MEMORA AI

MEMORA AI is your intelligent voice memory assistant that remembers important details, retrieves personal memories, and enables natural conversations in Tamil, Tanglish, and English using Gemini AI.

Comment

About MEMORA AI

Inspiration 💡

We often forget important details from our daily conversations, personal goals, tasks, and plans. Most AI assistants can answer questions, but remembering personal information and recalling it naturally across conversations can be challenging.

This inspired us to build MEMORA AI, an intelligent memory assistant designed to help people remember what matters. We wanted to create an assistant that feels more personal and accessible by supporting Tamil, Tanglish, and English.

What It Does 🚀

MEMORA AI combines AI-powered conversations with persistent personal memory.

  • Persistent Memory: Saves important information using an SQLite database.
  • Intelligent Memory Retrieval: Retrieves relevant saved information when users ask questions.
  • Multilingual Conversations: Supports Tamil, Tanglish, and Indian English.
  • Gemini AI Integration: Uses Google's Gemini AI to generate contextual responses.
  • Voice-Friendly Experience: Provides a conversational interface with voice-related features.
  • Personalized Assistance: Uses relevant stored memories to make responses more useful and personalized.

Our goal is to make AI conversations more continuous, natural, and personal instead of treating every interaction as an isolated question.

How We Built It 🛠️

We developed MEMORA AI using a full-stack architecture:

  • Frontend: React and Vite for the interactive user interface.
  • Backend: Python and FastAPI for API handling and application logic.
  • Database: SQLite for persistent memory storage.
  • AI Model: Google Gemini (gemini-2.5-flash) for AI-powered responses.
  • Languages: Tamil, Tanglish, and English.

The frontend communicates with the FastAPI backend through API endpoints. The backend manages memory storage and retrieval, prepares relevant context, and integrates with Gemini to generate responses.

Challenges We Faced 🧩

One of our main challenges was connecting persistent memory with meaningful AI responses. Saving a message in a database is not enough; the assistant must retrieve the correct information when the user asks a related question.

We also worked on configuring the Gemini API, handling backend and frontend communication, supporting multiple languages, and ensuring that memory remains available beyond a single interaction.

Another important challenge was distinguishing a genuine AI-generated response from a local fallback response. This highlighted the importance of reliable API integration, clear status indicators, and proper error handling.

What We Learned 📚

Building MEMORA AI helped us understand how full-stack applications, databases, REST APIs, and generative AI can work together.

We learned about persistent data storage, retrieving relevant context, integrating an external AI model, managing environment variables securely, and designing a multilingual conversational experience.

Most importantly, we learned that a useful AI assistant needs more than intelligent responses. It also needs reliable memory, relevant context, and a trustworthy user experience.

What's Next? 🌟

We plan to improve MEMORA AI with more accurate memory retrieval, better conversational context, stronger Tamil and Tanglish understanding, improved voice interactions, and more reliable AI error handling.

Our long-term vision is to make MEMORA AI a personal assistant that helps people preserve important information and recall it naturally whenever they need it.

MEMORA AI — Your Memory. Our Priority.

Built With

Share this project:

Updates

Submission history