symbiossis

AI therapy that adapts to you. Personalized conversations, local processing, privacy-first.

Comment

About the Project

Inspiration

Symbiossis emerged from a simple yet powerful belief: mental health support should be as accessible as your smartphone and as uniquely tailored as your most trusted therapist.

  • Traditional therapy is often out of reach for many, due to major barriers:
  • Geographic limitations restrict access for billions worldwide
  • Stigma prevents people from seeking the help they need
  • High costs place care out of reach for many
  • Lengthy wait times delay support for urgent situations
  • Concerns about privacy and data security in cloud-based solutions

My inspiration grew from seeing AI’s transformative potential to reshape healthcare delivery—opening access while preserving the authenticity and integrity of therapy. At the heart of this journey was the critical insight that the true value of therapy isn’t just in having a conversation, but lies in deeply understanding each person’s distinct psychological world.

What it does

Symbiossis isn't just another AI chatbot—it's a personalized therapeutic companion that understands you. Using advanced personality analysis and GPT-OSS models, it creates tailored mental health support that feels genuinely human and uniquely yours.

✨ How does it work?

Simply submit your biggest challenges in life, and you'll experience personalized therapeutic conversations adapted to your unique personality and communication style. Symbiossis works 24/7, for free, and gets better than traditional therapy through its innovative adaptation system.

🧠 Intelligent Therapeutic Analysis

The system analyzes your input to understand psychological patterns, identify underlying factors, and determine the most helpful therapeutic approach for your specific situation.

🎭 Personality-Driven Adaptation

Based on your therapeutic preferences, Symbiossis adapts its communication style, problem-solving approach, and therapeutic interventions to match how you process information and prefer to receive support.

🤖 GPT-OSS Integration

  • Dual-Mode Processing: 20B offline model + 120B online Ollama turbo fallback
  • Local-First Approach: Runs on your device by default for maximum privacy
  • Seamless Fallback: Automatically switches to cloud¹ processing when local models aren't available
  • Edge Runtime: Ultra-low latency responses with optimized performance

🛡️ Privacy-First Design

  • Zero Data Transmission: No data leaves your device by default
  • Local Profile Storage: User preferences stored in browser localStorage
  • GDPR Compliant: Full user control over data retention
  • Complete Offline Functionality: Works entirely without internet when possible

How I built it

I built Symbiossis using modern web technologies and AI frameworks. The frontend uses Next.js 15 with React 19 and TypeScript for a robust, type-safe application. For styling, I chose Tailwind CSS v4 with Framer Motion for smooth animations.

The AI integration leverages Ollama with GPT-OSS models, featuring a dual-processing architecture that runs locally first (20B parameters) and falls back to cloud processing (120B parameters) when needed. This ensures privacy while maintaining performance.

I use Legend State with automatic localStorage synchronization, allowing users to pick up conversations where they left off without any data transmission.

The user input is analysed and the answer is adapted based on individual communication preferences. Everything is designed to work completely offline when possible, with zero external data collection by default.

Challenges I ran into

Simple Therapeutic Adaptation: I built an easy and short system that adapts the therapist's communication style to each user's preferences. The system analyzes user input and automatically adjusts response style, tone, and therapeutic approach based on their individual communication patterns.

Local AI Performance: Running large language models locally while keeping the interface responsive required careful optimization. My dual-architecture approach with automatic fallback to cloud processing solved this elegantly.

Privacy vs. Personalization: Balancing deep personalization with absolute privacy was tricky. I achieved this through a local-first approach where all processing happens on the user's device by default, with encrypted cloud fallback only when necessary.

Accomplishments that I'm proud of

Privacy Innovation: I built a completely local-first therapeutic AI that processes everything on the user's device by default, with zero external data transmission. This maintains full GDPR compliance while delivering personalized therapeutic support.

Performance Excellence: I achieved sub-3-second response times in turbo mode with intelligent model switching, zero cold-start delays, and seamless fallback between local and cloud processing.

User Experience: I created an intuitive, accessible interface with smooth animations, responsive design across all devices, and a streamlined onboarding process that takes less than 2 minutes.

What I learned

Therapeutic AI requires careful balance: Privacy and personalization can work together when I design with local-first architecture. AI therapeutic quality depends heavily on rigorous validation systems and clinical oversight.

Local AI is the future of private AI: Running models locally provides unmatched privacy while maintaining therapeutic effectiveness. The key is intelligent fallback systems that maintain user experience.

Personality matters in therapy: Individual communication preferences significantly impact therapeutic outcomes. Adapting to how people prefer to process information creates more effective therapeutic relationships.

Trust is earned through transparency: Users need to understand how AI therapeutic systems work to build confidence. Clear communication about limitations and capabilities is essential for adoption.

What's next for symbiossis

Enhanced Personalization: I envision expanding my personality integration capabilities to create even deeper therapeutic connections through advanced session history analysis and adaptive learning.

GPT-OSS 2.0 Wait: I'm excited about the potential of GPT-OSS 2.0 and its multimodal capabilities that could open up entirely new dimensions for therapeutic AI interactions.

Smartphone Revolution: There's immense potential in expanding Symbiossis's mobile capabilities, making therapeutic AI even more accessible and integrated into daily life while maintaining my core privacy-first approach.

¹ Yes because we now live in a world where we can remotely use an IDE via SSH to connect to a locally purposed model... in the cloud. Thanks, Ollama Turbo.

Built With

  • aisdk
  • framermotion
  • gptoss
  • nextjs
  • ollama
  • tailwindcss
  • typescript
  • vercel
Share this project:

Updates