Neurix.ai

• 

  Thumbhnail

• 

  Dashboard

• 

  Landing Page

• 

  Session History

• 

  Challenges

Neurix.ai uses the multimodal and long-context capabilities of Gemini (specifically gemini-3-flash-preview and gemini-2.0-flash) to build a mental wellness system that goes beyond short, disconnected chatbot conversations 🧠.

Most mental health tools treat each interaction as isolated. Neurix.ai is designed around continuity 🔄. It remembers past conversations, understands emotional patterns over time, and adapts its responses based on the user’s emotional history rather than just the current message.

At the center of the system are three main components.

The first is real-time emotional understanding 💬. During live conversations, the generateChatResponse agent analyzes the user’s messages and dynamically adjusts its tone using a therapy-style selector prompt. Depending on the detected emotional state, responses may be more supportive, reflective, or cognitively structured. These responses are delivered through a real-time video avatar interface powered by Tavus 🎥, which helps the interaction feel more natural and personal rather than text-based and mechanical.

The second component is the long-context narrative engine 📖. Instead of storing conversations as separate chat logs, Neurix.ai uses Gemini’s large context window to analyze weeks of session transcripts together. From this, the system identifies recurring emotional themes, behavioral patterns, and gradual changes in mindset. Over time, it builds a cohesive view of the user’s mental health journey, highlighting growth, setbacks, and increasing resilience 🌱. This allows users to see progress that would otherwise be invisible in day-to-day conversations.

The third component focuses on predictive wellness insights 📊. By analyzing historical emotional data, the system looks for patterns that often appear before high stress, emotional fatigue, or burnout ⚠️. When such patterns are detected, Neurix.ai can gently suggest coping strategies or bring attention to potential risks early, before they turn into serious issues. The goal is not diagnosis, but awareness and prevention.

Together, these components turn Neurix.ai into something closer to a long-term mental wellness companion 🤝 rather than a reactive chatbot.

The idea for Neurix.ai came from noticing two major gaps in current mental health support. First, many people struggling mentally are not taken seriously or are told to “just deal with it.” Second, most digital tools lack memory 🧩. They forget the user after each session, which makes genuine emotional understanding impossible.

We wanted to build a system that treats mental health as a continuous process ⏳. A system that remembers, learns, and responds with context — not just what the user feels now, but how they’ve been feeling over time.

While building Neurix.ai, we learned that long-context reasoning is essential for emotional applications 🧠. Without it, AI responses remain shallow. We also learned that latency matters a lot in emotionally sensitive interactions ⚡. Even small delays can break trust, especially when using a video-based interface. Designing prompts for mental wellness also required care — responses had to be empathetic and useful without sounding clinical or unsafe.

Technically, the project was built using React, TypeScript, Vite, Tailwind CSS, and Framer Motion on the frontend 💻, with Node.js, Express, MongoDB, and Mongoose on the backend. Gemini handles emotional analysis, long-term reasoning, narrative generation, and predictive insights, while Tavus provides the real-time video avatar interface. Authentication is handled using Supabase 🔐.

One of the main challenges was balancing response quality with performance ⚙️. To solve this, we used Gemini Flash models for live conversations where speed is critical, and deeper analysis processes for background tasks. Managing long-term emotional data also required careful summarization so that important context was preserved without overwhelming the model.

Overall, Neurix.ai represents an attempt to move mental health AI away from short-term conversations and toward long-term understanding, emotional continuity, and proactive support 🌍.

Built With

• express.js
• google-gemini
• mongodb
• mongoose
• node.js
• react
• supabase
• tailwind-css
• tavus-api
• typescript