Ellocare

Project Description # H1 Ellocare is a specialized digital companion designed to bridge the generational gap and provide a "digital safety net" for seniors experiencing cognitive decline. By combining a family-curated knowledge base with the empathetic reasoning of the Gemini 3 Flash API, the app transforms static instructions (how to use a TV remote or who a specific family member is) into an interactive, conversational experience. It serves as a secondary, tireless memory for the elderly while offering caregivers a structured platform to manage daily checklists, health reminders, and vital safety information, ensuring that every precious memory and daily necessity is handled with dignity and care.

Inspiration ## H2 The heart of this project lies in personal observation. I watched my grandmother and mother struggle with the early stages of cognitive decline, where the digital world felt increasingly alien and daily tasks became sources of anxiety. I noticed a recurring friction: the elderly often feel like a "burden" for asking the same questions repeatedly, and family members, despite their love, can become exhausted. I wanted to build a "cross-generational bridge", an assistant that has the infinite patience of an AI but the personal context of a close family member.

What it does ## H3 Context-Aware AI Assistant: Seniors can ask questions like "Who is Tom?" or "How do I use the microwave?" and receive answers based specifically on data entered by their family. Visual Memory Box: A gallery for photos where users can annotate and "circle" important details (like a face or a specific object) to help with recognition. Life Navigation & Safety: Stores "Frequently Visited Places" with photos and one-click navigation, alongside "Pre-departure Checklists" (e.g., "Did I turn off the gas?") to reduce living hazards. Family Connectivity: A dedicated view for caregivers to set medicine reminders and a simplified "One-Click Video/Call" interface for the senior to reach their loved ones instantly. Dual-Role Interface: A secure management mode for family members to curate content and a simplified, high-contrast "Big Button" mode for seniors.

How we built it ## H4 Frontend: Built with React 19 and TypeScript for a robust, type-safe architecture. Styling: Tailwind CSS was used to create a "warm parchment" aesthetic that feels cozy rather than clinical. Icons: Lucide-React provided the clear, high-legibility iconography essential for senior accessibility. Intelligence: Integrated the Google Gemini API (gemini-3-flash-preview). We used a "RAG-lite" approach, where the user's curated memories and guides are injected into the system prompt to ensure the AI only speaks from personal truth. Deployment: Managed via Vite and Vercel, utilizing environment variables for secure API key management.

Challenges we ran into ## H5 UI for Seniors: Designing an interface that is "simple" without being "childish." We had to balance large touch targets with a sophisticated, respectful aesthetic. Context Pruning: Fitting a family's entire life history into an AI prompt without exceeding limits or confusing the model. We developed a "pruning" logic to only send relevant snippets. Environment Syncing: Ensuring the API_KEY was correctly injected into the Vite build process on Vercel while maintaining local development security. Image Annotation: Implementing a responsive HTML5 Canvas that allows users to draw and label photos on mobile devices smoothly.

Accomplishments that we're proud of ## H6 Patience-as-a-Service: Creating an AI persona that sounds like a "gentle grandchild", polite, patient, and clear. The Annotation Tool: Successfully building a feature where a son can take a photo of a remote and circle the "Power" button, which the senior then sees as a visual guide. Zero-Config Search: Making the search bar feel like a magic portal where any question about the user's life yields a correct, comforting answer.

What we learned ## H7 Inclusive Design: We learned that accessibility isn't just about font size; it's about reducing cognitive load and providing "safe" paths through an app. Prompt Engineering for Empathy: We discovered how much a system instruction can change the "warmth" of an AI response, which is crucial for users with dementia. The Power of Local Context: We realized that a "small" AI with the right personal data is more useful to a senior than a "large" AI that knows everything about the world but nothing about their family.

What's next for Ellocare ## H8 Gemini Live API Integration: Moving to real-time multi-language voice interaction so seniors can simply talk to the app without needing to type. Smart Wearable Sync: Connecting with smartwatches to track "Safe Zones" and automatically alert family if a senior wanders too far from a "Saved Location." Cloud Synchronization: Replacing LocalStorage with a real-time database (like Firebase) to allow multiple family members to update the memory box simultaneously from different cities. Photo Reconstruction: Using AI to "restore" old, blurry family photos, making it easier for seniors to reconnect with their past.

Built With

• canvas
• css3
• deep
• file
• gemini
• google
• gps
• html5
• javascript
• json
• lucide-react
• media
• react
• tailwind
• typescript
• vercel
• vite