

---

My First User

I tested Mindora with fellow university students dealing with late-night study pressure and project deadlines.

• What stood out to them: They loved that Step 2 (the 6-card check-in) meant they didn't have to stare at a blank chat box wondering how to start explaining their stress—clicking "Overwhelmed", "Studies & Exams", and "Staying up too late" let Gemma immediately greet them with a spot-on reflection.
• The Voice Difference: Hearing the response spoken aloud in a calm voice while watching the 4-4-4 breathing orb on the right panel made the app feel like a grounding companion rather than another text-heavy productivity tool.

---

Prize Categories

• Best Use of Gemma: Mindora uses Google's open-weight Gemma (gemma2:2b) model served locally via Ollama as its core empathetic reasoning engine—powering both the personalized check-in reflection (/session/check-in) and context-aware, non-diagnostic wellness conversations (/session/chat).
  Because it runs locally, users can safely share their heaviest mental health struggles without fear of privacy loss.

• Best Use of ElevenLabs: Mindora gives our local Gemma wellness companion a warm, natural human voice using the ElevenLabs Text-to-Speech API (eleven_flash_v2_5), transforming written reflections into comforting spoken guidance complete with an interactive Voice Orb and Play/Pause/Replay controls.
  This is the core of how Mindora helps with loneliness—instead of reading a cold chat window at 2:00 AM, users hear a soothing, empathetic voice that provides immediate auditory grounding and makes them feel truly understood when no one else is around.