SoulSpeak AI

Your AI-powered mental health companion offering empathetic conversations, private journaling, and mood tracking to support emotional well-being anytime, anywhere.

Comment

Inspiration

We built SoulSpeak to be a quiet, steady place where people can pour out their hearts and find a gentle, thoughtful response. It grew from a simple belief: that technology can carry warmth and safety when designed with care. This project is an invitation to anyone who needs a companion to listen, reflect, and help them notice the small steps of healing.

What it does

SoulSpeak is a single, cohesive experience that blends an empathetic AI companion, a private journaling space, mood tracking, and developer-grade observability into a product designed for real human healing.

Every user input receives an AI response. The AI companion is trained on therapeutic principles and large-scale empathetic dialogue so replies feel humane, validating, and actionable. Conversations are session-based by default to protect privacy, while still allowing the AI to use short-term context so reflections stay meaningful within a session. To guarantee availability during demos and real use, we run a robust key-rotation system across four Gemini API keys with intelligent cooldowns and automatic fallback to backup services when needed.

The journaling experience is a sacred, encrypted space. When a user writes, the platform stores an encrypted copy using AES-256-GCM; the system then produces a caring AI reflection tied to that entry. Tagging, search-ready metadata, and mood-linking help users discover patterns in their own stories while preserving strict data protection. Mood tracking uses a 1–10 scale with optional notes, streak tracking, and visual analytics so users can see their emotional arc over days, weeks, and months. AI-generated insights translate those signals into gentle, actionable takeaways that honor each person’s pace.

On the developer side, SoulSpeak includes a professional API testing dashboard and safe test modes so judges and engineers can validate functionality without consuming live quotas or altering user data. Health endpoints, synthetic checks, and live monitoring ensure the system is demo-ready and resilient under load.

How we built it

SoulSpeak is a modern full-stack serverless app crafted for speed, polish, and security. The frontend is React with TypeScript, Vite, TailwindCSS, and Framer Motion for a calm, accessible user interface. The backend runs as serverless functions on Netlify with Node.js for fast autoscaling. Supabase handles authentication while Neon PostgreSQL stores encrypted data with SSL and connection pooling. The AI layer uses Google Gemini as the primary model with a multi-key rotation and fallback strategy. Everything sensitive is encrypted end-to-end with AES-256-GCM and secured with token-based auth, CORS policy, and rate limiting. Lightweight observability and health checks keep the demo stage reliable and transparent.

Challenges we ran into

We navigated AI quota limits and remote service cold starts with an automated key-rotation and cooldown strategy that preserves session quality during peak demand. We balanced helpful, context-aware responses with strict privacy by giving the AI enough short-term context to be useful while encrypting and minimizing stored data. We tuned serverless latency and caching so the experience feels immediate and intimate even when the heavy work happens remotely.

Accomplishments that we're proud of

  • A compassionate AI companion that consistently responds to every user input with context-aware empathy.
  • End-to-end encrypted journaling workflow using AES-256-GCM for user confidence and safety.
  • A 4-key rotation and fallback strategy that keeps the AI available during demos and heavy usage.
  • Developer tooling for safe testing, monitoring, and reproducible demos that judges can run without risk.
  • A polished, accessible UI built for rapid iteration and emotional clarity.

What we learned

Careful prompt design and minimal, focused context produce the most humane responses. Session-first privacy increases trust, but excellent UX is required so users feel continuity without compromising safety. Lightweight monitoring and clear fallbacks are indispensable for demo environments.

What’s next

We will add multilingual empathy, deeper longitudinal context for long-term users, optional anonymous peer support, and PWA/native mobile apps for check-ins on the go. We will also offer an opt-in end-to-end encryption mode for users who want maximum control over their data.

Built With

Share this project:

Updates